The number that caught my attention in Anthropic’s IPO prospectus had very little to do with the price of its shares.
It was the amount the company could pay SpaceX.
Up to $120.5 billion.
At full delivery of the contracted capacity, the arrangements envisage payments of approximately $3.1 billion a month. Annualise that monthly figure and you get $37.2 billion, a potential spending rate, rather than guaranteed annual revenue.
That changes how I want to approach this IPO.
Much of the discussion will understandably focus on Claude: how capable it is becoming, how quickly businesses are adopting it, and how much of the economy Anthropic could eventually serve.
Those are important questions. I am optimistic about the opportunity.
But there is another investment story inside the prospectus.
Anthropic needs an enormous industrial system to turn that opportunity into a business. Someone has to supply the computing capacity. Someone has to connect the machines, deliver the electricity and finance the infrastructure.
Those companies can earn money while Anthropic is still working out how much of the value created by Claude it can retain.
That is where I want to focus this edition: the businesses getting paid as Anthropic grows, and the prices at which their shares could make sense.
First, understand why this IPO will be so compelling
It would be easy to build a cautious argument by dismissing the technology.
I think that would miss the point.
The exciting possibility is that AI becomes a larger part of the work businesses already pay people and software to perform. Coding is an obvious example, but the opportunity extends into research, financial analysis, customer operations and complex administrative processes.
As those systems improve, the potential budget expands beyond a subscription purchased by an individual employee. AI can become part of how an organisation delivers its services.
That is a substantial opportunity.
And Anthropic’s growth gives investors a powerful reason to take it seriously. Revenue reached approximately $33.2 billion in the first nine months of 2026, compared with roughly $2.5 billion a year earlier.
Faced with that trajectory, investors will be tempted to ask a familiar question:
How expensive can the shares really be if the business keeps growing this quickly?
My concern starts with the assumptions inside that question.
How long does exceptional growth continue? What must Anthropic spend to support it? How much pricing power survives competition? And how much of the eventual value accrues to each share?
An IPO valuation can require very optimistic answers to all four.
A company can deliver extraordinary progress while its shareholders earn a disappointing return. The starting price determines how easily those two outcomes can coexist.
The profitability question deserves a closer look
The prospectus reports a −4% operating margin for the nine-month period and a +12% adjusted operating margin. The adjustment excludes stock-based compensation and certain other expenses.
That distinction matters because shareholders ultimately own the economics of the business, including the cost of retaining the people who build it.
Equity compensation can be sensible. It can help a company recruit exceptional employees without paying the entire compensation bill in cash.
But the economic cost still has to land somewhere.
Existing shareholders can own a smaller percentage of the business as new shares are issued. Alternatively, the company can spend cash buying shares back to offset issuance. Neither outcome disappears because an expense is removed from an adjusted earnings measure.
There is also a separate accounting event to understand. Anthropic describes a potential $8.2 billion cumulative RSU expense triggered by the offering, based on its September figures.
I would not interpret that as an equivalent cash bill or an automatic forecast of selling pressure. Accounting recognition, dilution and employees becoming able to sell are different events.
But I would want to model all three before buying.
The risk is a gap between what investors initially celebrate, rapid growth and improving adjusted profitability, and what they subsequently have to absorb through reported results, a changing share count and additional stock becoming available for sale.
That gap could create a much better entry point later.
It could also prove manageable if operating performance is strong enough. The question is how much protection the purchase price gives us if the outcome is less favourable.
Now look at the other side of Anthropic’s spending
For Anthropic, compute is a resource it must purchase and turn into something customers value.
For its suppliers, that purchase can be revenue.
The same payment therefore sits on opposite sides of two business models. One company must earn an attractive return after making it. The other must earn an attractive return from providing the service.
Following that relationship can reveal opportunities that a discussion focused entirely on Claude’s future market share would overlook.
It also changes the questions we ask.
Instead of stopping at “How large could Anthropic become?”, we can investigate:
Who receives its spending? How much of that spending becomes profit? What additional investment does the supplier need? And can the supplier serve other customers if Anthropic’s requirements change?
