In the long arc of technological ambition, Anthropic's march toward a $2.9 trillion IPO stands as one of the most audacious bets capital markets have ever been asked to absorb — a company losing tens of billions annually, sustained by the very rivals it competes against, asking the world to believe the gap between spending and earning will one day close. The prospectus reveals not just a business, but an entire industry's wager: that the cost of building artificial intelligence at frontier scale will eventually be justified by what humanity is willing to pay for it. Whether that faith holds de
Anthropic's $2.9T IPO valuation masks billion-dollar losses and precarious dependencies
The entire sector is vulnerable to any adverse event
Mark
So Anthropic is going public at nearly $3 trillion. How does a company losing $42 billion a year get valued that high?
Mimi
Because investors believe the revenue growth will eventually outpace the spending. Revenues jumped from $4.6 billion to over $16 billion in just six months. If that trajectory holds, the losses become a temporary problem.
Luke
But "if" is the operative word. We don't actually know if that growth continues, or if margins will ever materialize. The company is spending $518 billion on future infrastructure—that's a bet, not a guarantee.
Mark
What about the customer concentration issue? That seems like a real problem.
Mimi
It is. A quarter of revenue comes from two customers. And nearly half of all sales flow through Amazon and Google, who also compete with Anthropic and supply its computing power. That's a conflict of interest built into the business model.
Luke
More than a conflict—it's a structural vulnerability. If Google or Amazon decided to compete more aggressively, or if they changed their terms, Anthropic's cash flow could be severely disrupted. The company even admits this in the prospectus.
Mark
So why would investors buy into this?
Mimi
Because the AI market is believed to be enormous—estimates range from $5 trillion to $30 trillion. If Anthropic captures even a small slice, the current losses look like an investment in the future.
Luke
But those estimates are guesses. No one knows what the actual market will be, or what customers will pay. And the bond market is already tightening—borrowing costs are rising. If capital dries up before these companies reach profitability, the whole structure collapses.
Mark
Is that likely?
Luke
Unknown. But it's the risk no one's really pricing in.
El Pulso
- Anthropic lost $42 billion last year and more than $8 billion in a single quarter, yet its valuation has ballooned tenfold in eighteen months — the arithmetic of belief outrunning the arithmetic of profit.
- Nearly a quarter of revenues flow from just two unnamed customers, and 47% of all sales pass through Amazon and Google, who are simultaneously investors, suppliers, and direct rivals — a concentration the company itself warns could create conflicts of interest and choke its cash flow.
- The entire AI frontier is structurally dependent on perpetual capital infusion, with hyperscalers projected to spend $800 billion on infrastructure this year and $1.1 trillion the next, binding competitors into a web of mutual vulnerability.
- Bond markets are flashing caution: yields sit near 20-year highs, credit spreads on hyperscale debt are widening, and the cost of insuring AI data centre debt is climbing — making the borrowing that sustains this model progressively more expensive.
- Anthropic claims it will post back-to-back operating profits this quarter, but the accumulated losses and $518 billion in future infrastructure commitments mean the IPO is less a graduation than a gamble on how long the window stays open.
In the long arc of technological ambition, Anthropic's march toward a $2.9 trillion IPO stands as one of the most audacious bets capital markets have ever been asked to absorb — a company losing tens of billions annually, sustained by the very rivals it competes against, asking the world to believe the gap between spending and earning will one day close. The prospectus reveals not just a business, but an entire industry's wager: that the cost of building artificial intelligence at frontier scale will eventually be justified by what humanity is willing to pay for it. Whether that faith holds depends less on Anthropic alone than on the patience of bond markets, the loyalty of a narrow customer base, and the continued willingness of Amazon and Google to prop up a competitor they also happen to be.
Anthropic is preparing to go public at a valuation of roughly $2.9 trillion — one of the largest IPOs in history — seeking to raise up to $100 billion. But the prospectus leaked to Reuters tells a story of a company burning through extraordinary sums while remaining tethered to a dangerously narrow set of relationships for its survival.
The financial picture is stark in its contrasts. Anthropic lost $42 billion last year on a net basis, including more than $8 billion at the operating level in a single quarter — nearly triple its full-year operating loss from 2024. At the same time, revenues are surging, reaching $16 billion in the first half of 2026 alone. The company spent $7.33 billion on compute and infrastructure last year and has committed to $518 billion more in cloud and data centre spending ahead. It says it will achieve operating profitability this quarter for the second consecutive time, but the accumulated losses remain the defining feature of its finances.
