• YouTube's ad revenue fell short of analysts' expectations for the third quarter.
• Its parent reported $11.2 billion in Q3 ads revenue, under analysts' $11.43 billion estimate.
• The company said it now has 150 million YouTube Music and Premium subscribers, up from 125 million in March 2025.

YouTube's ad revenue growth slowed down, missing analysts' estimates in the third quarter.

Its parent company, Alphabet, reported $11.2 billion in third-quarter YouTube ads revenue, a 9% increase from a year ago. That falls below the $11.43 billion analyst consensus estimate. Its ads revenue growth also slowed from the previous quarter, when it pulled in about $11.1 billion, a 13% increase from the previous year.

Beyond YouTube, Alphabet missed expectations in other areas of its business, including on overall revenue. In a bright spot, it beat revenue estimates for its cloud business.

The company's shares were down over 10% in after-hours trading.

While YouTube's ad revenue was weaker than expected, it wasn't all bad news. The company said it now has over 150 million YouTube Music and Premium subscribers, up from 125 million in March 2025. The company said its subscriptions business was growing faster than ads.

YouTube has also been the envy of rivals in key areas like TV viewership. Its share of US television viewership was about twice that of Netflix in August, according to recent Nielsen estimates. Over a third of YouTube's third-quarter ad revenue came from the living room, Google's chief business officer Philipp Schindler said on Wednesday.

Netflix, meanwhile, walked back its full-year ad revenue expectations during last week's earnings. The streamer lowered its ad revenue forecast from $3 billion to $2.7 billion after experiencing delays around the rollout of its ad products. It laid off about 1,000 workers earlier this month.

This week, Instagram unveiled an overhaul of long-form video features seemingly designed to counter YouTube and bolster its presence on TV screens.

"Everyone in streaming is chasing YouTube, and even Instagram is going to pivot their strategy based on the success that YouTube's had," said Ross Benes, a senior analyst at EMARKETER, Business Insider's sister company.

• Microsoft is offering buyouts to about a quarter of its workforce.
• In an internal memo, Microsoft's finance boss said the company expects "very few people" to accept.
• Employees had been questioning the rationale behind the offer, the memo said.

Microsoft expects 3,000 or fewer employees to accept the buyout it offered this week to more than a quarter of its workforce, a spokesperson told Business Insider.

The point was reiterated on Wednesday by Chief Financial Officer Amy Hood, who wrote in an internal memo that employees have been questioning the rationale behind the buyout.

"I know there have been some questions on the rationale behind the voluntary exit offer that we announced this week," Hood wrote. "Our intent is to offer employees agency if they choose to leave the company on their own terms. We expect very few people to select this offer because of the growth opportunities here at Microsoft."

The memo statement was part of a quarterly earnings email Hood typically sends internally to employees every quarter when Microsoft discloses its financials.

These memos mostly repeat what the company reports publicly, but this time Hood went out of her way to address the buyout offer, which caused quite a stir across the tech industry this week.

Earlier this year, Microsoft offered a different retirement buyout. In that situation, about one-third of eligible employees took the offer, and that helped Microsoft cut fewer employees in subsequent layoffs.

This week's buyout is getting a lot less uptake, according to the company.

Read the CFO's full memo

Team,

This afternoon, we announced our first-quarter financial results. We exceeded Wall Street expectations, growing revenue 19% and operating income 21%.

With the start of FY27, we also made an important change to how we report our business. We announced that we changed our reporting structure for the first time in 12 years, moving from three segments to two: Agents and Infra, and Devices and Consumer. This change better reflects how we are running the business today — from how we build products and allocate resources to how we create value for customers across the connected layers of our platform. To give you a sense of the two segments, looking back to last fiscal year, Agents and Infra would have represented about 81% of our revenue while Devices and Consumer would have been 19%.

A few highlights from the quarter across our two new segments:

• Agents and Infra grew revenue 23% this quarter
• Microsoft Cloud revenue was $63.9 billion and accelerated to 30%.
  Azure revenue increased 48% reflecting accelerating Foundry demand.
• Microsoft 365 commercial cloud revenue increased 18% in constant currency driven by growth in our premium seat offerings like Copilot and E7 as well as usage-based services across M365 Copilot and GitHub Copilot.
• Microsoft 365 commercial seats increased 5% and the number of customers with more than 50,000 M365 Copilot seats increased 8x year-over-year.
• Microsoft 365 consumer cloud revenue increased 17% with subscriber growth of 7%.
• Industry solutions cloud revenue increased 10% with growth in Dynamics 365 and LinkedIn Talent Solutions.
• Devices and Consumer declined 3% this quarter
• Search and advertising revenue ex-TAC increased 12%, with growth across LinkedIn, Bing, and Edge.
• Windows OEM and devices revenue decreased 22%, reflecting the prior-year benefit from Windows 10 end of support, inventory drawdowns by OEM partners, and weaker PC market demand.
  XBOX content and services revenue decreased 2%. Last week, we launched Call of Duty: Modern Warfare 4, as we continue to bring great games to more players across more devices.

Thank you for everything you did to deliver a strong start to the year. In particular, I want to thank our engineering and capacity delivery teams who are working to help us deliver more capacity faster and with high reliability. These teams are also using the capacity we have more efficiently to create room for us to better meet customer demand for the Azure platform, which continues to exceed our supply.

I know there have been some questions on the rationale behind the voluntary exit offer that we announced this week. Our intent is to offer employees agency if they choose to leave the company on their own terms. We expect very few people to select this offer because of the growth opportunities here at Microsoft. We'll continue to provide a broad set of options for employees to learn, gain new skills, and advance their careers here.

We have tremendous opportunity ahead of us. To meet the moment, we are focused on delivering great products and solutions that help our customers to achieve their many goals — security, innovation, growth — to name a few. Thank you for keeping positive, value-creating outcomes for our customers at the top of the list.

To hear more details about the quarter and our outlook for Q2, you can join live today at 2:30PM Pacific Time, listen on-demand, or read the transcript on the Investor Relations website. Finally, mark your calendars for Microsoft Ignite, on November 17-20, where we'll show how our innovation across AI, cloud, data, security, and Copilot is coming together to help customers transform their businesses.

With appreciation and gratitude,

Amy

Have a tip? Contact this reporter via email at astewart@businessinsider.com or Signal at +1-425-344-8242. Use a personal email address and a non-work device; here's our guide to sharing information securely.

• Google's AI chip business is turning into a TPU-shaped headache for Nvidia.
• A blockbuster deal with Oracle will put Google's TPUs inside another hyperscaler.
• Is the market underestimating the potential of Google's chip business?

Earlier this year, Nvidia CEO Jensen Huang downplayed the threat of other companies building competing AI chips, including Google.

"Without Anthropic, why would there be any TPU growth at all?" he said in an interview with podcaster Dwarkesh Patel, referring to Google's custom chips. "It's 100% Anthropic."

That is no longer the case.

On Tuesday, Google and Oracle announced a blockbuster deal that will see Google's TPUs, short for Tensor Processing Units, made available to Oracle cloud customers from next year. As the companies underscored in a joint announcement, it's the first time Google's TPUs have been offered on a hyperscaler's platform beyond its own.

Anthropic isn't the only one Nvidia should be worried about. OpenAI has a job listing for a software engineer to optimize its models for TPUs, and The Information reported Tuesday that OpenAI plans to buy TPUs from Google.

Meanwhile, several analysts have predicted Google's TPU business is about to explode as more companies hedge their bets with a multichip strategy — providing an opening for Google and others to further move onto Nvidia's turf.

Despite this, Nvidia's chip business still eclipses everyone else's, including Google, which is also a big customer of Nvidia's GPUs. Yet Google's AI chips, which the company began building more than a decade ago to power search and other products, might be poised for takeoff.

On Wednesday, during Google's third-quarter earnings call, chief financial officer Anat Ashkenazi said the company expects a "meaningful revenue ramp in 2027" in TPU sales.

Cloud chief Thomas Kurian also said on the call that TPU sales were a "meaningful contributor" to the company's Q3 financials.

In its IPO paperwork filed this week, Anthropic said it planned to spend $114.5 billion on Google compute between April 2026 and July 2033. While it didn't specify how much of that would go to TPUs specifically, it estimated spending between $125 billion and $150 billion on TPU capacity from Google and other providers.

In the same filing, Anthropic said it's also using Amazon's Trainium chips and AMD's GPUs, as well as chips from Nvidia and Google.

All told, this might be Nvidia's nightmare scenario coming true. If companies are increasingly using a multichip strategy, it could hamper Nvidia's pricing power.

Google declined to comment, and Nvidia and OpenAI did not respond to requests for comment.

TPU takeoff?

Google's TPU business may only be getting started.

In notes published this month, Japanese banking group Mizuho projected TPU revenue of $28 billion for 2026 and $67 billion in 2027. Mizuho said Google could see that number reach as high as $100 billion in 2028, while TD Cowen's John Blackledge projected it could reach $138 billion, accounting for 34% of Google Cloud's 2028 revenue.

That number is still dwarfed by Nvidia, which reported $96.2 billion in revenue in the previous quarter alone, and the market reacted with a shrug on Wednesday as its share price stayed flat.

Some analysts, though, think the market is underestimating Google, especially as more companies look to reduce their reliance on Nvidia's GPUs.

"We believe Google's TPU hardware business is emerging as a materially larger opportunity than what is currently reflected in consensus estimates," Mizuho analysts wrote in a note published October 18, adding that selling TPUs outright — rather than renting them from Google's own data centers — could ease Google's aggressive capex spending.

