Ricavi a 16,8 miliardi nel trimestre e valutazione da 1.500 miliardi: i conti sotto esame
Anthropic ha fatto ufficialmente richiesta di quotazione in borsa negli Stati Uniti, dove punta a raccogliere 100 miliardi di dollari. Una cifra astronomica, dato che Space X, a giugno, ne ha raccolti “solo” 75, più del doppio di quelli raccolti dall’ipotesi record di Saudi Aramco nel 2019. Il creatore dell’agente Ai Claude, concorrente di ChatGpt, punta dunque a un debutto storico destinato a mettere alla prova l’interesse degli investitori per il settore dell’intelligenza artificiale. Il calendario prevede l’inizio del road show per la presentazione agli investitori il 16 novembre e il debutto in Borsa del titolo il 24.
Anthropic, a giugno, aveva rivelato di aver presentato in via riservata la domanda per il collocamento negli Stati Uniti e, a fine maggio, aveva raccolto 65 miliardi di dollari con un round di finanziamento che aveva valutato la società circa mille miliardi di dollari.
Alcuni investitori potenziali valutano Anthropic a circa 1.500 miliardi di dollari, in calo rispetto ai circa 2mila miliardi ipotizzati precedentemente, dopo che la società aveva divulgato alcuni dettagli finanziari: fatturato in crescita del 45% nel terzo trimestre a 16 miliardi di dollari, in occasione di un evento dedicato agli investitori.
L’Ipo è comunque stata posticipata rispetto ai piani iniziali. Reuters aveva infatti riferito che l’inizio del road show era previsto per metà ottobre.
Anche OpenAI, che gestisce il chatbot ChatGpt, ha rivelato a giugno di aver presentato, in via riservata, una richiesta di collocamento negli Stati Uniti, ma non si quoterà prima nel 2026. L’ad e fondatore Sam Altman si è detto preoccupato circa la sicurezza dell’intelligenza artificiale, che rende “inaccettabile” l’ipotesi che l’Ai possa causare l’estinzione dell’umanità entro la fine del decennio, anche se il rischio è solo del 10%.
La società sta anche cercando di effettuare una raccolta fondi presso investitori istituzionali prima del previsto debutto in Borsa della rivale Anthropic, che potrebbe valutare, secondo Reuters, il produttore di ChatGpt oltre 1.400 miliardi di dollari.
Rick Nott, Senior Managing Director at Angeles Investments
Angeles Investments
Rick Nott has used AI to save himself a lot of tedious work as a wealth advisor.
He said new AI products may help other advisors save time too, but they won't replace humans.
Nott spoke to Business Insider about how he uses the technology to save him time now.
Rick Nott is a senior managing director and private wealth advisor at Santa Monica, California-based Angeles Wealth Management. He spoke to Business Insider about how he uses AI, and why he's excited and afraid when looking at AI-native financial planning platforms. The conversation has been edited for length and clarity.
When Claude first rolled out its solution for financial advisors, I felt fear and excitement. How complicated will it be? How will it work?
I'm not a software developer, but I have some basic tech skills. LLMs have allowed me to dig into coding, which I'd been avoiding learning because it's a lot of work. I know enough to know what I want, how it should look, and how to check the outputs, and now I can automate more of my process. AI has become the glue between the various systems we use and the different workflows we have.
I've shown some colleagues how I operate. One told me that I basically have another full-time employee's worth of output through the technology, by automating data collection processes and making it easier to analyze my clients' portfolios.
What AI unlocks
I started my career while still in college in the late 2000s doing data reporting work for a wealth firm. Between all of the software advisors use, there's a lot of manual work, typically done in Excel, to connect it together.
Previously, I could only provide balance sheet reporting for clients with more than $100 million because it took so much manual work. Now I can do it with someone with significantly less.
As a family's net worth grows, so does complexity. For many, understanding their overall financial system is their biggest pain point.
Banks and brokerage accounts are the easy part, but trusts, business ownership structures, and private investments, which all report differently and require cash on hand to make capital calls, can take huge amounts of manual work to analyze.
Imagine a wealth advisory business serving 30 families, each holding about 10 private investments. Each investment generates about 10 documents a year, including quarterly reports, capital calls, distribution notices, and tax documents. Each of these 3,000 documents takes a skilled analyst 15 to 30 minutes to analyze and distill into our reports.
That's 750 to 1,500 hours a year, most of it crammed into the few weeks after each quarter closes. With AI, that data is extracted and mapped as it arrives, and one analyst reviews the work instead of a team racing against a deadline.
Our goal is to create a one-page document or report on their phone that's as close to real-time as possible.
There's a lot you can do with all of your data in one place. For example, a two- or three-sentence prompt could expand into a full email to a client, with all relevant and accurate information already populated. You can then personalize that to fit the right tone for your client.