What makes the position genuinely precarious is the concentration of dependencies. Nearly a quarter of last year's revenues came from just two customers. More striking still, 47 percent of total sales flow through Amazon and Google — companies that are also major investors in Anthropic, suppliers of the computing power it needs, and direct competitors in the AI market. Anthropic's own prospectus acknowledges this creates complex dynamics that could give rise to conflicts of interest and threaten its access to compute, while also affecting cash flow in ways that are difficult to predict or control.
This is not a problem unique to Anthropic. Across the AI sector, the capital required to operate at the frontier has forced rivals into cooperation, creating layers of mutual vulnerability. Frontier labs cannot survive without constant new investment, and even the hyperscalers — Amazon, Alphabet, Meta, Microsoft, Oracle — are straining under projected capital expenditures of $800 billion this year and $1.1 trillion the next.
The sector's central question remains unanswered: will the gap between spending and revenues ever close, and will capital markets stay open long enough to find out? Bond yields are near 20-year highs, credit spreads on hyperscale debt are widening, and the cost of insuring AI data centre debt is rising. For companies already reliant on debt financing, the ground is shifting. Anthropic moves toward its IPO, but the foundation beneath the entire industry remains fragile.
Anthropic is preparing to go public at a valuation around $2.9 trillion, making it one of the largest initial public offerings in history. The company plans to raise up to $100 billion. But the prospectus details leaked to Reuters paint a picture of a business burning through staggering sums of money while remaining dependent on a narrow base of customers and rivals for its survival.
The numbers are arresting. Last year, Anthropic lost $42 billion on a net basis. In the September quarter alone, it lost more than $8 billion at the operating level—nearly triple the $2.98 billion operating loss it posted in 2024. Yet revenues are climbing fast: the company brought in $4.6 billion last year and more than $16 billion in the first half of 2026. The gap between what it spends and what it earns remains enormous. The company spent $7.33 billion on compute and infrastructure last year and has committed to spending another $518 billion on cloud services and data centres in the years ahead.
Anthropologic's valuation has inflated with remarkable speed. In a funding round a year ago, it was valued at $183 billion. By February of this year, that number had jumped to $380 billion. In May, it hit $965 billion. Now, on the eve of its IPO, the company is seeking a valuation ten times what it was worth just eighteen months ago. The company says it will achieve profitability at an operating level in the current quarter—its second consecutive quarter in the black—but the losses accumulated over years of heavy spending remain the dominant feature of its financial picture.
What makes the situation precarious is how narrow Anthropic's revenue base actually is. Nearly a quarter of its revenues last year came from just two customers. More troubling still, 47 percent of its total sales flow through Amazon and Google, which handle distribution and collect payments on the company's behalf. Both of these companies are also major investors in Anthropic, suppliers of the computing power it needs to operate, and direct competitors in the AI market. The company acknowledged in its prospectus that this concentration of relationships "creates complex dynamics that could give rise to conflicts of interest and adversely affect our access to compute." It could also, given that rivals are collecting nearly half of customer payments, significantly impact Anthropic's cash flow.
This tangled web of dependencies is not unique to Anthropic. Across the AI sector, the sheer scale of capital required has forced competitors to cooperate, creating layers of mutual vulnerability. Frontier labs like Anthropic and OpenAI cannot survive without constant infusions of new capital. Without the vast legacy revenues of the hyperscalers—Amazon, Alphabet, Meta, Microsoft, and Oracle—companies like Anthropic would collapse under the weight of their spending. Even the hyperscalers themselves are showing strain. Goldman Sachs estimates they will spend about $800 billion on capital expenditures this year, with another $1.1 trillion likely next year.
The entire sector now faces a critical vulnerability: its ability to keep raising money depends on convincing investors that the gap between current spending and revenues will eventually close. But the questions are hard to answer. What will businesses and consumers ultimately pay for AI? Will cheaper alternatives from "open weight" providers—companies offering adequate AI at a fraction of the cost—erode the margins that Anthropic and others need to justify their investments? Will capital markets remain open long enough for these companies to approach financial self-sufficiency?
The bond market is already sending warning signals. Yields across the curve are at or near 20-year highs and still rising. The spreads over Treasury bonds for hyperscale debt have widened, and the cost of credit default swaps for AI data centres has climbed. For companies already dependent on debt financing, borrowing is becoming more expensive. The sector remains vulnerable to any adverse event—a macro shock, a company-specific crisis, or a sudden tightening of capital markets—precisely because of the incestuous financial relationships that bind these competitors together. For now, Anthropic moves toward its IPO. But the foundation beneath it remains fragile.
Citas Notables
Creates complex dynamics that could give rise to conflicts of interest and adversely affect our access to compute— Anthropic prospectus, on its reliance on Amazon and Google for distribution and infrastructure
Quieres la nota completa?
Lee el original en The Age ↗