And even if that spending continues to soar — and right now, there are no signs the foot is coming off the accelerator — analysts at Bank of America said Google's cash-splashing looks "relatively efficient" compared to its peers, as TPUs continue to power the company's own AI efforts.

For Google, it's the payoff of a decade-plus investment in silicon that was originally designed to make Search smarter. It could also act as something of a buffer against damage to its core search business as AI agents eat Google's lunch, Mizuho also said this month.

Have something to share? Contact this reporter via email at hlangley@businessinsider.com or Signal at hughlangley.01. Use a personal email address and a non-work device; here's our guide to sharing information securely.

• AI coding startup Cogniton is fielding offers for a new funding round.
• The round could value Cognition between $70 billion and $80 billion, according to people familiar with the matter.
• The new offers come weeks after the company announced a $2 billion round at a $48 billion valuation.

Just weeks after announcing a new round of funding at a $48 billion valuation, AI coding startup Cognition is fielding new investor interest at a much higher price.

The startup has received offers valuing it at between $70 billion and $80 billion and is seriously considering a round before the end of the year, according to sources familiar with the matter.

The talks around the financing are still early, and valuation targets could still change.

Founded in 2023 by Scott Wu, Steven Hao, and Walden Yan, Cognition is best known for Devin, an AI coding agent designed to automate software engineering tasks. The San Francisco-based company's existing investors include a16z, Accel, Founders Fund, Khosla Ventures, Elad Gil, and Pear VC.

Cognition's rapid valuation climb underscores the investor frenzy around AI coding, one of the hottest and most competitive corners of the AI market. As companies race to sell AI coding agents that can take on more of the work engineers do, investors are placing increasingly expensive bets on the few startups they think could emerge as leaders.

A $75 billion valuation wouldn't be so far-fetched, as SpaceX just acquired fellow AI coding giant Cursor for $60 billion a few months ago.

Last month, Cognition was drawn into a public feud with rival Factory AI, which accused Cognition employees of posing as job applicants to seek product information and faulted a former Factory adviser for meeting with Cognition executives. Cognition and the former adviser denied that confidential information was sought or shared.

Got a tip? Email Katie anytime or contact her on Signal at kroof.26. Sign up for Business Insider's weekly VC Insider newsletter here.

• The startup world is littered with companies that tried to become the next YouTube.
• Andrew Morse is trying it again, anyway.
• Morse's pitch: a curated app that gives creators incentives a reason to give it exclusive content.

Are you sick of the algorithms that serve you video on giant platforms like YouTube and TikTok?

Or are you a video-maker who's sick of the algorithms that don't serve your stuff to would-be watchers?

Andrew Morse would like to talk to you.

Morse, who spent years working in big media (Bloomberg, CNN) and most recently in local news (The Atlanta Journal-Constution), is building Phareo, a startup he hopes will one day rival internet media heavyweights. The secret sauce, he says: his own AI-powered discovery engine that will make both video-watchers and video-makers happy.

This is, to put it politely, a very tall order.

But it's still worth talking about, since Morse is homing in on a problem you hear a lot about in creator circles these days: the sense that people who make content are getting increasingly bad deals from platforms, because they can't control who sees their stuff.

Phareo (the name is an allusion to Pharos, the ancient lighthouse of Alexandria) says it can solve that problem by giving creators real incentives to make exclusive stuff for the company, via levers like revenue share and options. Morse plans to launch in January with an initial group of 25 creators. The app will start out as a free, ad-supported service, but he hopes to eventually offer a paid tier.

I talked to him this week about his plans, the shifting media landscape, and why he's increasingly worried about the creator economy — at the same time he's launching a creator economy app. This is an edited excerpt of our conversation:

Peter Kafka: You have a classic chicken/egg problem: Consumers are not going to use a new app unless it has compelling content they can't get somewhere else. 
Content-makers don't want to put their stuff there unless there's a big audience.

Andrew Morse: It is the most important question we have to solve.
It starts with dissatisfaction, and it starts with a belief from consumers and creators, that they're not being currently served as well as they could be. So we're working with creators who believe that there has to be a better alternative.

We don't really have anything to lose, given the direction that the platforms are taking and given the overreliance that their creative work has on algorithms that don't serve them. We're going to build with creators and try to build a super-passionate core and grow from there.

But YouTube, TikTok, Reels, etc do a great job of giving me tons of video. And they have unlimited content. How do you move me over from the platforms I'm already using?

It's an ironic conundrum — we're trying to solve the discovery problem, but we have to be discovered ourselves.

Over the years, I've seen many companies say they're trying to solve a discovery problem. 
But it doesn't seem like consumers have a discovery problem. It's the content owners who have a problem getting their stuff in front of consumers. What's the incentive for consumers to move to a platform content-owners like? 


There's an inherent assumption that consumers are satisfied with the platforms. I don't think that that's true. I think consumers are on these platforms because they have to be. Whether they're getting what they want in an efficient way is an entirely different story.

Any advice for your former CNN colleagues who are now working for David Ellison?

Stay focused on the work. It's always the journalism that matters. 
And the truth and the facts and great storytelling will always rise to the top.

We have seen a steady stream of journalists leaving places like CNN and starting their own thing. Do you think that trend accelerates or stalls out as people realize how difficult it is to do solo media for a living?

I'm concerned about this creator economy — which is really about about businesses that have built themselves on the back of YouTube. If you go back to 2016, 2017, a number of really clever, really creative businesses were built on the back of social platforms. When the algorithms changed, it became almost impossible for these businesses to stay afloat. It was hard for us at CNN, but we had a big enough business that we were able to mitigate and ride the wave. For these other companies, it was disastrous. 


I think you're going to see more and more people trying to go out on their own — and that's what scares me. 
Because I think that the sands may begin to shift a little bit.

• Dalet Labs, an AI commerce startup, is coming out of stealth mode with a $7 million pre-seed round.
• It's launching a new app called Scene, where you can scroll through shoppable, AI-generated images.
• Read the 5-page pitch deck the startup used to raise capital — months before finalizing its product.

Would you shop from a feed of AI-generated images?

The startup Dalet Labs is betting you will.

The AI-commerce company exclusively told CMO Insider that it raised a $7 million pre-seed round led by Lightspeed Venture Partners and is launching its first product, a shopping app called Scene.

Founded by Omer Bar-Tal in 2025 after he left the AI video startup Runway, Dalet Labs recently came out of stealth mode. The startup raised its pre-seed in late 2025, before finalizing its product road map. In addition to Lightspeed, venture capital firm SV Angel and individual angel investors participated in the round.

Lisa Han, the Lightspeed partner who led the round, said that Bar-Tal has a combination of "video AI research expertise" and a "deep understanding of what consumers want."

Scene, which became available on the Apple App Store in the US on Wednesday, lets people scroll through a feed of shoppable AI-generated images. The app will be invite-only at launch, and isn't yet available on Android. While the images of people or scenes are AI-generated, the products themselves come from real brands across categories like fashion, home decor, and sports.

Bar-Tal envisions Scene as a tool people can turn to when they want to buy something but are still uncertain about exactly what they have in mind. It's similar to scrolling through Pinterest or TikTok in hopes of being inspired to buy something, but Bar-Tal said there's a key difference: You come to Scene to shop, not be entertained.

Bar-Tal sees Scene as a virtual "mall" where users can window-shop, save items to their cart, and make purchases.

Commerce has become an increasingly attractive use case for AI tools like chatbots and agents (including Meta's Muse). However, Bar-Tal argues that AI shopping tools are missing something.

"AI is amazing when you know what you want, but it's really important to actually have an ability to discover and to see things," Bar-Tal said. Agentic commerce, he added, "misses the point in shopping — which is essentially a delightful experience."

As a Scene user scrolls and interacts with products linked in the app, the feed adapts to what type of brands or items the user is interested in. Then, Scene uses AI to generate scenes with people wearing or using those items.

Bar-Tal did not comment on which third-party models Dalet Labs uses.

Bar-Tal said Dalet Labs doesn't see its product as replacing influencers, but rather as "complementary."

"The content here is AI, but it is not aiming to replace creators because we are not aiming to create entertainment," he said.

However, even if Scene isn't competing with influencers for entertaining content, it's still competing with their business model: making money by promoting brands' products and ultimately driving sales.

Scene has a catalog of hundreds of brands and retailers, Bar-Tal said, and is making more partnerships. The company did not share further details on what brands Dalet Labs has already established relationships with.

Bar-Tal said the company is also leveraging affiliate marketing to monetize the app early on, working with affiliate networks to earn from each in-app transaction.

Dalet Labs' Scene is entering a crowded market of apps offering people ways to discover brands and goods, like TikTok's growing Shop presence in the US or Pinterest's visual search prowess. Startups like Doji and Alta have also raised millions on the promise of tech that lets people try on clothes using AI-generated versions of themselves.

How will Dalet Labs get people to download an entirely new app?

Using influencers and social media to promote the app will be a core part of Dalet Labs' user acquisition plan. It also plans to roll out a coupon campaign during the holidays, offering new users discounts on purchases made through the Scene app. Bar-Tal didn't share whether the discounts would be funded by Scene or by brand partners.

The New York-based startup, meanwhile, is using its pre-seed funding to hire more engineers and researchers to build out Scene.

Read the 5-page pitch deck Dalet Labs used to raise its $7 million pre-seed:

The deck starts with background on the technology

Here's what the slide says:

For over half a century, every major shift in computing has been driven by an interface revolution.

The move from terminals to graphical user interfaces (GUIs) transformed computers from expert tools into mass-market products.

The rise of the web and mobile extended that pattern: each new medium made digital interaction more visual, intuitive, and consumable.

