Today’s big story: Anthropic raised $13 billion at a valuation of $183 billion. This comes a month after raising $5 billion at a $170 billion valuation.
Here are my thoughts:
1. Generative AI is here to stay. Like many others, I was initially skeptical/wrong, but these products have cemented themselves in the US economy. They are not a fluke or a fad, but permanent, much in the same way the World Wide Web, search engines, contextual advertising, personal computers, electric cars, apps, social networking, smartphones, and e-mail have proven to be. Millions of people daily use these tools, whether for coding, writing, generating images, or other tasks, similar to using Google or checking one’s Facebook account.
2. No clear winner. The AI market differs from past tech bubbles in that it can support a wide range of entrants, each commanding very high valuations. Unlike earlier eras dominated by duopolies or winner-take-all dynamics–such as Google vs. Yahoo, the Apple App Store vs. Google Play, or MySpace vs. Facebook–AI has not produced a single dominant player while forcing others into collapse. In prior cycles, losers often imploded, as happened with Research In Motion when it lost out to Apple. So far, nothing comparable has happened in AI.
Past bubbles were concentrated around a few interchangeable companies. By contrast, today’s AI and LLM market features many distinct players–Cursor, Windsurf, Cline, Claude, ChatGPT, Grok, Gemini, and others–each with different strengths and weaknesses and filling some sort of niche, with no obvious or clear-cut winner. For this reason, it’s not uncommon for people to have subscriptions for multiple companies, to fill in these gaps. This means the market can support many players.
With search, if you use Google there is little reason to use Yahoo or Bing, which are clearly inferior. But with LLMs, agents, and assistants, no single product is overwhelmingly better than the rest. That’s why treating AI as just another repeat of past tech bubbles–where a handful of firms like Apple, Facebook, Google, or Microsoft captured nearly all the marketshare while others disappeared or became worthless–is possibly wrong.
3. High interest rates = no problem. A common narrative in 2022-2024 was that high interest rates would lead to a collapse of valuations. Nothing of the sort has happened. Valuations have continued to surge despite high interest rates. Like above, this again shows the pitfalls of looking at AI from the lens of past bubbles or cycles. It’s not that rising interest rates cause recessions; rather, high rates typically reflect a strong economy, but that strength is independent of when the cycle ends. In other words, recessions often occur when interest rates are high, but this doesn’t imply high interest rates caused the recession.
Think about it from the perspective of a VC: If someone is expecting to make a significant real return (e.g. >50-100%) by investing in AI, dotcoms, etc. or whatever the hot sector is–does the difference between 1% or 5% interest rates factor much into this decision making process? Probably not. When greed and short-sightedness takes over, the difference between a few percent of the risk-free borrowing rate is immaterial.
4. Burn to earn. VCs are not as dumb or naive as often assumed by the media or laypeople, who dismiss such bets as irrational recklessness. Pundits said the same about Open AI in 2023, only valued at $30 billion at the time, and then seemingly overnight Chat GPT became a major commercial product rivaling Google, and now even calling into doubt the entire future of higher education due to widespread cheating. Or how Tesla went from being a niche brand to Tesla cars and trucks suddenly everywhere after Covid. These VCs are not as irrational as commonly assumed. They know if a product gains critical mass with the general public, it can go from being a niche to being everywhere.
Similar dismissive arguments were made for the apparent irrational valuations of Amazon, Tesla, Uber and others; then those companies began making billions. Uber lost money for 15 years, until making a billion dollars profit for 2023. The business model of raising billions in the attainment of this critical mass, has proven to work. Or in math/physics parlance, it’s like there is a phase shift or transition where a company that was otherwise losing money many years, at the flip of a switch, starts making billions.
5. Sam was right. FTX creditors should be seeing red for getting such a raw deal. As a consequence of the liquidation of the remaining assets of FTX following the arrest of Sam Bankman-Fried, FTX’s 8% stake in Anthropic, acquired for $500 million in 2022, was sold for just $900 million in March 2024. The timing could not have been worse. The FTX bankruptcy trustee forfeited $10-19 billion in potential profits by selling the Anthropic stake too early.
Following his arrest and extradition, SBF maintained his innocence, insisting that FTX still had sufficient funds. I believed him, even when nearly everyone else was convinced he was lying and that FTX had no assets left. By January 2023, $5 billion had already been recovered. Eventually, all FTX creditors were repaid in full–albeit at 2022 valuations–just as Sam had insisted. Technically, he hadn’t lied about FTX having the money. The collapse was set off when CZ’s tweets triggered a run on FTX’s assets at the worst possible moment, compounded by the platform’s own mismanagement and fraud.
From both a game-theoretic and legal standpoint, it made little sense for him to lie outright about something that could be easily verified. In high-stakes cases, even defendants facing severe charges typically remain silent rather than make unverifiable claims. And yet, billions were eventually recovered and distributed to creditors as he had promised. Sometimes, the person most responsible for the disaster is also the one most capable of fixing it. Instead, the FTX bankruptcy trustee made $1 billion in fees while creditors got in hindsight a poor deal.