The opportunity sits in the answers to those questions, rather than in the size of the announcement alone.
SpaceX is the example I keep coming back to
Anthropic announced that its initial SpaceX agreement provided access to all the compute capacity at Colossus 1: more than 300 megawatts, incorporating over 220,000 Nvidia GPUs.
Think about the commercial relationship that creates.
Claude competes with Grok. Yet the group behind Grok can also generate revenue supplying infrastructure to Claude.
For investors, that opens a second way to participate in Anthropic’s success.
SpaceX does not need every enterprise to choose Grok for the Anthropic relationship to have value. It needs to provide computing capacity on terms that produce an attractive return.
And if Anthropic raises fresh capital through its IPO, my inference is that a stronger balance sheet could improve its ability to support those purchases.
Only the net proceeds from newly issued shares go to Anthropic; shares sold by existing holders do not fund its operations. The eventual offering structure therefore matters.
But the mechanism is straightforward: fresh capital can strengthen a customer’s ability to pay its suppliers.
The IPO could benefit SpaceX even if buying Anthropic at the IPO price turns out to be a poor investment.
That is the central distinction I want readers to understand.
There is an important limit, however. The SpaceX arrangements allow cancellation after their initial firm periods, subject to notice. The headline amount should therefore never be treated as secured backlog.
For SpaceX, we still need to examine the cost of serving Anthropic, the capital tied up in the equipment and the value of that capacity if the customer changes its plans.
The next step is to turn the relationship into a cash-flow model.
There are other places to look—and some could offer better entries
SpaceX is the clearest example for explaining the thesis. Whether it offers the best investment depends on valuation.
I want to apply the same process across the infrastructure chain.
Optical connectivity is one area. Electrical distribution is another. Memory, specialised processors and infrastructure financing also deserve attention.
The challenge is to distinguish a genuine earnings opportunity from a share price that already assumes years of exceptional demand.
A supplier can win substantial business and still disappoint investors if the market has paid too much for it.
Diversification also needs closer examination. Five companies exposed to the same customers, financing conditions and construction cycle may provide much less protection than five different tickers suggest.
I would therefore test each candidate against a less enthusiastic environment: slower orders, pressure on prices, delayed projects and lower valuation multiples.
Which companies could keep generating cash? Which would need to keep borrowing? Which have customers and activities outside the most aggressive AI spending programmes?
Those questions are central to how I want to position.
My approach to the IPO
I am bullish on the opportunity Anthropic is pursuing.
My starting position is that I would rather investigate the beneficiaries of its spending than commit to its shares at an IPO valuation that may leave little room for disappointment.
The eventual price could change that judgement. So could new evidence about profitability and capital requirements.
But I feel no obligation to buy on the first day.
Consider a simplified investment where revenue rises 50%, the valuation multiple falls 40%, and the share count increases 5%. Holding other factors constant, the value per share falls approximately 14%.
The business grew substantially. The shareholder still lost money.
That is why I want our research to begin with the return required from today’s price, rather than the excitement surrounding the listing.
My preferred opportunity is a company that can benefit from AI’s expansion and still offer an acceptable return if the market becomes less enthusiastic about it.
In the Premium research: mapping the winners and deciding what to pay
We will start with SpaceX and work through the wider chain of potential beneficiaries.
For each company, I want to establish the commercial link, the potential contribution to earnings, the capital required and the valuation we are being asked to pay.
The research will cover:
The beneficiary map: direct counterparties, indirect suppliers and the evidence connecting each to the spending.
The SpaceX economics: what the Anthropic relationship could contribute after operating costs, investment and financing.
Three scenarios: continued expansion, strong adoption with pricing pressure, and a slowdown in infrastructure spending.
The entry-price framework: what each business would need to deliver to justify its valuation, and where a margin of safety could emerge.
The Anthropic watchlist: the valuation, dilution and operating milestones that could make its own shares attractive later.
The objective is to build a selective way to participate in AI’s emergence, with a clear understanding of what could go wrong.