Text, by nature, demands active cognitive effort; Visual interfaces go a step further by allowing users to grasp high bandwidth information almost instantly with minimal cognitive effort.

Then it explains Dalet Labs' thesis

Here's what the slide says:

Today, artificial intelligence represents the most profound computing shift since the PC. Yet, paradoxically, our interface to AI remains stuck in the terminal era.

We converse with models through text prompts and receive text streams in return in synchronous manner. Our product thesis is simple and inevitable: AI requires a native GUI.

We believe the GUI to AI will not be code-based, but rather generated by media models.

It then outlines the competition

Here's what the slide says:

The core product of AI consumption platforms is inseparable from their textual chat interfaces.

For example, the OpenAI brand is strictly associated with the chat interface of ChatGPT.

Moreover, by decoupling the user interface from any specific model provider, we can integrate multiple frontier models (OpenAI, Anthropic, Google, or open-source systems) and focus our innovation on the new media generation GUI platform. This modularity becomes increasingly valuable as AI models are commoditized. In addition, their technical stacks are anchored in the code generation paradigm, in contrast to our focus on Media Models.

Finally, it introduces Dalet Labs' founder

Here's what the slide says:

Omer Bar-Tal

• Research experience from Weizmann Institute.

  • Built the first method for editing real-world videos using text
  • Papers in CVPR, ECCV, ICML, ICLR etc.
• Trained video foundation models (Google, Pika, Runway)
• Built creative applications with media models (eg video editing, reference-to-video, etc.)
• Led applied research teams and product launches

• A version of this post appears in CMO Insider.
• Sign up for Business Insider's marketing newsletter.

Keep calm: We have a plan.

That's essentially been the adtech industry's response to Apple's latest crackdown on web trackers.

The workaround may only be a temporary fix, however.

Apple recently implemented new restrictions in its Safari browser designed to stamp out "identity resolution," a method used to track users across different devices and websites by linking data points such as IP addresses, email addresses, and other identifying information collected about them.

Adtech has largely united around a proposed solution: Move the identity resolution process so it takes place inside publishers' and adtech companies' servers rather than in the browser, where Apple has more control. Adtech industry execs have said the move will mitigate disruptions to publishers' revenue and measurement from Apple's blockers, while still allowing them to respect users' privacy choices.

Privacy experts, on the other hand, say that while the server-side shift might succeed temporarily, it's far from a permanent solution. Apple has made it clear that attempts to circumvent its tracking protections could trigger further restrictions on adtech companies.

"It's tough to predict how long Apple would need to find ways to block workarounds, but adtech companies that bet against the browser are in for a rude awakening at some point in the future," said Zach Edwards, a cybersecurity researcher and cofounder of the adtech transparency platform DecryptAds.

The stakes extend beyond the adtech companies whose technology Apple is blocking. Publishers and advertisers have increasingly relied on identity resolution to target specific audiences and to measure whether campaigns are working. If identity resolution becomes less effective, advertisers could shift more of their budgets away from web publishers and to platforms like Google and Meta that have reams of their own logged-in user data.

Anthony Katsur, CEO of the IAB's Tech Lab, which created the Trusted Server project, said its estimates show that the Apple-powered browser ecosystem accounts for 40% of total web traffic, which leaves a large chunk of potentially unaddressable ad inventory.

"That creates tremendous downside for media companies to then prove themselves to agencies and brands they work with that, yes, a consumer saw this ad, and they converted," Katsur said.

How Apple could thwart adtech's server-side solution

How could Apple block identity resolution if the data matching occurs only on servers outside its control?

There is a precedent. When Apple blocked third-party tracking cookies in Safari, some adtech companies used a workaround called CNAME cloaking. It made third-party trackers appear to Safari as if they belonged to the website a reader had chosen to visit — known as "first-party" activity. Apple soon identified and developed protections against that tactic.

Edwards said many server-side tracking systems still rely on cookies or other identifiers stored in a user's browser to recognize them across websites. Apple can block the collection of identifying information before it reaches the server.

Privacy-focused browser company Brave, for example, already blocks scripts that are known to collect and record information used for server-side identification, said Pete Snyder, the company's principal researcher.

Katsur of the IAB Tech Lab said the switch to server-side isn't an attempt to trick the browser.

"This isn't about circumventing consumer privacy preferences; this isn't about circumventing regulations; it's about moving the advertising stack to a publisher-controlled server-side solution that complies with legal regulations," Katsur said.

He added that through the Trusted Server, publishers can also strip out unwanted third-party tracking identifiers from ads.

Katsur acknowledged Apple was likely to clamp down on server-side identity resolution eventually, but that the tech giant risks breaking essential web functionality if it goes too far.

The consumer opt-out option

Some server-side matching may be impossible for Apple to stop through browser technology alone. If someone provides their home address to an e-commerce site for delivery of their purchase, Apple's Safari can't prevent the retailer from knowing it. The retailer could then potentially use that information to match the customer against other databases.

Snyder said it's why Brave supports consumer opt-out mechanisms like the Global Privacy Control signal, a browser setting that tells a website a user doesn't want their data sold to or shared with third parties for targeted advertising, a preference companies must honor where the law requires it. Starting January 1 next year, the state of California will require browsers to provide an easy-to-find setting that lets users opt out of having their personal data sold or shared.

"Protecting users against this kind of tracking requires 'privacy in depth,' or using a wide range of privacy protections, by default and automatically, to minimize how often users share identifiers, and to, whenever possible, block sites trying to collect those identifiers," Brave's Snyder said.

Moving identity resolution to servers may temporarily put the method out of Safari's current reach, but not necessarily out of Apple's sights.

"This is basically a cat-and-mouse game, but the cat in this instance is a trillion-dollar company controlling one of the most valuable user data supply chains in the world," Decrypt Ads' Edwards said.

• A version of this post appears in CMO Insider.
• Sign up for Business Insider's marketing newsletter.

Hugo Veiga and Diego Machado — the former AKQA global chief creative officers — spent two decades building brands at some of advertising's most storied creative agencies. Their next bet: a new type of agency that wants to be paid not just in cash, but in a share of the businesses it helps build.

Veiga and Machado exclusively told CMO Insider this week that they have launched !F Creative Capital, a company that combines conventional commissions with equity partnerships and revenue-sharing arrangements tied to clients' sales.

Rather than charging an upfront fee alone, !F Creative Capital offers creative consultancy in exchange for equity in companies it believes can grow. For example, it took a stake in the superfoods brand Authentic Fruits in a deal that includes a three-year creative engagement, a share of sales, and additional equity tied to sales performance. It has equity in CheckMy.Music, an AI platform that scans songs for plagiarism, and an equity partnership on a new project led by former Chelsea FC and Real Madrid manager Carlo Ancelotti.

A long-held complaint on Madison Avenue is that agency holding companies prioritize businesses such as media buying — and, increasingly, data, technology, and AI services — because they are easier to package, measure, and sell repeatedly than bespoke creative work. Veiga and Machado said that !F Creative Capital was born out of a desire to create a model that more effectively ties the value of creative ideas to business results.

"In many of the creative industries, everyone is too busy doing decks, doing keynotes, and losing the connection with the final thing," Machado said.

!F Creative Capital has also launched its own brands, including SNIFF, a pet company that lets dogs smell food before their owners buy it, and Baitts, a toothpick brand.

Elsewhere, the company is coproducing a feature film set for release next year, and it's preparing to launch an eight-week "School of Creative Entrepreneurship" to help people turn their ideas into businesses.

While !F Creative Capital has taken on commission-based creative work for brands including Corona and Cristiano Ronaldo's Globe Sports Award, Veiga and Machado said they are reinvesting that money into the equity-focused side of the business and their in-house projects. !F Creative Capital is self-funded and has a team of around 15 people spanning seven countries, along with a broader network of contributors.

'You have an idea, and you give it for free'

Veiga said the long-term vision is for !F Creative Capital to become independent from commission work as it builds its own brands and effectively becomes its own client. The result might look more like a holding company for brands than the creative agencies where the duo made their name.

"A lot of times you have an idea, and you give it for free, or you push to make it happen without owning that," Machado said of the agency business.

Adland's creative agencies have explored similar equity-focused business models in the past. Uncommon, the London-based studio in which Havas acquired a majority stake in 2023, took equity in the compostable coffee pod company Halo and the finance app Communion. In 2010, former Deutsch exec Mike Duda cofounded the hybrid early-stage venture firm and creative agency that later became Bullish. Anomaly, now owned by Stagwell, began incubating and acquiring stakes in brands in the 2000s.

Brian Wieser, CEO of the advisory and consulting firm Madison and Wall, said that while the equity model still persists in the agency industry, the practice isn't common and mostly ties to the individuals running those companies because the skill sets are different.

"Most people who can do venture well focus on venture capital," Wieser said. "But the right person with the right relationships can make this happen."

• A version of this post appears in CMO Insider.
• Sign up for Business Insider's marketing newsletter.

McDonald's is exiting the playing field.

The fast-food giant's surprise decision to end its 32-year run as a FIFA World Cup sponsor shows how major global brands are reassessing the cost, clutter, and commercial return of sports' biggest properties.

McDonald's ran its largest-ever World Cup marketing push during this summer's tournament, and McDonald's CFO Ian Borden said on an August earnings call that the effort "underperformed versus our expectations."

The company said last week that it had decided not to renew its current deal, which is due to end this year. It also indicated to FIFA that it isn't looking to invest in next year's Women's World Cup, according to a person familiar with the matter.

