Anthropic at $900 Billion. The Valuation Just Lapped OpenAI.

Anthropic is days away from closing a $50 billion funding round at a $900 billion valuation. That number is not a typo, and it is not a forward projection. It is the price tag investors are right now writing checks against, and it puts Anthropic ahead of OpenAI by valuation for the first time in either company's history.

Three months ago Anthropic closed a $30B round at $380B. The new round more than doubles that mark. Board approval is expected at a meeting this month, with an IPO window opening as early as October. We have been watching the numbers run for a while. Here is what changed.

The Round Itself

Reporting from Bloomberg and TechCrunch over the last week pegged the round at roughly $50B in new capital, with the valuation sitting between $850B and $900B depending on which source you read. By the weekend the number had hardened around $900B, with sources noting investor demand could push it higher before the deal is signed.

The most striking detail is the timeline. Investors were reportedly given a 48-hour window to submit allocations on or around May 1. That is not a normal pace for a deal of this size. It tells you two things: the round is heavily oversubscribed, and Anthropic is moving fast enough that it is willing to leave money on the table to lock the cap table before the IPO conversation gets serious.

Read top to bottom: Anthropic's valuation has roughly 22x'd in 21 months. Most of that gain happened in the last six. The compounding shape here is closer to a hardware company riding a hyperscaler tailwind than to a typical software business. There is precedent (NVIDIA itself between 2023 and 2024), but precedent is a short list.

The Revenue Has Caught Up to the Hype

Valuations of this size used to be a faith trade. Anthropic is no longer in that category. The revenue numbers leaked over the last month genuinely justify a frontier-leader multiple, and the trajectory is still vertical.

ARR went from $9B to a reported $44B in five months. That is a 4.9x in less than half a year. For context, the comparable five-month stretch for OpenAI in the post-ChatGPT period was about 2x. This is not a typical growth curve for a company this size, and it is the part of the story that is most defensible against gravity.

At a $900B valuation against $44B ARR, the implied multiple is roughly 20x. That is high in absolute terms, but for context: NVIDIA traded at a 25x revenue multiple at peak. OpenAI's last round priced it at roughly 32x. If you believe the run rate holds for two quarters, Anthropic is actually the cheaper of the two frontier labs by this measure.

Where Did the Revenue Come From

Three drivers, in rough order of contribution.

First, Claude Code. The internal coding agent rolled out paid tiers in late 2025 and turned into one of the most-used coding products in the developer tool category by early 2026. The revenue share for Claude Code itself is not public, but the Anthropic team has been pointing at it as the primary driver of Q1 acceleration in conversations with investors. We track the harness side of that story on our harnesses page.

Second, enterprise API at the high end. Reporting from Sacra and the company itself indicates customers spending $1M+ per year on Claude have grown from a dozen two years ago to over 500. Customers spending $100K+ per year are up 7x in the last year. That is a thick middle of the curve, not just whales.

Third, distribution muscle that did not exist 18 months ago. Claude is now first-party on AWS Bedrock, Google Vertex, and Snowflake. The Amazon and Google compute partnerships funnel enterprise procurement into Anthropic by default. When AWS sales reps walk into a Fortune 500 and pitch generative AI, the path of least resistance is Bedrock plus Claude.

The Compute Math

The other thing the new round funds is silicon. Anthropic announced two compute deals in the last six weeks that, taken together, lock in roughly 10 gigawatts of capacity coming online through 2027.

The Amazon deal is up to 5GW for training and inference. The Google plus Broadcom deal added another 5GW, anchored by Anthropic-tuned TPU capacity through Google Cloud. The headline investment commitments include up to $25B from Amazon and up to $40B from Google ($10B immediate, $30B contingent on milestones).

For context, 10GW of training capacity is roughly the entire installed base of frontier AI training infrastructure as of mid-2024. Anthropic alone has now contracted for that much over the next 18 months. The new $50B round is the company's share of bringing that capacity online and paying for the training runs that fill it.

What This Does to OpenAI

OpenAI's last round priced the company at roughly $800B against $122B raised. That deal was anchored by Amazon, NVIDIA, SoftBank, and Microsoft. The new Anthropic mark prices ahead of that, on less raised capital and (as of April) higher ARR.

The strategic implication is more interesting than the headline. For two years OpenAI was the default frontier lab in every enterprise AI conversation. Anthropic was the safety-conscious alternative. The conversation has flipped. Today Anthropic is the revenue leader, the valuation leader, and (as of last week) the only frontier lab that walked away from a Pentagon contract on principle. Read our coverage of that decision for the connecting story.

OpenAI is still bigger by user count (900M weekly actives on ChatGPT) and consumer brand. But on the metrics that drive enterprise AI procurement (revenue per customer, retention, infrastructure spend, enterprise distribution agreements), Anthropic now leads. That gap is the story for the next 12 months.

The IPO Question

October 2026 is the IPO window the board is reportedly evaluating. If they file in October at the current valuation, Anthropic would be the largest tech IPO since Saudi Aramco in 2019. The public markets pricing for an AI company at $900B+ is uncharted. The closest comparable is NVIDIA in 2023, and NVIDIA had four decades of public-company financial discipline before that re-rate.

The bear case is that an IPO at this valuation requires Anthropic to keep growing 4x annualized for at least four more quarters to grow into the multiple. That is an aggressive bet on a company that just turned three.

The bull case is that AI infrastructure spend is still front-loaded, the enterprise revenue ramp is accelerating not decelerating, and the compute deals lock in cost advantages that competitors will spend years matching. Both cases are credible. Pick your priors.

Our Take

The $900B headline is the easy story. The interesting story is the speed. A frontier lab going from $9B to $44B ARR in five months is something the public markets have never priced before. A company more than doubling its valuation in 90 days while shipping models, signing 10GW of compute, and walking away from defense contracts on principle is the kind of compound action that breaks valuation models.

We will be watching three things over the next 60 days. First, whether the round actually closes at $900B or pushes higher on demand. Second, whether the May ARR figure is real once third parties get a look at it (Sacra, The Information, anyone with a finance source). Third, the IPO filing date, which will tell you what the board actually thinks the long-term clearing price is.

We track the AI funding and pricing landscape on our models tracker and Anthropic provider page. The numbers update as the deals close. If you want the agent-readable feed, it lives at our developers page.

One last thing. A year ago the consensus was that the AI industry would consolidate to two frontier labs, and that one of those would be much larger than the other. The consensus is half right. There are two. They are within striking distance of each other, and the second one just lapped the first.