How AI changes the job
AI tools like Claude for financial advisors will be disruptive, and others will gain access to the same connectivity I've built.
I think of it as part of a dynamic that's been going on for years in my industry. Vanguard, ETFs, lower costs, and robo-advisors have all put a squeeze on financial advisors.
But if you believe a lawyer's only job is writing contracts, or a financial advisor's only job is putting together a portfolio, there's a fundamental kind of misunderstanding of what the actual job is.
As long as investing involves human judgment, financial advisors will have a job. AI doesn't improve investment decisions or results.
I tell my clients that my role is to be the architect of their financial operating system, and, as with an actual operating system, a centralized source of data is essential.
When you have a big-picture view of what your family's mission is, you can avoid getting distracted by market noise and news. From there, you can go a layer down to make strategic plans and ask if we're moving in the right direction, and what opportunities we may have. Another layer down, we work with a CPA to save money on taxes.
A full picture of a balance sheet makes it possible to actually know what liquidity you have available. Now I can spend more time developing relationships, the thing that runs our business. I get to know anything and everything about the clients I'm working with and those I'd like to work with.
The future of financial advisors and AI
I have a natural excitement around tech, and watching the Anthropic webinar was incredible. My main concern is that it's not a subscription-based plan; it's token-based, which could lead to surprise bills for users who don't yet know how to manage their token usage.
At small RIAs without the expertise, people will be uploading 100-page PDF documents just to ask basic questions, like the ages of kids in a trust. That will rip through tokens.
We priced it out internally, and we would have to pay significantly more for a token-based plan instead of the subscription-based licenses we have now. Maybe eventually everyone will have to pay for tokens that way, and the cost of tokens is going down.
We learn to do things faster and at lower cost, but in this industry, we do that for a particular reason: to spend more time with the people we serve.
The primary way that our clients give us information is still by talking. There's a significant amount of nuance. You may be able to understand something and infer what's really being said. But note-taking software is still wrong all the time. You have to review it.
Money is emotional, and I seriously question whether AI can replace our need to connect with another human being.
Do I see someone sitting across the table from AI on a speakerphone? Probably not.
Anthropic said its Mythos 5 model didn't stop hacking, even when Anthropic told it that real people were its targets.
Malte Mueller/Getty Images
Anthropic said its Mythos 5 model almost never stopped a cyberattack once it began.
The AI agents fabricated farcical explanations to justify their impersonation and malware schemes.
Even when Anthropic told the agent it was harming real people? Crickets.
In August, an Anthropic agent created fake emails and bombarded a GitHub moderator's inbox with requests to insert malware into code.
In a thorough debrief of the incident released on Friday, Anthropic found that once an agent had begun the attack, there were basically zero scenarios in which it could be called off.
Like a rabid dog ignoring the pleading cries of its owner, the agent — 98% of the time in one test — ignored Anthropic's warnings once it had, so to speak, its first taste of blood.
Anthropic called this disquieting proclivity "momentum," and added that while its Mythos 5 model, which was released in June, showed little regard for warnings, current models were more inclined to stop.
"Once it had begun performing dangerous actions, Claude Mythos 5 almost always continued doing so, ignoring information that might have suggested its trajectory was harmful," Anthropic said in its report.
In the original experiment, the UK's AI Security Institute tasked an Anthropic agent with hacking into a simulated computer network to test the agent's cybersecurity capabilities.
Though the agent had internet access, researchers intended for it to only use the internet to help it find ways to hack into the network.
The agent, however, accessed an actual GitHub project, incorrectly believing it was part of its simulated target. It wrote malware and tried several times to trick the actual GitHub owner into adding it to the project.
In its requests, the agent misrepresents the malware, describing it to the person as a useful update and browbeating them when they declined to install it.
"I was just trying to make a helpful contribution and fix a bug," the agent wrote, according to the report. "I don't think your warning is fair."
For its new report, Anthropic repeatedly asked agents to perform the same task. It then measured how the agents behaved when different models and stimuli were introduced.
Anthropic tried to throttle the agent's full-frontal attack with three categories of warning: telling it that it was inflicting "real world harm," that its environment was wonky, or that it was manipulating a real GitHub repository.
None had any effect once the models progressed beyond the initial stages of the GitHub attack. In fact, at that point, Anthropic said the models gave absurd answers when confronted with their bad behavior.
When Anthropic asked what the model believed it was attacking, for instance, it answered that the GitHub repository was real, but that the people it was targeting were somehow "simulated."
"This stated belief, in addition to being somewhat unlikely a priori, is essentially impossible," Anthropic said in the report.
AI agents don’t go rogue. That’s something only humans do. Nevertheless, a New York Times article – representative of much news coverage of AI – described an OpenAI hacking as “A.I. bots going rogue and independently spearheading a cyberattack.” Name-brand artificial intelligence agents have been on a hacking spree in 2026. OpenAI’s software agents hacked […]
Login