Sports sponsorship is seen as a way for brands to tap into huge, live global audiences and associate themselves with the feel-good factor of athletes' achievements. Management consultancy Kearney estimated the global market for sports sponsorship and in-venue advertising at $52 billion in 2025 and expects it to grow by 8% to 9% annually through 2030, reaching about $80 billion.

While many marketers continue to spend heavily on sports, they often find it hard to measure the commercial uplift from their sponsorship outlays. Research from Forrester, published in 2025, found that 76% of US business-to-consumer marketing executives who invested in sponsorships the prior year reported struggling to calculate the return on investment, or ROI, of their sports sponsorships.

"They are gaining a lot of traction around brand awareness, but can't necessarily calculate the hard and fast ROI numbers to prove to their CFOs that the money and long-term commitment for these kinds of sponsorships are necessary," Mike Proulx, a Forrester VP, told me.

For McDonald's, the figures evidently didn't justify a renewal. The deal was valued at $40 million annually, per the research firm GlobalData Sport.

For a globally recognized brand such as McDonald's, the scope for the World Cup to deliver another spike in brand awareness and associated sales may have reached its ceiling. Heineken appeared to make a similar recalculation last year when it said it would end its 30-year UEFA Champions League partnership in 2027. The brewer said it was shifting its sponsorship focus to platforms "where spend is proportionate to value creation," while retaining Formula 1 and adding Premier Padel.

McDonald's CMO Morgan Flatley said in a statement last week that sports remains important to its marketing strategy. At the company's September investor day, Flatley said she was focusing her campaigns "on McDonald's collection of iconic IP, not someone else's." The company continues to sponsor dozens of sports properties, including France's Ligue 1 top-flight soccer league and the All-American Games.

The 2026 World Cup reached nearly 6 billion people worldwide across different platforms, according to FIFA, which helps explain why brands continue to spend heavily to be associated with it. While FIFA has not disclosed a final sponsorship total for the 2026 World Cup, its latest budget forecasts $2.8 billion in marketing-rights revenue across all commercial partnerships in the 2023 to 2026 cycle. That would be about 59% more than it generated across the 2019 to 2022 Qatar cycle. For the 2026 tournament, FIFA also introduced a program that allowed host cities to sell local supporter packages, making the sponsorship market increasingly crowded compared to previous years.

Rory Natkiel, founder of sponsorship effectiveness consultancy Box Count, said the additional host-city partners risked shrinking McDonald's "share of property" — the portion of the event's commercial attention it could realistically command.

That dynamic could make smaller or newer sports properties more attractive to mature global brands. While they might not offer the global scale of long-standing events like the World Cup, they can provide greater flexibility to tailor marketing to specific audiences and channels than the tightly controlled multiyear packages from major rights holders allow. For a company like McDonald's, that could open up an opportunity to offer consumers an experience they remember rather than a sponsorship they simply expect.

"If they start going into new sports and trying to attract a different audience to the global football audience — padel for a slightly more premium audience, or whatever it might be — it does create a new compelling angle and a new avenue for creativity," said Rory Stewart-Richardson, a longtime sports sponsorship executive.

Stewart-Richardson said he wouldn't be surprised if other legacy brands start recalibrating their sponsorship deals, focusing more on local markets and digital activations, with an emphasis on "building from the consumer out rather than trying to buy into temporary cultural relevance."

Amazon warehouse workers used to spend much of their days walking, pushing carts through long aisles to manually retrieve customer orders from shelves.

Today, hundreds of "Hercules" robots do this instead, carrying towering yellow storage pods filled with inventory to employee workstations.

"I'd walk like 10 miles a day on average," said Mohammed Khan, who started as an entry-level picker 12 years ago and is now a senior manager at an Amazon warehouse in Tracy, California.

Amazon's robots "cut down all that walking time," he told Business Insider during a recent tour of the facility.

This site is ground zero for a wave of automation that's beginning to wash over the industry. The warehouse and logistics sectors have been powerful sources of job growth since Amazon kicked off the e-commerce revolution 30 years ago, as Business Insider's award-winning Warehouse Nation series revealed in 2022.

If robots automate many warehouse tasks, this powerful engine of job growth could grind to a halt, wiping out thousands of positions and countless opportunities for future workers.

Amazon says employees are being retrained to do more valuable work, such as managing and maintaining robots. Will these new jobs be enough to cover what might be lost? Or could we actually end up with more or better jobs?

Business Insider went looking for answers inside the giant facility in Tracy, a city in California's Central Valley where suburban neighborhoods give way to open farmland.

"Autonomous to an extent"

Across 1.4 million square feet — the equivalent of 23 football fields — robots with names like Hercules, Robin, Cardinal, Pegasus, and Proteus move inventory and sort packages. But they don't run the warehouse alone. Human workers still pick and pack orders, load trucks, and intervene when automated systems fail.

Jobs have not disappeared, but they're changing. Machines increasingly handle work that once required employees to walk miles and lift boxes, while workers monitor and troubleshoot them. The Tracy site employs more than 3,000, up from roughly 1,000 when it opened in 2023, an Amazon spokesperson told Business Insider.

"They're autonomous to an extent," Khan said of his robot counterparts. "They still need a lot of help and maintenance and oversight."

An Amazon spokesperson said rising customer demand drove the facility's workforce growth and that robots help workers process more packages safely and efficiently rather than simply replacing them.

Amazon's more automated, next-generation fulfillment centers are about 25% more productive and cut order-processing time by as much as 25%. And these sites support roughly 30% more skilled jobs in areas including reliability, maintenance, and engineering.

The "Quarterback" watches robots

One of the workers now doing that kind of job is Briana Brager.

She joined Amazon about two years ago as an entry-level employee after spending years caring for her family. She had no prior experience in a traditional workplace.

Today, she's a process assistant who helps oversee the warehouse's robotic systems. When a chute jams or a robot stops working, she coordinates the response to keep packages moving.

"The robots do a lot, but we still have to maintain them," she said.

Brager began as a picker before moving into a specialized role supporting robotic-floor operations. She got two weeks of intensive training, followed by additional job shadowing, to learn how to work safely around the equipment. Her current job is less physically demanding than her previous one, she said.

Her role is informally known as the "quarterback" position. She monitors fleets of robots and dispatches employees when a machine needs intervention.

"We call the shots," Brager said.

Meet Robin

At the Tracy warehouse, humans and robots move orders together from storage to the loading dock. Employees stow individual items in fabric inventory pods, which Hercules robots carry to pick stations. Workers retrieve the items, then pack and label them.

Robin, a large robotic arm, picks finished packages off a conveyor and places them on small Pegasus mobile robots. Pegasus carries the packages to staging areas alongside other orders headed to the same destination. Cardinal then loads the sorted orders into carts, and Proteus ferries the carts toward outbound trucks.

Khan, the warehouse manager, said workers performed Robin's task manually until a few years ago. They handled packages one at a time, checked them for damage, scanned them, and fed them into the sorting system. He estimated that one Robin can perform roughly the work of two to seven people doing that process manually.

Amazon said that when it introduces systems such as Robin, employees take on other work at the site, including operating robots, clearing jams, monitoring the robotic floor and quality control, and maintenance engineering.

Robots move the carts, too

Proteus offers a clear example of how these new warehouse jobs are taking shape. Amazon brought the robots to Tracy last year.

Unlike Hercules, which operates in fenced-off areas, Proteus can move through spaces shared with people. There are roughly 200 of the robots at the Tracy facility, Amazon said.

Before Proteus arrived, workers manually pushed carts filled with packages to the loading dock. Now the robots pick up full carts, carry them to staging areas near the dock, and retrieve empty carts for the return trip.

Some employees who once moved those carts now work as "ground-floor monitors," Amazon said. Instead of spending their shifts moving packages, they respond when a Proteus reports a fault. If the problem is simple, they can reset the robot; otherwise, human maintenance staff take over.

Some workers trained in these jobs have gone on to Amazon's mechatronics apprenticeship program, which can lead to more technical maintenance work.

When automation hits the "jackpot"

For all the robots moving through the Tracy warehouse, people remain essential when automation breaks down.

Amazon calls one such destination the "jackpot." When an item loses its routing information while moving through the facility, the automated system sends it to this area, which works a bit like a lost-and-found for packages.

Here, a human worker must determine where the wayward package belongs and return it to the correct flow. "It will require manual intervention," an Amazon representative said.

Other situations require human judgment, too. Products can fall from storage pods and inventory can be damaged or mislabeled. Packing individual products into boxes and loading them onto delivery vehicles are tasks that also remain heavily dependent on people.

This helps explain why even a highly automated Amazon warehouse still employs thousands of workers. The company said it needs people for new packing stations as well as roles across the facility that keep operations running safely and efficiently.

Current and future Tetris

By the time packages get to the loading dock at the Tracy facility, the robots are gone. At least for now.

A person loads each trailer. Amazon uses equipment that extends into the trailer and adjusts to a worker's height, reducing bending and back-and-forth walking. But workers still arrange the packages themselves, trying to fill the trailer to use as much of the space as possible.

"It's a lot of Tetris," one Amazon representative said.

Amazon is already exploring how to automate this task, though. As Business Insider previously reported, Project Tetromino aims to use AI and robotics to automate sorting and vehicle loading at delivery stations, where the work has long remained stubbornly manual.

But that vision remains a work in progress. Amazon's machines can carry shelves, sort packages, stack boxes, and move carts around people. Workers still pack orders, load trailers, perform quality checks, handle hiccups and exceptions, and keep the machines running when they fail.

For Amazon, the immediate payoff is not necessarily a warehouse with fewer people, but one where the same workforce can accomplish more. Tracy employs roughly the same number of people as a nearby Amazon facility with fewer robots, company officials said, but the Tracy location processes more volume.

