Anthropic is in talks to bring Nvidia in as an anchor investor in what would be the largest initial public offering ever attempted, Reuters reported on Friday evening, citing two people familiar with the matter. The company is seeking to raise as much as $100 billion at a valuation near $2 trillion. Nvidia is weighing up to $10 billion of that.
Both sides said little. Anthropic declined to comment and Nvidia did not respond to Reuters’ request. The sources cautioned that the plans are still under discussion and could change.
Strip out the superlatives and one structural fact remains: the investor being courted to validate the price is also the supplier being paid. Nvidia sells Anthropic the compute its models run on. Under this plan it would also help set what the market pays for Anthropic’s equity.
Article Brief
Key Takeaways
5 Points30s Read
- The dealAnthropic is in talks to bring Nvidia in as an anchor investor, with the chipmaker weighing up to $10 billion, Reuters reported citing two people familiar with the matter.
- The sizeAnthropic is seeking as much as $100 billion at a valuation near $2 trillion, which would be the largest IPO ever attempted.
- The scale problemA $100 billion raise equals roughly 73% of the record $137 billion raised by all US IPOs, excluding SPACs, in the first eight months of this year.
- The circularityNvidia, Amazon and Google are each both backers of Anthropic and suppliers it has committed to pay for compute.
- What is not settledBoth sources were anonymous, Anthropic declined to comment, Nvidia did not respond, and the sources said the plans could change.
Start with the size of the ask
A $100 billion raise is not merely large for a single company. It is large for the entire market.
US initial public offerings, excluding blank-cheque vehicles, raised a record $137 billion in the first eight months of this year, according to Dealogic figures cited by Reuters. Anthropic alone would be asking for roughly 73% of that — one listing seeking nearly as much as every US flotation combined across a record-setting year.
That is the real test embedded in this deal, and it is a test of the buy side rather than of Anthropic. Someone has to write those cheques. Anchor investors exist precisely to answer that question early: they commit to a defined slice before the offering is marketed widely, which gives everyone behind them a reason to believe the book will fill. Arm lined up Nvidia and Amazon that way. Saudi Arabia’s Public Investment Fund did the same for SpaceX.
The supplier is also the investor
This is the part that deserves more scrutiny than the headline number.
The template was set in November 2025. Microsoft, Nvidia and Anthropic announced a set of strategic partnerships under which Nvidia would invest up to $10 billion and Microsoft up to $5 billion. In the same arrangement, Anthropic committed to purchase $30 billion of Azure compute capacity — running on Nvidia chips — and to contract for up to a further gigawatt of capacity.
Set those numbers beside each other. Up to $15 billion went in from the two companies. Thirty billion was committed back out to one of them, for hardware built by the other. A second $10 billion from Nvidia at the IPO would extend the pattern into the public market.
Nor are they alone in occupying both seats. Amazon and Google are among Anthropic’s largest backers and among its largest compute suppliers. In April the company said it would commit more than $100 billion over a decade to Amazon’s AWS while using over a million of Amazon’s Trainium2 chips. It has separately agreed with Google and Broadcom to add multiple gigawatts of TPU capacity.
So the register of people funding Anthropic and the register of people Anthropic pays are close to the same list.
None of this is improper, and none of it is hidden. Strategic investment by suppliers is ordinary in capital-intensive industries, and a chipmaker taking a stake in a major customer is a recognisable move. But it does change what an anchor commitment signals. An anchor investor is meant to be an independent vote of confidence in a price. A supplier’s stake in a customer that is contractually committed to buying its hardware is a different instrument, and it carries a different amount of information about whether $2 trillion is the right number.
For Nvidia’s own shareholders, $10 billion is not a balance-sheet event — the company is measured in trillions and has taken strategic stakes before. The thing worth tracking on Nvidia’s stock is the disclosure question that tends to follow arrangements like this. When a supplier funds a customer that is committed to buying its hardware, analysts eventually ask how much of reported demand is genuinely independent. Vendor financing has attracted that question in technology for decades, and it usually arrives not when the deals are announced but when growth slows.
