Field Notes
Big AI Lab Revenues Projected Out from May 2026 to 2035 in Takeoff Scenarios
Bamboo at a shrine outside Kyoto, Japan. April 2026
Recently it was leaked somewhere that Anthropic was at $44 billion in ARR. Then Brad Gerstner implied on the All-In Podcast last week that OpenAI should be around $35 billion by saying that they were at a combined $80 billion in revenue.
I’m going to go with Zephyr’s assumption that we get to $200 billion ARR for the big labs by the end of the year.
For posterity, let’s just make a table of AI lab revenue projections based on various scenarios. Let’s call them:
- “Conservative”: current track but kind of looks like a typical software company.
- “Base Case”: agents become the enterprise software/labor layer.
- “AI-2027 / Situational Awareness”: it’s a fast-ish takeoff scenario.
| Year | Conservative OpenAI | Conservative Anthropic | Base OpenAI | Base Anthropic | AI-2027 / SA OpenAI | AI-2027 / SA Anthropic |
|---|---|---|---|---|---|---|
| 2026E | $89B | $111B | $89B | $111B | $89B | $111B |
| 2027 | $142B | $178B | $177B | $223B | $222B | $278B |
| 2028 | $198B | $250B | $310B | $390B | $498B | $627B |
| 2029 | $258B | $324B | $481B | $604B | $947B | $1,191B |
| 2030 | $310B | $389B | $673B | $846B | $1,563B | $1,964B |
| 2031 | $356B | $448B | $875B | $1,100B | $2,266B | $2,848B |
| 2032 | $399B | $501B | $1,067B | $1,342B | $2,945B | $3,703B |
| 2033 | $439B | $551B | $1,238B | $1,556B | $3,534B | $4,443B |
| 2034 | $474B | $596B | $1,387B | $1,743B | $4,065B | $5,110B |
| 2035 | $507B | $637B | $1,525B | $1,918B | $4,552B | $5,723B |
Now, you might think it’s insane to talk about single-year revenue for a single company in the trillions, but I assure you if you read various projections from the AI 2027 paper or Situational Awareness, this has been speculated upon for a few years now.
Combined Revenues
It’s a little cleaner to look at this as combined revenue since you really don’t necessarily care who wins; you just want to quantify the magnitude of the winning.
| Year | Conservative | Base | AI-2027 / Situational Awareness |
|---|---|---|---|
| 2026E | $200B | $200B | $200B |
| 2027 | $320B | $400B | $500B |
| 2028 | $448B | $700B | $1,125B |
| 2029 | $582B | $1,085B | $2,138B |
| 2030 | $699B | $1,519B | $3,527B |
| 2031 | $804B | $1,975B | $5,114B |
| 2032 | $900B | $2,409B | $6,648B |
| 2033 | $990B | $2,794B | $7,978B |
| 2034 | $1,069B | $3,130B | $9,175B |
| 2035 | $1,144B | $3,443B | $10,276B |
Potential Market Cap / Valuations
If you put even vaguely software-like multiples on these revenue numbers, the valuation outcomes get silly very quickly. The exact multiple obviously compresses as the numbers get bigger, but even with compression you end up with outcomes that are hard to compare to today’s public market.
For a current anchor: OpenAI officially closed at $852B post-money in March 2026, while Anthropic officially closed at $380B post-money in February 2026. Reuters has also reported that Anthropic is considering a raise at over $900B to nearly $1T, so the live combined mark is probably somewhere between $1.23T closed/official and $1.75T to $1.85T rumored/current.
| Year | Conservative valuation | Base valuation | AI-2027 / SA valuation |
|---|---|---|---|
| 2026E | $1.8T | $2.0T | $2.4T |
| 2027 | $2.6T | $4.0T | $6.0T |
| 2028 | $3.1T | $6.7T | $12.9T |
| 2029 | $3.5T | $9.8T | $22.4T |
| 2030 | $3.8T | $12.9T | $33.5T |
| 2031 | $4.0T | $15.8T | $43.5T |
| 2032 | $4.3T | $18.1T | $49.9T |
| 2033 | $4.5T | $19.6T | $55.8T |
| 2034 | $4.5T | $20.3T | $59.6T |
| 2035 | $4.6T | $20.7T | $61.7T |
When do they overtake the big tech companies?
Current annualized baselines:
- Alphabet $440B
- Microsoft $332B
- Meta $225B
- Amazon $726B (we should note this is inflated by the commerce side of things)
Alphabet reported Q1 revenue of $109.9B, up 22%; Microsoft reported Q3 FY26 revenue of $82.9B, up 18%; Meta reported Q1 revenue of $56.31B, up 33%; Amazon reported Q1 net sales of $181.5B, up 17%.
| Scenario | Company | Meta | Microsoft | Alphabet / Google | Amazon |
|---|---|---|---|---|---|
| Conservative | OpenAI | No by 2035 | No by 2035 | No by 2035 | No by 2035 |
| Conservative | Anthropic | No by 2035 | No by 2035 | No by 2035 | No by 2035 |
| Base | OpenAI | 2029 | 2030 | 2031 | 2035 |
| Base | Anthropic | 2028 | 2029 | 2030 | 2032 |
| AI-2027 / SA | OpenAI | 2028 | 2028 | 2029 | 2030 |
| AI-2027 / SA | Anthropic | 2028 | 2028 | 2028 | 2029 |
Implications
Needless to say, if we really have $10 trillion in annual revenue accrued to AI companies in 10 years, there is a lot of stuff both upstream and downstream of them that will reprice. The obvious stuff right now is chips, memory, power, and datacenters, which I wrote about a lot previously, including here, here, and here.
If you take the bull case seriously, these labs are probably each a 10x by 2030, so on a roughly five-year timeframe. The annoying part is that you cannot cleanly buy the labs directly in the public market, so you have to think in proxies. For Anthropic, possible public-market proxies include SK Telecom, Zoom, Google / Alphabet, and Amazon. For OpenAI, the list is more like Microsoft, AMD, CoreWeave, and SoftBank, which says its cumulative OpenAI investment should represent about 13% ownership after completion. These are all imperfect exposures, but that is the game while the actual labs remain private.
Notes
I will add to this as more information comes out on revenue figures.