Anthropic’s $100B IPO: Why Nvidia Is Funding Its Customer
Anthropic wants up to $100 billion from public investors at a $2 trillion valuation. Nvidia may put in $10 billion — then keep selling Anthropic the compute.
Anthropic is reportedly chasing up to $100 billion in an IPO at roughly a $2 trillion valuation. Nvidia is considering up to $10 billion as an anchor investor.
Don’t clap just yet. When the bloke selling the picks and shovels helps fund the gold rush, he is not merely backing your dream. He is protecting his revenue.
That doesn’t make Nvidia’s reported interest in Anthropic stupid. Quite the opposite. It may be very smart. But founders, investors and anyone tempted to call this a clean vote of confidence should understand what is actually happening: AI is becoming a closed-loop capital machine, where the chip supplier funds the model company, the model company buys compute, and the public market is asked to finance the next lap.
The $100 billion question is not whether Anthropic can list
Reuters reported on September 11 that Anthropic is in talks to bring Nvidia in as an anchor investor for what could become the largest IPO in history. The reported target: as much as $100 billion raised, at around a $2 trillion valuation. Nvidia is reportedly considering an investment of up to $10 billion. Neither company confirmed the talks, and the plans can still change.
That last bit matters. Until a prospectus lands and the shares price, this is a live negotiation — not a fait accompli. But the numbers tell us where the market has gone.
Anthropic, maker of Claude, reportedly raised $65 billion in May 2026 at a $965 billion post-money valuation. Reuters also reported its annualised revenue run rate had passed $65 billion by the end of July, up from about $9 billion at the end of 2025. If the IPO lands anywhere near the numbers being discussed, the public market will be asked to value the company at more than double its May valuation only months later.
That is not a normal growth-company listing. It is an attempt to turn a private AI arms race into a public balance-sheet event.
And to be fair, Anthropic is not selling a PowerPoint and a hoodie. In February, the company announced a $30 billion Series G at a $380 billion post-money valuation. It said its run-rate revenue was then $14 billion, that more than 500 customers were spending over $1 million annually on an annualised basis, and that eight of the Fortune 10 were customers.
There is real commercial traction here. That is why this story deserves attention rather than lazy bubble-chat.
But revenue is not the same as free cash flow. A company can grow like hell and still chew through money at a rate that would make a casino blush. Frontier AI has a nasty habit: every increase in capability and customer demand can require another mountain of infrastructure spending.
Nvidia is not just investing — it is underwriting demand
Nvidia already has deep commercial and financial ties to Anthropic.
On November 18, 2025, Anthropic announced that it would purchase $30 billion of Microsoft Azure compute capacity powered by Nvidia systems, while contracting for additional capacity of up to one gigawatt. Nvidia committed to invest up to $10 billion in Anthropic, and Microsoft committed up to $5 billion.
Read that again slowly.
Anthropic needs compute. Nvidia supplies the architecture behind much of that compute. Microsoft sells the cloud capacity. Nvidia and Microsoft invest in Anthropic. Anthropic spends billions buying the compute. Everyone gets a bigger number to put in a presentation.
Again: this does not mean the demand is fake. Claude is used by real companies for real work. But it does mean the traditional line between supplier, customer, investor and strategic partner is now badly blurred.
That is the overlooked risk in this IPO. A $10 billion Nvidia anchor cheque would not just tell public investors, “Nvidia believes Anthropic will win.” It would also tell them, “One of Anthropic’s most important infrastructure partners has every reason to see Anthropic remain enormous, well-funded and thirsty for chips.”
That is strategic alignment. It is also circularity.
I’ve seen versions of this in other industries. A financier backs an operator, the operator spends with a supplier, the supplier extends terms or puts capital back into the ecosystem, and everyone tells themselves it is a flywheel. Sometimes it is. Sometimes it is just a very expensive way to delay the moment somebody asks whether the underlying unit economics work.