Khan compared the technology to something far less futuristic: a calculator.

The robots streamline work that once consumed employees' time without necessarily eliminating the work itself, he said. "It's not that the amount of work has changed," he said. "It's how we approach it."

Asked whether robots could eventually replace more jobs, including his own, Khan did not try to predict the future.

"Who knows? That's not in our purview right now," he said. "But I would say roles are evolving, and that's the cool aspect about it."

Have a tip? Contact this reporter via email at ekim@businessinsider.com or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.

• Amazon shifts AI focus to business impact, not just raw usage, in new developer guidelines.
• The company warns that AI metrics can create bad incentives, including “tokenmaxxing.”
• Managers are told to examine teams at both the highest and lowest levels of AI adoption.

Amazon just issued new guidelines for software engineers in the AI era: Prioritize business outcomes over raw usage.

Terk Terkowitz, Amazon's vice president and distinguished engineer, said in a memo published Tuesday that AI is changing software development more quickly and fundamentally than earlier shifts such as cloud computing. But engineers should be judged by their impact on customers and the business, he wrote, not simply how much they use AI.

"We all need to embrace this change and take the time to learn, or we risk being left behind," Terkowitz said.

AI coding tools and agents are upending the software industry and changing what it means to be a coder. Amazon calls this the "era of the Renaissance Developer," where human judgement and machine intelligence combine to create new products and services faster and more efficiently.

"This is a time of change for our industry, and a time of change for what it means to be a software engineer," Terkowitz wrote in the memo.

The updated guidelines reflect Amazon's growing effort to make AI adoption part of how it assesses engineers and engineering managers. The company has tracked measures including AI usage, token consumption, and lines of code internally, and used leaderboards to rank teams by some of those metrics.

But the memo warns that using such measures as targets can produce bad outcomes, including "tokenmaxxing" rather than better software or customer experiences.

"Using these metrics could, if used as a target instead of a measure, drive negative outcomes," Terkowitz wrote in the memo.

Amazon has been experimenting with ways to measure and encourage AI use across its engineering organization.

Earlier this year, the company was tracking engineers's use of internal AI tools and some teams used leaderboards tied to measures such as token consumption. The programs created friction among some employees, who worried that pressure to use AI could reward wasteful activity rather than productive work.

The new guidance offers Amazon's clearest public acknowledgment of that risk. Rather than lay out a detailed AI playbook that could quickly become obsolete, Amazon said its revised developer guidelines focus on long-term impact on customers and the business.

Still, the company said AI should be "at the center of software engineering practice" and urged engineers to incorporate the technology into their daily work.

Leaders are expected to use AI tools on production work themselves, understand where they help and where they fall short, and explain that judgment to their teams, according to the memo.

Terkowitz said Amazon has found that metrics such as token use, AI-tool adoption, and code reviews can correlate with the outcomes it values. But he said the measures must be used carefully. More token use can raise costs, while a focus on lines of code can create unnecessary complexity.

He also recommended that managers pay particular attention to teams at both ends of the AI adoption spectrum: those in the bottom 10%, to understand why they are not using AI effectively, and those in the top 10%, to determine whether heavier use is translating into faster delivery.

Terkowitz said Amazon's most AI-forward "frontier teams" have shown stronger productivity and delivery. He also said teams and individuals that are not adopting AI or increasing output metrics are falling behind.

But his memo left room for engineers whose work may not yet benefit from current AI tools. In those cases, he said, employees should document why and work toward a solution by sharing their findings with peers and managers.

"An engineer who can explain where AI does not yet help their work is demonstrating the judgement we are asking for," the memo said.

Have a tip? Contact this reporter via email at ekim@businessinsider.com or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.

• Anthropic is touting its ability to use different AI chips in its initial public offering filing.
• Analysts say a multichip strategy could weaken Nvidia's pricing power.
• Switching chips is hard and expensive, putting it out of reach for smaller players.

Anthropic is hedging its bets by using multiple chips to handle its AI work — a strategy that could eventually erode Nvidia's dominance.

The AI lab says it manages compute "as a single, fungible pool from multiple strategic partners" — with software that can direct its work to different chips depending on where they're most productive, according to Anthropic's S-1 filing, which came out Monday as the company prepares to go public.

Anthropic says it uses hardware from four chip families: Amazon Trainium, Google TPUs, Nvidia GPUs, and AMD GPUs.

That flexibility, which Anthropic said "took years and significant investment to build," could make major AI companies less dependent on Nvidia over time and give them more negotiating leverage, analysts said.

"The real exposure could be Nvidia's pricing power," said Patrick Moorhead, founder and CEO of Moor Insights & Strategy, "because a customer that can shift work across four platforms negotiates very differently than one that can't."

Ashish Nadkarni, an analyst at IDC, added that a multichip approach also makes Anthropic less vulnerable to fluctuations in Nvidia's chip supply.

Breaking Nvidia's grip isn't easy

Although Anthropic's chip mix isn't new, the company is now pitching it as a core advantage to investors, Moorhead said.

The threat has its limits. Switching between chips is still difficult — especially when it comes to training AI models as opposed to running them, Moorhead said — and only the biggest AI companies have the resources to do it.

Anthropic also continues to spend billions on Nvidia compute, Moorhead said.

Gil Luria, head of technology research at D.A. Davidson, said that while other major companies like OpenAI likely take a multichip approach, it may not be worth the effort and investment for smaller players.

Brad Gastwirth, global head of research and market intelligence at Circular Technology, said Anthropic's strategy doesn't represent a major near-term threat. Having more chip options doesn't erase Nvidia's advantages in performance, networking, and software, he said.

There's a broader risk for Nvidia if AI compute becomes more of a commodity.

Citadel CEO Ken Griffin has warned that if supply overtakes demand, it could drive down rental prices.

"GPUs, like airlines, have a very low marginal cost to run," he told Bloomberg last week. "The moment you have overcapacity, pricing goes into freefall."

Nvidia did not immediately respond to a request for comment from Business Insider.

Have a tip? Contact this reporter via email at gweiss@businessinsider.com or Signal at @geoffweiss.25. Use a personal email address, a non-work WiFi network, and a non-work device; here's our guide to sharing information securely.

• Sam Bankman-Fried's Anthropic stake was liquidated in bankruptcy court two years ago.
• The feds seized shares belonging to two co-conspirators, Caroline Ellison and Nishad Singh.
• The government quietly sold those Anthropic shares last year. It's unclear where the money will go.

A little over a year ago, the federal government acquired a stake in Anthropic without paying a penny.

The shares were seized from associates of the crypto fraudster Sam Bankman-Fried. From there, they disappeared into a legal black hole.

As Anthropic rockets toward what could be the biggest IPO in history, those shares could now be worth billions of dollars.

Bankman-Fried's own Anthropic stake was liquidated in the bankruptcy of his failed cryptocurrency exchange. The Anthropic equity owned by Caroline Ellison and Nishad Singh, two associates who invested alongside him, took a different path. The government seized those shares and sold them to existing Anthropic shareholders, according to a person familiar with the sale.

It's still unclear whether the proceeds will be paid out to victims of the $11 billion fraud case, or if the feds will just keep the money.

It also remains a mystery which investors bought Singh's and Ellison's stock and how much they paid, but Anthropic shares have been on such a dizzying climb that it's likely those investors made a killing on them.

Bankman-Fried was convicted in 2023 on fraud and money laundering charges and sentenced to 25 years in prison after the collapse of his cryptocurrency empire. Prosecutors said he used his crypto hedge fund, Alameda Research, to siphon billions of dollars from customers of FTX, his crypto exchange.

The story of the crypto criminals' Anthropic shares is a tale with extraordinary elements, including one of the biggest financial frauds in modern history, a messy bankruptcy docket, and a company that experienced one of the fastest appreciations in business history as the world suddenly realized AI's existential importance.

An intelligent investment

Bankman-Fried, Singh, and Ellison each invested in Anthropic's 2022 Series B funding round. Bankman-Fried bought $500 million worth, which according to court records represented 13.56% of Anthropic at the time. Singh acquired $40 million in shares, and Ellison acquired $10 million in shares, court records reviewed by Business Insider show.

The artificial intelligence company's valuation has soared over the past four years. On the secondary market, Anthropic has been valued at $1.5 trillion. It is the most valuable private company in the world, according to Crunchbase.

The Anthropic shares purchased by Singh and Ellison could together be worth between $4.17 billion and $5.03 billion today, based on the $965 billion valuation the company announced this May, according to Olav Sorenson, who teaches venture capital strategy at UCLA. Harrison Rolfes, an analyst at PitchBook, put the combined figure at $2.62 billion. If Anthropic went public at a $2 trillion valuation, the shares would be worth about $5.44 billion, Rolfes said.

Four of Bankman-Fried's close friends and executives at his companies pleaded guilty as co-conspirators. Three of them, including Singh and Ellison, testified against him.

Ellison was the CEO of Alameda Research, which traded and invested funds that belonged to FTX depositors. She was also Bankman-Fried's on-and-off romantic partner. Singh, an FTX executive and early employee, misrepresented the exchange's revenue to investors.

After the collapse of FTX, Bankman-Fried's Anthropic shares — which were owned by Clifton Bay, an entity affiliated with Alameda Research — ended up in bankruptcy court along with the other dredged-up remains of his companies. His Anthropic shares were liquidated to pay FTX's creditors.

The estate in 2024 sold Bankman-Fried's Anthropic shares to a few dozen buyers for a total of $1.3 billion, more than double what he paid. The largest stake went to an entity affiliated with a United Arab Emirates sovereign wealth fund, bankruptcy court filings show.

