Anthropic’s $100B IPO: Can AI Make Real Money?
A reported $100 billion IPO would force Anthropic to prove a $2 trillion AI business can produce more than hype, compute bills and spectacular revenue.
A reported $100 billion IPO is not a victory lap. It is a $2 trillion question Anthropic would have to answer in public: can AI produce real returns, or just spectacular revenue and even more spectacular compute bills?
Anthropic is reportedly weighing an IPO that could raise more than $100 billion and value the Claude maker at about $2 trillion. That is not a funding round. That is a public referendum on whether AI is a business or merely the most expensive group project in history.
I’m not dismissing Anthropic. Far from it. It has built a serious product, attracted serious customers and become one of the few companies that can credibly stand beside OpenAI in frontier AI.
But when the cheque gets that large, the question changes.
It is no longer: Can Claude write code, analyse documents and save companies time?
It becomes: Can Anthropic turn all that intelligence into returns worthy of a $2 trillion valuation without needing an ever-larger pile of investor cash and borrowed computing power?
That is a much nastier question. Public markets have a habit of asking it repeatedly.
The $100 billion number is the story
Bloomberg reported that Anthropic expects to match or exceed the size of SpaceX’s record IPO. Other reporting has put the potential capital raise above $100 billion, with a possible valuation around $2 trillion.
That is still reported, not settled. The final size, timing and valuation are not the point yet. The scale is.
Anthropic is not preparing to list because it wants a bit of brand recognition or a nice liquidity event for early staff. It needs access to public-market capital because frontier AI is now an infrastructure business as much as a software business.
That means chips. Data centres. Power contracts. Networking gear. Research talent. Long-term compute commitments. More chips.
This is the inconvenient truth many people miss when they look at an AI chatbot and think, “That’s just software.” The interface is software. The engine underneath is an industrial project with a software margin attached.
Anthropic’s reported deals underline the point. Bloomberg has reported major computing arrangements, including a $9 billion deal with Riot Platforms and a separate $10 billion computing deal with a cloud startup. These are not the costs of building another project-management app. They are the costs of competing at the frontier against OpenAI, Google, Meta and whoever else decides to spend a small country’s GDP trying to own the next platform.
A $100 billion IPO would give Anthropic something hugely valuable: a war chest that is public, liquid and renewable. It could issue shares, raise debt more easily, pay employees with stock and keep signing enormous infrastructure contracts without asking private investors to swallow another eye-watering valuation step-up.
That is the bull case.
The bear case is simpler: if you need $100 billion just to keep your seat at the table, the table may not be as profitable as everyone hopes.
Anthropic has real traction — but valuation is still a brutal master
Let’s give credit where it is due. Anthropic is not some bloke with a slide deck, a hoodie and a claim that he has reinvented search.
Claude has become a genuine enterprise product, especially in coding, analysis and knowledge work. Anthropic has also positioned itself as the more safety-conscious rival to OpenAI, which matters to corporations buying AI for regulated and high-consequence work.
The company has reportedly been growing revenue hard. Bloomberg reported in August that Anthropic’s annualised revenue had passed $65 billion. If that figure holds, it puts the company in rare territory: enormous revenue growth with a product that has clearly escaped the laboratory.
But revenue is not profit. And annualised revenue is not free cash flow. Anyone who confuses those two deserves the lecture they get from the market.
At a $2 trillion valuation, Anthropic would be asking investors to believe several things at once:
1. That enterprise AI spending will remain massive rather than turn into a procurement clean-up. 2. That Anthropic can protect pricing even as models become cheaper and more interchangeable. 3. That its computing bill does not grow faster than the gross profit it creates. 4. That Claude becomes embedded enough in business workflows that customers cannot casually swap it for OpenAI, Gemini, a Meta model or an open-weight alternative. 5. That frontier-model providers capture most of the economic value, rather than the companies using the models or the chip and cloud suppliers underneath them.
That last point is the one I’d be most careful with.
In every gold rush, the obvious winners are not always the people digging. Sometimes it is the bloke selling shovels. In AI, Nvidia has already demonstrated that beautifully. Cloud providers, power developers, data-centre operators and network companies are also taking their cut before many AI application businesses have worked out what their margins should be.
