Back in the spring, I put together a list of who's actually making money from the AI boom, and I said at the time that net worths at this level move faster than I can update a blog post. That turned out to be an understatement. A few people asked for a follow-up, so here it is: same cast, same order, four months later (as of September 19th).

Jensen Huang (Nvidia) - Nvidia briefly hit a $5 trillion valuation this fall, pushing Huang's net worth to around $180 billion, and Huang has spent the season doing actual diplomacy: flying to Beijing with the Trump delegation to talk chips with Xi Jinping after initially being left off the list.

The China question: How does a market that used to be worth $17 billion a year to Nvidia and is now functionally zero under export controls still remain what analysts watch most closely, even with $500 billion of order backlog on the books through next year?

Andrew Feldman & Sean Lie (Cerebras Systems) - They actually did it: Cerebras IPO'd in May at $185 a share, popped 68% on day one, and raised over $5.5 billion in what's been called the biggest semiconductor IPO ever. Both founders are now comfortably billionaires (Feldman around $3.1B, Lie around $1.6B), and the stock has kept climbing since, up double digits in a single September session, plus a new Finland data center and a Callosum partnership for good measure.

Larry Ellison (Oracle)- Still rich off the OpenAI Stargate deal, but Oracle's fall has been messier than the spring made it look. Reports surfaced of internal layoff lists being drawn up even as the company posts record cloud revenue, because building AI infrastructure at this pace is burning cash faster than the contracts pay it back. Oracle publicly denies any delay to OpenAI's data centers; investors clearly aren't fully convinced, given how much the stock has swung on capex questions alone.

Elon Musk (xAI /Tesla/SpaceX) - This is the one that actually changed shape. SpaceX-xAI, merged at $1.25 trillion back in February, went public on the Nasdaq in June at a $1.77 trillion valuation, popped over 20% on debut, and briefly made Musk the first trillionaire in history. Four days later, flush with newly public stock, SpaceX bought Cursor-maker Anysphere outright for $60 billion, turning what I'd described as an "option" into an actual deal. Musk's net worth has since been doing exactly what you'd expect: it fell below $700 billion in July, then rebounded to nearly $900 billion in August.

Jeff Bezos (Project Prometheus) - Barely existed when I wrote the original post; now it's fully real. Project Prometheus closed a $12 billion Series B in June at a $41 billion valuation, up from $6.2 billion eight months earlier with Goldman, BlackRock, and JPMorgan all writing checks. The pitch is still the same: not a chatbot, an "artificial general engineer" that designs chips, jet engines, and drugs, trained on physics and experimental data rather than internet text.

Dario & Daniela Amodei (Anthropic) - Anthropic filed confidentially for an IPO, first among the frontier labs, which surprised people given OpenAI's head start after a $65 billion round in May pushed its valuation to $965 billion, edging past OpenAI for the first time. It's targeting a listing this fall, possibly this month, with revenue run-rate having gone from $9 billion to $47 billion in under half a year.

Sam Altman (OpenAI) - Filed for IPO a few days after Anthropic, initially targeting September at up to $1 trillion, then quietly pushed the date to 2027. In the meantime, secondary markets have been pricing OpenAI shares at roughly a 44% discount to what people paid in the March primary round. Altman is still drawing his $76,001 salary and still holds no equity in the company. The structurally strangest part of this whole list remains exactly as strange as it was in spring.

Ilya Sutskever (Safe Superintelligence) - Had a genuine soap-opera summer. Co-founder Daniel Gross left at the end of June, reportedly to talk to Meta (Zuckerberg apparently explored buying all of SSI at one point), and Sutskever stepped into the CEO chair himself. Nvidia doubled down anyway, putting in $5 billion and promising a 10x compute boost, a vote of confidence in a company that still has no public product.

Mira Murati (Thinking Machines Lab) - Rockier than expected: she fired her CTO, Barret Zoph, in November over what sources described as "unethical conduct," with Soumith Chintala taking over. Despite that, Thinking Machines is reportedly in talks to raise at a $50 billion valuation, five times where it stood in the spring, with Nvidia now a strategic partner on top of being an investor.

Ioannis Antonoglou & Misha Laskin (Reflection AI) - The round I described as a rumor actually closed, and closed bigger: $2.5 billion at a $27.5 billion valuation, above even the $25 billion figure being floated at the time. They've since signed an MOU with the Department of Energy to power the Genesis Mission across 17 national labs and partnered with Dell on an AI-factory product, a fast pivot from "coding agent startup" to "national infrastructure vendor."