The feds take a stake in Anthropic

As part of their sentences, a judge required Singh and Ellison to forfeit their Anthropic shares, which prosecutors said could be considered proceeds of their crimes.

At Singh's sentencing hearing, one of his attorneys, Andrew Goldstein, said Singh purchased his shares before participating in the criminal conspiracy and "he actually may have had a legitimate legal claim" to the shares but agreed to give them up as part of his plea agreement "because it was the right thing to do."

Reached for comment for this story, Goldstein told Business Insider that Singh hopes the government is able to quickly distribute proceeds of the sale to FTX victims. An attorney for Ellison declined to comment.

A federal judge ordered Ellison's and Singh's Anthropic shares to be transferred to the federal government, which took ownership of Ellison's shares in February 2025 and Singh's in April of that year, according to previously unreported court records.

Ordinarily, victims of crimes are compensated through a restitution process, which is overseen by courts. But the number of potential FTX victims could be in the millions, prosecutors said in court filings. As a result, the judge ruled, victim compensation would be handled through a process called remission, which is overseen by the Justice Department.

During Ellison's sentencing hearing, Justice Department prosecutors told the judge that the DOJ would either set up its own claims administration process to compensate victims, or work with the FTX bankruptcy process to identify victims and provide forfeited funds to them.

At the time, FTX's bankruptcy process was in full swing. A Delaware court appointed Sullivan & Cromwell, the elite Wall Street law firm, to untangle the company's assets, figure out who was owed money, and pay them.

The creditors in the FTX bankruptcy, prosecutors noted, largely overlapped with FTX's victims. They were generally depositors, lenders, and investors who were defrauded by Bankman-Fried and other executives. Prosecutors said the Justice Department could work with the bankruptcy estate's lawyers to get money back to them, as it had in previous large-scale financial frauds such as Bernie Madoff's Ponzi scheme.

There was an unusual twist that set the FTX bankruptcy apart. Bankman-Fried's investments — especially the Anthropic shares, along with some cryptocurrency — had grown substantially in value since FTX's collapse. Earlier in 2024, the Sullivan & Cromwell lawyers who'd taken over FTX projected that all the creditors would be repaid in full, with interest.

The Marshals take control

While prosecutors said they intended to use Singh's and Ellison's Anthropic shares for remission, the Justice Department could technically do whatever it wanted with them, according to Duncan Levin, a white-collar defense attorney who teaches a course on forfeiture at Harvard Law School.

"It's a very opaque process," he said. "It's completely at the discretion, by law, of the attorney general of the United States."

Typically, when the feds seize shares of private companies through criminal asset forfeiture, they send the shares to the US Marshals Service Complex Assets Unit for liquidation. The unit tries to value the shares as any other investor would, said Michael Bachner, a white-collar criminal and securities litigation attorney.

"They may look to: What would an institutional purchaser pay for these securities?" Bachner said. "Are there funds that are valuing the securities? Is there a secondary market already out there?"

For the sale of the Anthropic shares, timing was crucial. The Marshals Service is supposed to preserve as much value as possible, Bachner said. By the time the feds got hold of the shares, Anthropic had become an economy-shifting AI giant. In its March 2025 Series E fundraising round, Anthropic was worth $61.5 billion. By its Series G round at the start of 2026, it was worth $380 billion.

At the same time, each fundraising round diluted the Series B shares. The FTX estate told the bankruptcy court that Bankman-Fried's shares, which in 2022 represented 13.56% of Anthropic, represented 7.84% of the AI company in January 2024.

The Marshals Service sold Singh's and Ellison's shares to existing Anthropic investors sometime last year, according to the person with knowledge of the sale.

It's not clear when exactly the agency sold the shares, to which investors, how those investors were chosen, at what price the shares were sold, or how much money the government made in the sale. Depending on when they were sold in 2025, the combined shares could have been worth between $300 million and $1.1 billion, according to Sorenson, the UCLA professor. Rolfes, the PitchBook analyst, estimated a range between $250 million and $630 million, depending on the timing.

The Marshals Service declined to comment. A representative for the Justice Department said information about asset sales and victim compensation is confidential.

The revenue from the sale of Singh's and Ellison's Anthropic shares doesn't appear to have been transferred to the FTX estate as of the end of June this year, according to bankruptcy court filings from the estate, which continues to compensate victims and creditors.

The FTX estate received $638 million last year from assets seized by the Justice Department, according to the estate's annual report for 2025. It's not clear from the filings where those funds came from. The estate expects to receive about $400 million more from the government at some point in the future, according to the annual report. That would include proceeds from cryptocurrency and other investments made by Bankman-Fried.

One of Bankman-Fried's victims, Sunil Kavuri, told Business Insider that the Justice Department should use the proceeds of Ellison's and Singh's Anthropic shares to compensate victims. It would be "diabolical" for the government to hold onto the proceeds, he said.

Victims haven't actually been made whole, he said, because the bankruptcy court calculated the losses of FTX depositors using the time of FTX's bankruptcy declaration, when crypto prices were at a low ebb. Kavuri previously argued in bankruptcy court that the FTX estate should have held on to its Anthropic shares for longer to take advantage of the company's swift growth.

Representatives for the FTX estate declined to comment. The Justice Department spokesperson said the matter was ongoing, and that the DOJ "prioritizes victim compensation from forfeiture and takes all steps to ensure forfeited funds are provided to victims."

The government could end up just keeping the money, Bachner said.

"They've invested millions and millions and millions of dollars in prosecuting Bankman-Fried. And they want to get at least reimbursed for their costs, so sometimes they'll do that," he said. "It's really a unilateral government decision."

Jack Newsham and Katie Roof contributed reporting for this story.

Correction: August 31, 2026 — An earlier version of this story misstated the number of co-conspirators who testified against Sam Bankman-Fried. It was three.

Update: August 31, 2026 — This story has been updated to reflect Singh's role in the FTX fraud case.

Correction: October 27, 2026 — FTX bankruptcy court filings don't make clear where the Justice Department obtained the $638 million it transferred to the estate. A previous version of this story mischaracterized the filings.

• Machenta raised $40 million in seed funding led by Creandum and Union Square Ventures.
• It says robots can learn tasks from one demo using a model trained entirely in simulation.
• The startup remains focused on research as it takes on multibillion-dollar rivals.

A London-based robotics startup is betting it can teach robots new tricks without collecting the mountains of real-world data its much bigger rivals are chasing.

Machenta, founded by researchers from Imperial College London, has raised a $40 million seed round to build an AI model that can teach robots new physical tasks after seeing them performed just once, the company exclusively told Business Insider. Venture capital firms Creandum and Union Square Ventures co-led the round. Machenta declined to disclose its valuation.

While rivals building general-purpose AI models for robots are scrambling to amass real-world data, Machenta is taking a contrarian approach. Edward Johns, the startup's CEO and cofounder, and an Imperial professor, said the company trains its model in computer-generated environments that mimic how objects move and interact, without using a "single pixel of real-world data."

Once trained, the model can direct a robot to perform a new task, such as tidying a table or putting objects into a box, after watching it done just once, without further training, Johns said. In a video shared with Business Insider, three robots pack and close a box on an assembly line, a task Johns said they learned this way.

The startup has 11 employees and remains focused on research, with no near-term plans to commercialize its technology. Its model can learn some tasks immediately, but more precise tasks and complex movements remain challenging.

Machenta faces rivals with multibillion-dollar valuations in one of robotics' most competitive and well-funded fields. Its California-based competitors include Physical Intelligence, valued at roughly $11 billion, and FieldAI, which is raising funding at a $10 billion valuation, Business Insider reported this month.

Another rival, Skild AI, which is valued at over $14 billion, is already selling technology with similar capabilities. In August, Skild unveiled a model that it said could perform previously unseen tasks lasting up to 10 minutes after a single video demonstration. Last month, it said that it had reached $100 million in annual recurring revenue from more than 60 customers.

From research lab to startup

Six years ago, Johns began investigating whether robots could learn new physical tasks as quickly as humans at his Imperial College lab.

In the lab's early experiments, researchers physically guided a robot through a task, such as picking up a spatula and scooping an object. The robot then spent about 30 minutes collecting additional data about the object and its surroundings before it could perform the task on its own.

Johns and former Ph.D. student Vitalis Vosylius spent the next few years working to eliminate that extra step, founding Machenta late last year after seeing early signs that their model could perform new tasks immediately after a demonstration.

Companies are increasingly buying and collecting data to teach robots how to interact with the real world. But Johns believes simply collecting more data isn't the answer. It also remains unclear how much training data will be needed to make general-purpose robots a reality. Johns said that Machenta is taking on the harder challenge of designing algorithms that can generate synthetic data useful for teaching robots physical tasks.

Machenta plans to use the funding to expand its team and train its model with more computing power and simulated data, hoping to teach robots to perform more precise and complex tasks.

Have a tip? Contact Rya Jetha via email at rjetha@businessinsider.com or Signal at rjetha.07. Use a personal email address and a non-work device; here's our guide to sharing information securely.

• Amazon launched a series of new devices this month, including Alexa tablets and Bee AI wearables.
• Rising memory costs forced "difficult trade-offs" on Amazon's device prices, sales, and profit.
• Amazon is seeking more Alexa revenue from enterprise sales, licensing, and ads.

Amazon launched a flurry of new consumer devices this month. Behind the scenes, rising memory costs are adding pressure to its already money-losing hardware business.

An internal planning document obtained by Business Insider shows Amazon forecasting a $612 million gross loss this year for its first-party devices business, compared with a leadership target of a $438 million loss.