What has to be true for $2 trillion
The growth underneath this is not in dispute, and it is extraordinary. Anthropic’s annualised revenue run rate passed $65 billion by the end of July, up from roughly $9 billion at the close of 2025 — more than seven times higher in seven months, on the company’s own figures.
Apply the proposed valuation to that and Anthropic would price at about 31 times its current annualised run rate. For a business compounding at this speed, that multiple is defensible on its face.
The forward case is where the assumptions concentrate. Reuters has previously reported that the valuation depends in part on company projections of roughly $190 billion to $200 billion of revenue in 2028. Against the midpoint, $2 trillion is about 10 times forward revenue, which sounds modest — but only if you accept a projection that requires revenue to roughly triple again within two years, produced by the company seeking the money.
The May round is the other reference point. Anthropic raised $65 billion then at a post-money valuation of $965 billion. Reaching $2 trillion would mean the price roughly doubled in about four months, with no public market having yet expressed a view.
Line the marks up and the slope is the story. Anthropic was valued at $183 billion a year ago. The November partnerships carried it to roughly $350 billion. May’s round took it to $965 billion. The proposed listing would put it near $2 trillion — close to eleven times higher in about twelve months, with each step set privately by a small group of investors, several of whom sell the company its compute.
Public markets are where that sequence meets people who have no commercial relationship with Anthropic at all. That is the point of a listing, and it is why the identity of the anchor investors matters more here than in an ordinary offering.
The clock is doing some of the work
The listing is expected to complete before the US midterm elections in November. That detail is easy to skim past and it is worth holding onto, because it compresses everything above into a narrow window.
A deal of this size normally takes its time finding the price. A deadline shortens the period in which the market can push back, and it raises the value of having committed buyers lined up in advance. It also means the offering lands into whatever conditions November brings rather than conditions the company gets to choose.
TECHi has been tracking the run-up to this for months, from the AI IPO supercycle that SpaceX opened in June to the structure of an eventual Anthropic listing. The direction has been consistent. The speed is what has changed.
Why leverage is the thing to watch, not the multiple
There is a recent and specific reason to look past the headline valuation at how the position is financed.
In July, a fund built entirely around the AI trade lost roughly two-thirds of its value in a single month — not because its thesis was wrong, but because it held the right idea with too much borrowed money. TECHi went through its regulatory filings and found a book that flipped from 61.9% bearish to 99.6% long in one quarter before the drawdown. The stocks it was forced to sell recovered most of the damage without it.
The lesson transfers. The risk in an offering this size is not principally that Anthropic’s business disappoints. It is that a $100 billion raise priced near a record has to be absorbed by buyers, some of whom will be levered, into a market that has spent the year setting records. Concentration plus leverage is what turns an ordinary drawdown into a forced one.
What is actually known, and what is not
Known: Anthropic is in talks, Nvidia is considering up to $10 billion, the target is as much as $100 billion at roughly $2 trillion, and the listing is expected before the midterms. All of it comes from two people familiar with the discussions, reported by Reuters, and Nvidia’s potential role had not been reported anywhere before Friday.
Not known: whether any of it holds. The sources explicitly said the plans could change. Neither company has confirmed a figure, a date or a structure. No prospectus exists publicly. The 2028 revenue projection underpinning the valuation is the company’s own and has not been tested by a public filing.
Anthropic’s compute costs are also moving beneath all of this. It is diversifying away from sole reliance on Nvidia GPUs — Trainium2 at Amazon, TPUs with Google and Broadcom — and has formed an in-house team to design custom chips tailored to Claude. A company actively reducing its dependence on Nvidia silicon, while inviting Nvidia to anchor its IPO, is holding two positions at once. Both can be rational. They are still worth naming.
The number to watch is not the valuation. It is how much of the $100 billion is spoken for before the deal is marketed, and by whom. If most of the book is filled by companies that also sell Anthropic compute, the price will have been set largely by the supply chain rather than by the market.