Bret Taylor & Clay Bavor (Sierra) - Raised $950 million in May at a valuation north of $15 billion, up from $4.5 billion a year earlier, and crossed $200 million in revenue not long after. Sierra now claims over 40% of the Fortune 50 as customers and has been on an acquisition spree of its own, picking up a French startup (Fragment) and a Tokyo-based enterprise AI company (OPERA TECH) along the way.

Arthur Mensch, Guillaume Lample & Timothée Lacroix (Mistral AI) - Europe's answer to the American labs is reportedly in talks for a round near €20 billion, roughly double where it sat a year ago, with a second dedicated data center now planned for Sweden on top of the one outside Paris. Revenue is on track for roughly €1 billion by year-end, up from $16 million at the end of 2024.

Winston Weinberg & Gabe Pereyra (Harvey) - Raised $550 million at a $15.6 billion valuation, about 42% higher than where they were in March, on the back of ARR that's roughly doubled over the same period. Their client base went from 1,300 law firms in March to more than 3,000 by September, and they picked up an AI-safety startup, Guardrails AI, as their fourth acquisition of the year.

Ali Ghodsi (Databricks) - Kept re-pricing every few months upward: $134 billion in February, talks of $165–175 billion by June, reportedly landing closer to $188 billion by midsummer. Ghodsi has been publicly candid that this is "a terrible year" to go public given how crowded the IPO calendar already is with SpaceX, Anthropic, and OpenAI all jostling for the same window, though he's told investors a 2027 listing remains on the table.

Scott Wu, Steven Hao & Walden Yan (Cognition) - The wildest trajectory on this entire list. Cognition (maker of the Devin coding agent) went from a $10.2 billion valuation in September to $26 billion in May to $48 billion by September, nearly quintupling in a year, as revenue climbed from $37 million to almost $1 billion annualized. Devin now reportedly writes over 90% of Cognition's own code.

Daniel Nadler (OpenEvidence) - The "ChatGPT for doctors" company doubled its valuation to $12 billion in January, taking Nadler's own net worth to roughly $7.6 billion. Clinical queries on the platform have gone from 10 million to 16.5 million a month since last summer, and Nadler's stated next act is, in his own words, "medical superintelligence", a bigger ambition than the note-taking tool this started as.

Michael Truell, Sualeh Asif, Aman Sanger & Arvid Lunnemark (Anysphere / Cursor) - No longer an independent company. SpaceX exercised its option and bought Cursor outright for $60 billion in June, days after its own IPO, converting what had been a standing acquisition option into a done deal. Cursor's revenue had reportedly reached $2.6 billion annualized by the time the deal closed, a premium well above the roughly $50 billion valuation it had been raising independently.

Edwin Chen (Surge AI) - Stayed exactly as quiet as the "boring infrastructure layer wins" framing suggested it would. No major new funding news surfaced this season, which for a data-labeling company that was already doing over $1 billion in annual revenue without needing to make headlines is arguably the whole point.

Colette Kress & Jay Puri (Nvidia) - Their fortunes track the stock, and the stock has mostly gone one direction: up, alongside Nvidia's run toward a $5 trillion-plus market cap this fall. Both remain the clearest example on this list of "stay at the right company long enough" as its own form of founder-level return, no new company required.

Aravind Srinivas (Perplexity) - Tripled revenue to roughly $750 million, and Nvidia is reportedly in talks to invest at a $30 billion valuation, up 50% from a year ago. I'll admit I underestimated Srinivas's staying power the same way a lot of people did back when Perplexity looked like an also-ran against OpenAI and Google.

Alexandr Wang (Scale AI) - Now Meta's Chief AI Officer, running Superintelligence Labs, with Zuckerberg personally offering nine-figure packages (at least one reported case over $200 million) to poach researchers from OpenAI, Google, and Apple. Whether Meta's superintelligence push pays off is genuinely unresolved; what's not unresolved is that it's the most expensive AI talent war anyone has run.

Liang Wenfeng (DeepSeek) - Became, per Bloomberg, the single wealthiest AI-model founder on the planet at roughly $36 billion, funded entirely by a quant hedge fund most people outside China have never heard of. DeepSeek has reportedly raised a further 70 billion yuan and is now hiring for coding-agent products, a sign of where the company thinks the next fight is.