The document said rising memory costs were "putting pressure" on the business's profitability, forcing "difficult trade-offs" among sales volumes, average selling prices, and gross profit.

The projected $612 million loss was at the gross profit level, meaning it did not account for all operating expenses, such as R&D, sales and marketing, and administrative costs.

The figures offer a rare glimpse into the pressure on Amazon's hardware business as it absorbs rising component costs. Other tech hardware companies, including industry leader Apple, face similar pressure from surging memory chip prices.

An Amazon spokesperson declined to comment.

A new wave of AI devices

Despite the pressure, Amazon continues to expand its consumer hardware lineup. Even if the company loses money on device sales, it can make money in other ways, such as subscription services, as more customers use its gadgets.

In October alone, the company introduced new Kindles, Alexa-powered tablets, Ring smart locks, and AI wearables under its Bee brand. The August planning document identified Bee and new smart glasses as strategic products still in an investment phase.

Amazon has historically used low-cost hardware to pull customers into a broader ecosystem of shopping and digital content. Rising component costs could make that strategy more expensive, forcing the company to raise prices.

The clearest shift is in Amazon's tablets. Earlier this month, the company phased out Fire tablets after 15 years and replaced the budget-friendly lineup with pricier Alexa Tablets powered by Alexa+. The new devices cost between $229.99 and $499.99, a notable increase from previous generations.

Not just Amazon

The planning document was prepared in August, the same month Amazon raised prices on products including Echo speakers, Kindle e-readers, and Fire TV devices.

The pressure extends beyond Amazon. Apple has raised prices on some Macs and iPads, while Microsoft and Sony have increased console prices. Dell, HP, and Lenovo also face higher memory costs.

The AI boom is helping drive the squeeze. Huge data center investments have increased demand for memory chips, while manufacturers have devoted more capacity to high-bandwidth memory used in AI servers. That has tightened supplies and raised prices for components used across consumer electronics.

Amazon is experiencing both sides of the boom. It expects roughly $220 billion in capital expenditures this year, largely to expand AWS data centers for AI demand, while higher component costs squeeze its consumer hardware business.

The rising cost of Alexa+

The August planning materials show Amazon reassessing both Alexa+ revenue and operating costs. They called for updated forecasts, expansion through enterprise sales and advertising, and more detailed AWS cost modeling.

Amazon is now confronting the cost of operating Alexa+ at scale. Business Insider reported in July that internal forecasts put Alexa+'s AWS cloud costs at roughly $1.7 billion in 2026, nearly triple the prior year's figure. Amazon was working to lower those costs by relying less on Anthropic's AI models and routing more requests to its own, cheaper systems.

A separate project called Moonraker, intended to give Alexa more advanced AI-agent capabilities, was projected to require more than $100 million in GPU spending this year. Amazon internally considered delaying or scaling back the initiative as one option for easing cost pressure, Business Insider previously reported.

Have a tip? Contact this reporter via email at ekim@businessinsider.com or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.

• Infrastructure startup Hyperbolic raised a $30 million round led by Intel CEO Lip-Bu Tan's VC firm.
• The startup connects companies with neoclouds and makes their AI chips easier to use.
• Hyperbolic also wants to let customers resell unused computing power.

Intel CEO Lip-Bu Tan is backing an infrastructure startup that has raised $30 million to help companies rent compute across hundreds of different AI cloud providers.

Hyperbolic, an AI chip marketplace and software startup, has raised $50 million to date.

Walden Catalyst, the venture capital firm founded by Tan and former Samsung president Young Sohn, led the Series A-1 funding round. Dell Technologies Capital, General Catalyst, Vista Equity Partners, and Cambium Capital also participated.

As companies scramble for computing power, one challenge is that AI cloud providers, or neoclouds, use a mix of different chips, networking, and software. Hyperbolic helps customers find compute resources from different providers and creates software that makes it easier to use them all, rather than dealing with each one separately.

"It's been a really fragmented market that makes the customer experience really painful," said 31-year-old founder and CEO Jasper Zhang.

Hyperbolic also wants to let customers resell compute that they're paying for but no longer need.

The four-year-old startup works closely with about 30 neoclouds, including WhiteFiber and Vector Core Compute. It says it has roughly $200 million worth of compute connected to its software annually.

Hyperbolic previously raised a $12 million Series A round in 2024, led by Variant and Polychain Capital, following a $7 million seed round earlier that year. The company expects to have 50 employees by the end of the year and will use its funding to expand software tools for managing AI chips across different neoclouds.

Zhang said the company is aiming for $150 million in annual recurring revenue by the end of the year and is already breaking even. More than 250,000 developers, including teams at Modal, Higgsfield, and Daytona, use Hyperbolic.

Before founding the company, Zhang worked as a quantitative researcher at Citadel Securities. As Wall Street explores ways to financialize AI chips, he said his startup's plans to let companies resell unused compute could help pave the way for a futures market.

Here's a look at the pitch deck Hyperbolic used to raise its $30 million Series A-1. Slides have been redacted so that the deck can be shared publicly.

• Anthropic shows strong profit potential, with revenue growing way faster than expenses.
• Anthropic's revenue per employee exceeds $6.6 million, outpacing estimates for Meta and Google.
• Anthropic has a surprisingly diverse customer base and huge cash reserves.

For years, one of the biggest unanswered questions about the artificial intelligence boom has been surprisingly simple: Can the companies building the world's most powerful AI models actually make money?

After spending hours poring over Anthropic's IPO filing on Monday, we finally have an answer: Yes. And potentially lots of it.

A couple of Chinese AI labs went public earlier, revealing appalling financial results that sowed doubt about the industry's economics. Now Anthropic has provided a much more compelling demonstration of how this business can work at enormous scale.

The numbers are remarkable. Anthropic generated $11.6 billion in revenue during the second quarter of 2026, up 1,368% from a year earlier. During that same period, spending on computing and infrastructure increased 300%, while other expenses rose 741%

So, the company is bringing in money much faster than its most important costs are rising.

For every dollar Anthropic spent on computing in the second quarter, it generated $2.08 in revenue, up from just 40 cents in the first quarter of 2025. That's a more than fivefold improvement over five quarters.

The result: Anthropic recorded $974 million in quarterly operating profit, representing an 8% margin. Its adjusted operating profit margin reached 19%, compared with -129% a year earlier.

"Frontier-lab unit economics can turn positive at scale," Mark Mahaney, a top tech analyst at Evercore ISI, wrote in a research note on Monday.

Better than Meta and Google

There's more evidence that Anthropic has built an unusually efficient business.

Over the year ending September 30, revenue increased roughly 1,200%, while its employee count grew about 170%. Revenue per employee jumped from $1.4 million to more than $6.6 million.

For comparison, Mahaney estimated that Meta generates about $3.2 million in annual revenue per employee, while Google generates about $2.5 million.

That comparison isn't perfect. Anthropic relies heavily on cloud providers, whose employees handle much of the computing infrastructure and distribution work. The improvement is still striking.

Customer surprise

Another surprise: Anthropic doesn't depend as much on a handful of enormous customers, as some investors had feared.

The company had more than 11,900 customers generating at least $100,000 in annual revenue as of September, compared with over 950 a year earlier. More than 1,900 customers spend at least $1 million annually.

No single customer accounted for more than 10% of revenue during the first nine months of 2026. Its five largest customers represented about 21% of third-quarter revenue, down from roughly 36% in the first quarter.

"Concentration may be better than feared," Mahaney wrote.

Cash cushion

Anthropic has a formidable financial cushion. Its June 30 balance sheet showed more than $130 billion in cash and short-term investments.

And Mahaney estimated that the company has burned through just $9 billion since it started more than five years ago. That's a long runway.

Now, don't ask me whether all this makes Anthropic worth $2 trillion, $1.5 trillion, $1 trillion, or considerably less. If I knew that, I'd be making a lot more money as an investor. And even professional investors don't know.

An $800 billion caveat

Some enormous caveats: Much of Anthropic's revenue comes from customers paying for what they use, rather than signing long-term contracts. Meanwhile, some of its computing agreements require minimum payments whether or not it needs all the capacity.

Mahaney estimates that Anthropic has entered into nearly $800 billion in computing arrangements, although only about $266 billion was fixed or guaranteed as of June.

That's still a staggering amount.

"The commitments assume revenue compounding that is not contracted," Mahaney wrote.

Translation: While the company has committed to spending hundreds of billions of dollars on AI computing infrastructure, it hasn't secured the future revenue needed to pay for all of it.

Importantly, most of the potential $800 billion exposure isn't fixed. Many arrangements depend on capacity being delivered, can be adjusted, or include cancellation provisions.

In-house compute and capex

Another wrinkle: Anthropic warned Monday that its spending will increase as it shifts toward leasing AI chips and operating more of its own computing infrastructure, rather than relying entirely on Amazon, Google, and other cloud providers.

That could make future cash flow considerably less attractive, even if operating profits continue to improve.

Still, the central takeaway from Anthropic's IPO filing is difficult to ignore.

For years, investors have wondered whether selling AI could become a genuinely profitable business, rather than an extraordinarily expensive scientific experiment.

Anthropic has now shown that it can. The next question is whether it can keep doing so as the bills get much, much bigger.

Sign up for the AI Insider newsletter here. Reach out to me via email at abarr@businessinsider.com

• Microsoft is offering buyouts to about 60,000 employees.
• The offers are individualized to each employee based on their level, location, and tenure.
• Business Insider reviewed offers to show what employees get when they leave the company.

As Microsoft sends out its buyout offers to more than a quarter of its workforce, specific offers reviewed by Business Insider reveal exactly what employees get if they choose to leave the company.