The IPO market is being asked to fund the next phase of the AI war
The easy take is that Anthropic’s float would be a referendum on Claude versus OpenAI. Too narrow.
The bigger issue is whether public-market investors are prepared to bankroll frontier-model companies with capital requirements closer to national infrastructure projects than normal software businesses.
Reuters reported that Anthropic has committed more than $100 billion over a decade to Amazon Web Services, plans to use more than 1 million Amazon Trainium2 chips, and has agreements with Google and Broadcom for multiple gigawatts of TPU capacity. It is also building an in-house custom-chip team, according to the report.
This is what people miss when they compare AI companies to the great software firms of the last 20 years. Software businesses traditionally became wonderful because each extra customer cost comparatively little to serve. Frontier AI may become wonderful too, but it is being built on an industrial-scale appetite for chips, power, cooling, data centres and debt-worthy infrastructure.
That changes the game.
If Anthropic floats at roughly $2 trillion, it will give the market a liquid benchmark for the entire frontier-AI race. It will also put relentless pressure on quarterly execution. Private companies can tell a long story. Public companies must answer every three months whether revenue growth, gross margin, compute spend and customer concentration are behaving themselves.
And that could be useful. Frankly, a bit of sunlight would do this sector good.
The contrarian view: Nvidia may be the safer way to own Anthropic’s upside
Here’s the part the AI evangelists won’t love.
If you believe Anthropic will keep growing at an extraordinary rate, buying Anthropic at a $2 trillion valuation may not be the cleanest way to express that belief. Nvidia potentially benefits whether Anthropic wins, OpenAI wins, or several major labs stay in the fight long enough to keep spending heavily on infrastructure.
Anthropic has model risk, product risk, safety risk, pricing risk and the burden of having to turn staggering revenue into durable profits. Nvidia has competition and cyclicality risks of its own, obviously, but it sits closer to the toll booth.
That does not mean Nvidia is automatically cheap, or that Anthropic cannot become a generational public company. It means the phrase “Nvidia is investing” should not end your thinking. It should start it.
The irony is that Nvidia’s potential anchor investment could make Anthropic’s IPO easier to sell precisely because it validates the supply chain that Anthropic must keep paying. Public investors need to decide whether that is a moat or a dependency.
It can be both.
This is also a warning for ordinary founders
Most founders will look at this story and think the lesson is: raise more money.
Wrong lesson.
Anthropic is not your benchmark unless you are operating a frontier-model business with global enterprise demand, deep technical talent, strategic cloud relationships and a genuine need for massive compute. For almost everybody else, a giant round is not proof of strength. It is a commitment to a much harder future.
Big money raises the cost of being wrong. It invites more scrutiny, more hiring mistakes, more nonsense projects and a valuation that can turn a good business into a disappointing outcome.
The founders who win the next decade will not be the ones pretending they are mini-Anthropics. They will be the ones using AI to build leaner, sharper businesses that sell a useful outcome before spending like lunatics.
What this means for you
If you are a founder: separate customer demand from investor excitement. Ask whether customers renew, expand and pay enough to cover the real cost of serving them. If your AI product needs expensive inference, measure gross margin by customer and use case — not across a spreadsheet designed to impress a seed investor.
If you are an operator: do not buy AI because it looks clever in a board deck. Pick one costly, repetitive workflow. Set a baseline for time, error rate or revenue. Run a contained trial. Keep the tool only if it produces a measurable commercial result. The winners will compound practical gains, not collect AI subscriptions like Pokémon cards.
If you are an investor or saver: distinguish the business from the capital structure. Ask who funds the company, who sells it infrastructure, what the company is contractually committed to buy, and whether a strategic investor is also a major supplier. That is not cynicism. That is due diligence.
Anthropic’s reported IPO is a massive vote for AI’s future. It is also a reminder that capital can make a business look inevitable right up until the bill arrives.
Back the companies with real demand. Just don’t confuse a $10 billion cheque from the chipmaker with proof that the economics are bulletproof.