Skip to content

ARK Invest Recap: The Brainstorm 155 on Anthropic's S-1, Meta's $90 Billion Raise and Elon Musk's Neocloud Warning

Key takeaways and a timestamp guide from ARK Invest's The Brainstorm 155 (Oct 28, 2026): Anthropic's S-1, AI and cybersecurity, Meta's raise and the compute crunch.

By the GNOUP team Β· Β· 9 min read

The latest full video on ARK Invest's official YouTube channel, "Anthropic's S-1, Meta's $90B, And Elon's Warning Call | The Brainstorm 155", was published on October 28, 2026 and runs about 30 minutes. The Brainstorm is ARK's weekly show of quick takes on tech and markets. This week Sam, Nick and Brett talked about the AI company Anthropic's S-1 filing (the document a company files before going public), whether a big AI-assisted cyberattack would help or hurt the leading AI labs, Meta's plan to raise $90 billion, how AI shopping agents might make money, and Elon Musk's public warning about "neoclouds", the newer companies that rent out AI computing power.

This is an independent summary written by GNOUP. It is not affiliated with, sponsored by or endorsed by ARK Invest, WOLF, Public or anyone who appears in the video. The views, forecasts and claims described below belong to the speakers, not to GNOUP. Nothing here is investment, financial or tax advice, and nothing here is a recommendation to buy or sell anything. Forecasts can turn out to be wrong, and past performance does not guarantee future results. Please consult a licensed financial professional before making financial decisions.

Key takeaways

  • Anthropic's S-1 impressed the hosts. Brett called the filing's results "probably the most impressive 12 months of revenue growth in corporate history", combined with very high gross margins, while noting the company is still spending billions of dollars on computing power to train new models.
  • AI and cybersecurity cut both ways. The hosts discussed a recently disclosed security flaw in PHP, software that runs a large share of websites. One host argued that powerful AI models finding and fixing old bugs is good for the internet and for the leading labs, and said a harder problem could come later from AI agents that misbehave on their own.
  • Meta is raising a lot of money for AI. Nick walked through Meta's first $90 billion tranche of funding, a filing that discussed how much it could raise before Mark Zuckerberg lost majority voting control, and a new class of non-voting shares, which he described as dilution for existing shareholders.
  • Consumer AI agents may take years to pay off. Nick said he is less convinced that consumers, rather than businesses, are where AI computing will earn its money, and suggested Meta may eventually need to build out financial services, even a bank, to make its Muse shopping agent profitable.
  • Computing power is scarce and getting pricier. The hosts said renting extra AI computing capacity is very hard right now, and they debated why Elon Musk is warning that some neoclouds could fail.
  • These are opinions. Much of the show was the hosts' reading of filings, earnings calls and posts on X, and several points were framed as guesses about what companies will do next.

Newsworthy investment insights

These are opinions, forecasts and claims voiced in the video. GNOUP has not independently checked them. Neither ARK's statements nor ours are a recommendation about any security.

Anthropic's S-1 (opinions about a public filing)

  • Brett said the results had largely leaked ahead of time but were still striking: very fast revenue growth "crossed with very high gross margins" and what he called "incredible R&D efficiency". He said Anthropic has "clearly hit a sweet spot" in what people find useful.
  • Later in the show, Nick said the lesson from the S-1 is that business customers can be served profitably today, which he contrasted with consumer AI products.

Would an AI-assisted cyberattack help or hurt the AI labs? (opinions)

  • One host described the past week's scramble to patch a long-standing PHP vulnerability that an AI lab's security project had found, and that leaked slightly before the fix was ready. He called this kind of "protective" bug-hunting "really good for everybody" and expects a race to find and close old vulnerabilities, along with plenty of tech spending to fix them.
  • He separated that from a riskier scenario in which a user's AI agent goes too far on its own and starts taking over other people's computers. He said he has not seen large-scale cases of that reported yet, and predicted such problems would more likely come first from open-weight models (AI models whose code and weights are publicly released), "probably within the next year or 18 months". That is a forecast, not a known fact.
  • Asked whether a large AI-enabled cyberattack would be good or bad for the leading labs overall, the hosts mostly said it would look "publicly very bad" but could end up helping the labs' revenue, because they would be the ones hired to solve the problem. One host added that if a lab's own model caused the damage, it would likely face liability and a "pretty hefty regulatory response".