Anthropic may absolutely become one of the great businesses of this era. But a great product and a great investment are not synonyms. Price matters. Always has.
The public market is about to become the adult in the room
Private markets can tolerate ambiguity for a long time. They are designed for it.
A private investor can say, “Yes, the company is burning cash, but look at the growth.” They can mark a position up once a quarter, ignore daily volatility and hope the next funding round produces a higher valuation.
Public investors are less romantic. They want disclosure. They want comparable numbers. They want to know what the company earns from customers, what it spends on compute, what it owes under long-term contracts and whether a model improvement actually produces more profit.
That is why an Anthropic prospectus will matter far more than the headline valuation.
I want to see customer concentration. If a small number of giant firms drive a hefty chunk of revenue, that is risk.
I want to see gross margin trends. If revenue is booming but cost of revenue climbs just as quickly, the business is running hard without getting much fitter.
I want to see commitments for chips, cloud capacity and power. Some of those deals will be smart strategic locks on scarce infrastructure. Others may look ugly if demand softens or model efficiency improves faster than expected.
And I want to see stock-based compensation. AI firms are in a savage war for researchers and engineers. Paying talent well is sensible. Pretending dilution does not count is not.
The best outcome for Anthropic is not merely raising a record amount of money. It is setting a standard of transparency that proves frontier AI can be judged like a business rather than worshipped like a movement.
The overlooked angle: this IPO could punish smaller software companies
Here is the second-order effect most founders and investors should not ignore.
A giant Anthropic IPO could drain attention, capital and index capacity from the rest of the market. Bloomberg’s analysis of the coming mega-listing cycle made this point plainly: huge IPOs can create natural buyers through index inclusion, but they can also force investors to sell existing holdings to make room.
That matters because the market does not have infinite appetite. If public investors are preparing to absorb a $100 billion AI offering, they may be less generous toward a $300 million SaaS listing, a middling cybersecurity float or a software company whose growth story has already been dented by AI.
The irony is delicious. AI is supposed to make software companies more productive. But the AI capital cycle may make many software companies less investable.
For founders, this is not an abstract capital-markets chat. If you run a software business with weak retention, generic features or a pricing model that assumes customers will keep paying for work AI now does cheaply, the market is going to get less forgiving.
You do not want to be pitching “AI-enabled workflows” six months after Anthropic has taught the market what actual AI scale looks like.
The contrarian view: a huge IPO may be good discipline
Everyone loves to sneer at a big valuation. Fair enough — plenty deserve it.
But I can see a constructive case here. Taking Anthropic public may impose discipline that private capital has softened. It would force clearer reporting, tougher capital allocation and a more honest conversation about which AI use cases pay today versus which ones make a good conference demo.
That is healthy.
It may also force a split in the AI market. A small number of frontier labs will become capital-intensive infrastructure companies. Everyone else will need to stop pretending they are building the next model lab and instead build profitable applications on top of whichever models win.
That would be good for operators. Less cosplay. More customers.
What this means for you
If you are an investor, do not buy the phrase “AI exposure” and call it analysis. Ask where the profit pool sits. Is it in chips, cloud capacity, models, applications, services or the customer’s own productivity gain? Those are different bets with wildly different economics.
If you are a founder, stop treating AI as a feature checklist. Your job is to identify one expensive, repetitive decision or workflow and make it materially faster, cheaper or more accurate. If you cannot show the economic improvement in dollars, hours or headcount avoided, you do not have a business case. You have a demo.
If you run an established company, negotiate AI vendors like a grown-up. Demand usage data, exit rights, pricing protections and a plan for what happens when the model changes. Do not hand a supplier your core workflow, customer data and margin structure because their chatbot had a nice landing page.
And if Anthropic does pursue a $100 billion IPO, read the prospectus. Properly. Skip the victory-lap headlines and look for the ugly bits: compute commitments, concentration, gross margins, cash burn and dilution.
That is where the truth will be hiding.
The biggest AI IPO in history may create fortunes. It may also remind a market drunk on intelligence that businesses still run on the old stuff: customers, margins, cash and discipline.