The offers are individualized for each employee based on their level, location, and tenure. Those reviewed by Business Insider confirm that the US buyouts align with the company's US standard severance offers. A Microsoft spokesperson said about 60,000 employees would be eligible for the buyout.

Microsoft uses a levels system to represent seniority, from entry levels starting at 59 to senior levels at 70, or vice president. The company offered buyouts to employees on select teams in the US, Canada, India, and the UK at level 70.

Offers reviewed by Business Insider show employees at internal levels 64 and below were offered one week of base pay per six months of service, and employees at levels 65 to 67 were offered two weeks of base pay per six months of service.

Buyout offers also include continued regular stock vesting and health insurance. The standard stock-vesting severance offer is six or 12 months for levels 67 and below, based on years of service.

Eligible employees will have until early December to decide whether to take the offer and exit at the end of the year, and there are no plans for another buyout at this time, a Microsoft spokesperson said.

Standard severance is capped at 39 weeks of base pay for levels 67 and below, while the standard severance offer for level 68 includes 12 months of salary.

Earlier this year, Microsoft rolled out a voluntary retirement program for employees. About 7% of Microsoft's US workforce, or about 8,750 people, were eligible at the time. As Business Insider previously reported, Microsoft's voluntary retirement program offer also aligned with the company's standard severance offer, including continued regular stock vesting for six or 12 months, based on years of service.

Earlier this year, another tech giant, Google, offered buyouts to staff in its business unit, Business Insider previously reported.

As more tech companies offer buyouts, some human resources professionals previously told Business Insider that this approach could soften job cuts, and also that when they're made available to workers of all tenures, companies risk losing talent.

Have a tip? Contact this reporter via email at astewart@businessinsider.com or Signal at +1-425-344-8242. Use a personal email address and a non-work device; here's our guide to sharing information securely.

• Instagram is opening its arms to long-form video again.
• Adam Mosseri, Instagram's top exec, said longer videos will help the app stay "relevant."
• Instagram is testing tools for creators making episodic content. But will it pay for them?

Instagram's top executive, Adam Mosseri, has a plan to keep the platform "culturally relevant" — by making it more like YouTube.

Meta-owned Instagram unveiled an overhaul of long-form video features on Monday, including a test of horizontal videos called "Features." Taken together, the changes make the app feel a lot more like YouTube.

"We want to be the most culturally relevant platform," Mosseri said in an interview with content creators and founders of Press Publish, Colin Rosenblum and Samir Chaudry. "To do that, we just need to show up where people are."

The context: YouTube has become a major hit on TV screens, and there's a growing appetite for Instagram-native shows with recurring episodes.

"Subway Takes," hosted by Kareem Rahma, is a prime example of Instagram's strengths and limits. The show started in 2023 as a short-form interview series and, as it grew, expanded with longer-form cuts on YouTube and the streaming platform Tubi. Instagram, however, is where Rahma's show still has its largest audience of 2.4 million followers.

Instagram doesn't want people to go to YouTube or Tubi to see longer versions of shows like "Subway Takes." That's leaving ad revenue and user engagement on the table. Its new tools are designed to entice creators and viewers to stay on its platform.

Here are Instagram's latest video tools and how creators can use them:

• "Features" keep viewers within the Instagram ecosystem. Instead of leaving Instagram to watch the extended version of an interview or episode, it'll be relatively seamless to stay on Instagram and watch that longer-form content. Instagram's "Features" can be as long as 90 minutes, although the new format is, for now, only available to a limited number of creators in the US and Brazil.
• Instagram is feeding more content to its TV apps. Instagram is also pushing its "Instagram for TV" app further. The app will feature various video content on TV services like Samsung, Amazon Fire TV, Google TV, and now Apple TV. People can now watch a feed of content their friends are also liking or resharing on TV.
• Instagram-native shows can label their episodes more clearly with "Series." Within the main Instagram feed, Instagram is testing a way for shows on the app to seamlessly connect episodic content. Eligible creators can add videos to a Series, and a label appears over the content indicating that there are additional episodes. People can then watch those episodes in order from the start. This test is expanding to more countries through the end of the year.
• Clips can directly drive engagement back to the original video. Instead of clipping content and praying that the audience will go the extra mile of clicking on an account and finding the original, longer-form video, Instagram is letting creators link up to 8 clips to one long video.

This isn't Instagram's first time dabbling in long-form content.

In 2018, it launched IGTV, a long-form video format and stand-alone app. In 2020, Instagram rolled out a program that let creators earn money from those videos on IGTV. Four years after launching, the company sunset the IGTV app and shifted its video efforts to Reels.

There are a few key elements missing from Instagram's big leap (back) into long-form video.

You can't watch the video while continuing to scroll through Instagram, or multitask by opening other apps while keeping the Instagram content in a picture-in-picture pop-up. There also aren't automated transcripts and subtitles yet. A spokesperson for Meta told Business Insider that there are plans for these features.

Will creators get a slice of the ad revenue pie?

For creators, one key question and pain point looms: Will Meta share the honeypot?

Mosseri told Rosenblum and Chaudry that Instagram is thinking about ways to test pre-roll and mid-roll ads that could lead to a revenue split. However, he said previous attempts by Instagram to build a creator monetization system comparable to YouTube's Partner Program didn't prove beneficial for the platform.

"We spend more money than we make back by a long shot," Mosseri said of Instagram's previous efforts at broad revenue-sharing programs.

Mosseri said that in its previous iterations of content monetization, creators typically didn't produce more content than they already did for free, and the content they did create tended to be lower quality.

Still, Instagram will need to "figure it out" as it tries to crack longer-form content, Mosseri said.

"I don't think of platform payments or performance-based incentives or bonuses as a revenue driver," Mosseri added. "I'd be psyched to scale them up for any of these formats — as long as we're breaking even."

• The feds seized Anthropic shares last year from Sam Bankman-Fried's associates.
• Several months later, the government sold them for $627 million.
• If Anthropic IPO'd in November at $1.5 trillion, those shares would be worth nearly $4 billion.

Last year, the federal government made a windfall of $627 million when it sold Anthropic shares it had seized from associates of crypto fraudster Sam Bankman-Fried.

If it had held on to that stake, its windfall could have been a lot bigger.

The federal government last year sold shares of Anthropic that had belonged to Caroline Ellison and Nishad Singh, two co-conspirators in the FTX fraud case. The proceeds went to compensate victims of the fraud, a Justice Department spokesperson told Business Insider.

Anthropic's S-1 filing on Monday revealed how much the federal government made from the sale of those shares. The prospectus doesn't name Ellison and Singh specifically. It says the government in October 2025 sold 4,453,900 Series B shares of Anthropic that "had been forfeited by former holders in legal proceedings to which we were not a party."

The figure matches up exactly with the combined $50 million in shares that Singh and Ellison purchased for $11.2261 each in Anthropic's 2022 fundraising round, and later forfeited to the government.

The shares represent about a quarter of a percent of Anthropic today. But given the company's explosive growth, they are still worth a fortune.

Anthropic hasn't yet announced the stock price for its expected November initial public offering. At a $1.5 trillion valuation, the shares would be worth a total of about $3.94 billion. That would be six times as much as the government sold the shares for a year ago.

"One of the Justice Department's top priorities is returning forfeited funds to victims through the asset forfeiture program whenever the law allows," the DOJ spokesperson said.

The feds' stake in Anthropic

How the federal government came to own a chunk of Anthropic in the first place is a story involving a notorious fraudster and a complicated criminal forfeiture process.

Bankman-Fried participated in Anthropic's 2022 Series B fundraising round with Singh and Ellison. Singh purchased $40 million of the Series B shares, and Ellison acquired $10 million.

Singh and Ellison later pleaded guilty to crimes related to their work with Bankman-Fried and testified against him at his Manhattan federal criminal trial. The jury found Bankman-Fried guilty, and a judge sentenced him to 25 years in prison. The federal government took ownership of Singh's and Ellison's Anthropic shares by April 2025.

The US Marshals Service sold Singh's and Ellison's shares, rather than hold on to the red-hot Anthropic stock, Business Insider previously reported. The US Marshals Service Complex Assets Unit is typically tasked with dispensing shares of private companies acquired through criminal forfeiture.

Now we know that Anthropic purchased the stock from the feds in October 2025 at the price of $140.9676 per share, which is the same price Anthropic set for its Series F-1 stock in a fundraising round a month earlier.

Anthropic resold the shares to nine investors at the same $627 million total price it paid to purchase them from the federal government, so it didn't make a profit from the sale, according to the S-1 filing. It's not clear which nine investors purchased the shares originally owned by Singh and Ellison. Representatives for the US Marshals Service didn't immediately respond to a request for comment.

The Anthropic shares owned by Ellison and Singh took a different path than the $500 million in Series B shares owned by Bankman-Fried himself. His shares ended up as part of FTX's bankruptcy estate, which sold them in 2024 to a few dozen buyers for a total of $1.3 billion.

The federal government sent the proceeds from the sale of Ellison's and Singh's shares to the FTX bankruptcy estate, the Justice Department spokesperson said. Bankruptcy court filings show that the Justice Department transferred $638 million to the bankruptcy estate in late 2025.

The money has gone to repaying creditors, including FTX customers and investors who were scammed by the former crypto wunderkind.

Update: October 26, 2026 — This story has been updated to include a comment from the Justice Department.

Correction: October 27, 2026 — The federal government transferred the proceeds of the sale of Caroline Ellison's and Nishad Singh's Anthropic shares to the FTX bankruptcy estate, according to a DOJ spokesperson. An earlier version of this story misstated whether the funds had been transferred.