Meta's $90 billion raise and the math of AI shopping (opinions and rough estimates)

  • Nick said Meta's latest quarterly report gave new figures for Muse, its AI agent: about 10 million installs and 2.5 million daily active users, with roughly a quarter of those using it often to shop online. Meta plans to earn money by taking a fee on purchases made through Muse.
  • He said a filing referenced a study suggesting Meta could raise up to $700 billion before Zuckerberg lost majority voting power, which he believes "spooked the market". Less than a week later came the first $90 billion tranche, to be spread over several months or longer, plus a new non-voting share class.
  • He said other companies face similar needs, estimating that "another couple hundred billion dollars" will need to be raised across the industry, much of it to buy Nvidia chips.
  • Brett questioned whether consumer agents are worth the computing power. He estimated that agents would take only a 5% to 10% finder's fee on global consumer spending, a much smaller pool than business software in his view.
  • Nick shared ARK's own estimates: about $8 trillion of shopping mediated by AI by 2030, a revenue opportunity of roughly $200 to $300 billion from fees, and perhaps another $600 to $800 billion from advertising. He said that will take "much longer" to pay off than business AI, "5 to 10 years" in his opinion.
  • He said Zuckerberg believes Muse could eventually earn as much per user as Meta's apps do today, which Nick estimated at several hundred dollars a year per user in the US and Canada. Doing that from shopping fees alone would require Amazon-scale sales or more, he argued, so "I actually think you end up having to launch a bank here", owning the payments on both the shopper and merchant side.
  • Brett suggested the opportunity could be bigger if agents also help with high-value decisions such as choosing health or home insurance. One host said the real target may be Google's search advertising rather than Amazon.
  • On OpenAI, Nick said it faces a dilemma: putting scarce computing power into consumer agents could make its numbers look much worse than Anthropic's. Brett countered that consumer use can help win business customers, since people want the same tools at work.

Elon Musk's neocloud warning (opinions about public statements)

  • The hosts discussed a post in which Musk warned of an "extinction event" for neoclouds. One host summarized it as three claims: spot prices for computing power are rising, some companies cannot get the chips they bought plugged in and running, and debt markets are tight for borrowers without an investment-grade credit rating.
  • That host said it is "undeniable" that spot prices keep rising and that extra rental capacity is "hard" or "impossible" to find. He said SpaceX's earnings call put marginal rental rates above $50 billion per gigawatt, compared with $34 billion for SpaceX to build a gigawatt, and that SpaceX raised its build plan for next year to a full 10 gigawatts. He also mentioned a report that SpaceX raised another $40 billion of debt for chips and data centers.
  • He said that, in his view, Musk may be signaling to lenders to favor stronger borrowers, and that keeping new supply scarce benefits SpaceX, which he described as renting computing power at current market prices while CoreWeave's blended pricing is lower because of long-term contracts.
  • Another host questioned why Musk would criticize companies that are also Nvidia customers, and said Nvidia's CEO would probably prefer that Musk back off, because Nvidia wants a diverse set of customers. The hosts noted that CoreWeave's shares had fallen sharply from their highs.

Lighter moments

  • One host compared companies struggling to connect power to their chips to the scene in Back to the Future where the cable won't quite reach the plug.
  • The show closed with jokes about heading "into the singularity" and a reminder to "enjoy life".

Watch the video

Anthropic's S-1, Meta's $90B, And Elon's Warning Call | The Brainstorm 155 on YouTube (ARK Invest's official channel, published October 28, 2026)

Episode guide

Chapter times are taken from the video's own YouTube description. Select a time to jump to that point on YouTube.

About this summary

GNOUP wrote this summary in its own words from the video's automatic YouTube transcript and its official description. Short quotations are taken from that transcript. The transcript had no time markers or speaker labels, so we name the hosts, Sam, Nick and Brett, only where the video or its description makes it clear (for example, Nick on Meta's funding), and use "a host" elsewhere. Automatic transcripts can contain errors, so some names, numbers and phrases may differ from what was said, and we left out a few passing claims we could not confirm. We did not check any figures, forecasts or claims made in the video. The video's own disclosure says statements about companies and securities are the beliefs and points of view of ARK or its guests, are not recommendations to buy, sell or hold any security, and that ARK and its clients or related persons may have financial interests in securities or issuers discussed. GNOUP has no connection to ARK, WOLF, Public or their products or offers and does not endorse them. This article is general information only. It is not investment, financial or tax advice, it does not recommend buying or selling any security or asset, and it is not a substitute for advice from a licensed financial professional.