Anthropic is negotiating with Nvidia to bring the chipmaker in as anchor investor for an initial public offering that would raise as much as $100 billion and value the company near $2 trillion, according to people familiar with the talks cited by Reuters. If the deal holds, it would be the largest IPO in history by a wide margin, and it would deepen a commercial relationship that already binds the two companies together.
Nvidia is considering an investment of up to $10 billion in the offering, one source told Reuters. The plans remain fluid, and both companies have stayed quiet: Anthropic declined to comment, and Nvidia did not respond to requests for comment. Timing has moved too. Reporting earlier this month indicated the listing slipped from October to November so that the prospectus can include third-quarter financials, a window that closes ahead of the US midterm elections.
Why Nvidia would write the check
An anchor investment is not charity. Nvidia agreed in November 2025 to invest up to $10 billion in Anthropic as part of a broader partnership under which Anthropic committed to spend $30 billion on Microsoft Azure capacity running Nvidia chips. A stake taken at IPO would tie the AI lab’s fortunes even more tightly to Nvidia’s, and it follows a pattern: Nvidia has become a serial anchor investor in companies preparing to list, and several of those companies then spend much of the money they raise on Nvidia hardware.
The numbers behind the listing explain why the offering is so large. Anthropic raised $65 billion in May at a $965 billion post-money valuation. Its annualized revenue run rate passed $65 billion by the end of July, up from roughly $9 billion at the end of 2025, driven largely by Claude Code and enterprise adoption. Projections shared with investors point to $190 to $200 billion in revenue in 2028, and the reported $2 trillion target prices the company against that curve.
The compute bill behind it
Anthropic has been buying compute on a scale no private company has attempted. Over the past 11 months it signed roughly $517 billion in compute leases, locking up 14.8 gigawatts of capacity across vendors. Amazon’s AWS is getting more than $100 billion over a decade alongside over a million Trainium2 chips, and Anthropic added multiple gigawatts of Google TPU capacity in a deal with Google and Broadcom. It has also started building an in-house chip design team to custom-tune silicon for Claude.
That spending is the point of the IPO. Training and serving frontier models requires capital that even the largest private rounds cannot cover, and a public listing converts the revenue run rate into a currency for buying infrastructure. It also gives early backers, including Amazon and Google, a path to liquidity without forcing them to sell in the private market.
| Metric | Value |
|---|---|
| Target raise | Up to $100 billion |
| Target valuation | ~$2 trillion |
| Nvidia anchor stake | Up to $10 billion under discussion |
| Revenue run rate, July 2026 | $65 billion+ |
| Compute leases signed in 11 months | ~$517 billion / 14.8 GW |
| IPO window | November 2026, before midterms |
Governance question worth watching
One detail from the filing coverage stands out: after the IPO, three outside directors will control Anthropic’s board, and the Long-Term Benefit Trust, the body that can appoint a board majority, holds no equity. Public market investors buying in at a $2 trillion valuation will effectively be funding a company whose governing structure is deliberately insulated from shareholder pressure. Whether the market prices that as a feature or a bug is one of the more interesting questions of the listing.
Public pension funds and index vehicles will end up holding this stock whether or not they have views on AI safety, and the trust structure means their voting power will be limited from day one. Comparable listings, including recent large tech debuts, have faced similar questions without lasting damage to demand. But the Anthropic arrangement goes further than most, and the S-1 will have to explain it in plain language for the first time.
There is competition for the title of largest IPO, too. OpenAI confidentially filed a draft S-1 in June 2026 and could list in 2027 at an $852 billion valuation, and Databricks sits at a $188 billion private mark. Anthropic moving first, at this size, would reset expectations for every AI listing that follows.
For Nvidia, the investment also hedges a strategic risk. Custom silicon from Google, Amazon and Anthropic itself is eating at the edges of its data center franchise. Owning a large stake in the company spending the money is a way to profit from the shift even if it accelerates.
Nothing is signed. Anchor talks at this scale have collapsed before, and the final size and structure of the IPO will only be clear when the prospectus hits the SEC. But the direction is set: the largest AI lab outside Big Tech intends to go public this year, at the largest price ever asked, with its biggest supplier standing at the front of the queue.
The revenue picture deserves a closer look, because it is what carries the valuation. A run rate above 65 billion dollars annualized puts Anthropic ahead of where Google’s cloud business stood at a comparable age, and the growth is concentrated in exactly the products enterprises pay for without negotiating: coding assistants, agent tooling and API access. Claude Code alone is credited with driving much of the jump from the end-of-2025 figure. When a company’s revenue grows more than tenfold in twelve months, a 2 trillion dollar ask is aggressive but no longer absurd on paper, and the November delay exists precisely so the prospectus can show another quarter of that curve rather than a stale snapshot from July.