Guillermo Rauch (Vercel) - Revenue run-rate went from $100 million to $340 million in about fourteen months, and Rauch has been publicly signaling IPO readiness, calling Vercel "a work-in-public company." A broader software sell-off has frozen most of the IPO pipeline behind SpaceX, Anthropic, and OpenAI, so Vercel's debut is likely waiting on that window to reopen rather than on the company itself being ready.

Yan Junjie (MiniMax) - Still under-covered relative to Liang Wenfeng's DeepSeek, and that gap hasn't closed this season. MiniMax remains the quieter half of the "Chinese AI billionaires the English-language press mostly treats as an afterthought" story I flagged in the spring.

Shanka Jayasinha (Onto AI) – The ‘hybrid architecture’ lab filed a Form with the SEC on September 8 showing a $100 million round that was fully subscribed. Although rumored to be around the $3B mark from a few references online, no public statement was made around the valuation from the company yet.

Amjad Masad (Replit) - Raised $400 million at a $9 billion valuation in March, tripling in six months, and crossed into billionaire territory himself in the process. Revenue is tracking toward $1 billion annualized by year-end, up from $2.8 million reported in 2024, on the strength of Agent 4 and its canvas-based "vibe coding" pitch.

Brett Adcock (Figure AI) - Sits at a $39 billion valuation and has been busy demonstrating an eight-hour autonomous robot work shift on its Helix 2 neural network. Adcock's stated bet is that whoever ships the first humanoid capable of five hours of unsupervised work in an unfamiliar home "becomes the largest company in the world". Still a prediction, not yet a fact, but he's clearly not backing off it.

Naveen Rao (Unconventional AI) - Left Databricks to build what he calls "brain-scale efficiency without the biological baggage," and was in talks for $1 billion at a $5 billion valuation with a16z leading. It's the most quietly ambitious bet on this list: rethinking the physical substrate of computing itself, rather than what runs on top of it.

Fei-Fei Li (World Labs) - Raised $1 billion from Autodesk, a16z, Nvidia, and AMD to keep pushing on "spatial intelligence", AI that understands three-dimensional space rather than flat text. It's a similar wager to Bezos's Prometheus in spirit, if not in specific target: physical-world reasoning.

Surya Midha, Brendan Foody & Adarsh Hiremath (Mercor) - Now being talked about at a $20 billion valuation, up from $10 billion in the fall and $2 billion the spring before that. All three are still in their early twenties, and Forbes has confirmed what the growth curve implied: they're the youngest self-made billionaires in the world.

And the honorable mentions, since a few of them had bigger seasons than the main list:

Demis Hassabis (Google DeepMind) - Got promoted, not fired: in August he moved up to Chair of Google DeepMind and Chief Scientist of Alphabet, handing day-to-day Gemini operations to Koray Kavukcuoglu. It's being read across the industry as a signal that senior researchers are choosing longer-horizon science over product-release clocks. Yann LeCun, notably, has been the loudest voice making that case publicly.

Yann LeCun (Advanced Machine Intelligence) - Left Meta in late 2025 after twelve years as Chief AI Scientist and became Executive Chairman of AMI Labs in Paris, which closed a roughly $1.03 billion seed round in March at a $3.5 billion valuation, reportedly Europe's largest seed round ever. He's building world models on JEPA architecture, still betting that scaling language models isn't the whole path to real intelligence.

Anton Osika & Fabian Hedin (Lovable) - Went from a $1.8 billion valuation last spring to $13.3 billion by September, with revenue climbing from $200 million to roughly $600 million annualized along the way. Both founders are now dollar billionaires and have said they intend to give away half their fortunes, a nice echo of the Amodei siblings' pledge from the original list, from a very different corner of Europe.

Aidan Gomez (Cohere) - No longer flying solo: Cohere is merging with Germany's Aleph Alpha in a deal expected to value the combined "sovereign AI" company at around $20 billion once it closes, nearly triple where Cohere sat on its own last September.

So: still close to 100 billionaires, still a few trillion dollars combined, and the picks-and-shovels pattern hasn't budged: chips, data infrastructure, data labor, developer tools, and increasingly physical-world engineering are still where the money concentrates, not the chat interfaces most people actually touch. What's changed since spring is the mechanism. Four months ago, this was a story about private valuations doubling every few months.

Now, it's a story about those valuations hitting public markets with real daily prices attached and at least one of them, OpenAI's, immediately trading at a discount to what people paid privately six months earlier. That's a different kind of surprise than the one I was writing about last time, and it's the one worth watching heading into whatever comes next.