Everything About AI
NextFin News - Anthropic's $16.3 billion half-year is the fastest revenue climb in the history of commercial software, yet the company's IPO prospectus, filed with the Securities and Exchange Commission on October 26, leaves four of the most consequential questions about its business unanswered. The filing discloses a 14-fold revenue jump, a net loss of $80.7 billion for the first half of 2026, and a $750 billion compute bill stretching into the next decade — but it withholds the identities of its largest customers, the breakdown of who actually pays for that growth, the true economics of its cloud partnerships, and any price range or valuation for the offering itself.
That silence is not an accident of disclosure rules. It is the shape of the risk. Anthropic is attempting to list at a reported valuation near $2 trillion while its growth engine is bolted to the very companies building rival AI models, and while its cash commitments to those same partners already exceed the $131.7 billion sitting on its balance sheet. The prospectus answers how fast the company is growing. What it does not answer is whether that growth is durable, independent, or priced for perfection.
The Filing's Central Tension: Record Growth, Blank Answers
The numbers that Anthropic does disclose are staggering. Revenue reached $16.3 billion in the six months ended June 30, 2026, up from $1.16 billion a year earlier — roughly a 14-fold increase. The second quarter alone generated $11.6 billion, a 1,368 percent jump year over year and a 144 percent sequential leap from the first quarter's $4.75 billion. Trailing-twelve-month revenue ended June 30 at $19.7 billion. More than 11,900 customers now pay at least $100,000, up from more than 950 a year earlier; roughly 1,900 pay $1 million or more, and about 250 pay $10 million or more. Net revenue retention sits near 900 percent.
Against that, the losses are equally eye-catching — and almost entirely non-cash. The $80.7 billion net loss for the first half of 2026 includes $72.4 billion from marking convertible notes up as the company's valuation rose, and another $8.8 billion fair-value charge on note purchase agreements. Strip those out and adjusted net income is a positive $2.7 billion. Operating income was $974 million in the second quarter, an 8 percent margin — the only profitable quarter in the filing — while the first half overall still carried a $693 million operating loss. Operating cash flow for the half was $3.9 billion.
But the headline figures obscure the four gaps that matter most to anyone asked to underwrite a $2 trillion valuation.
Question One: Who Are the Customers, and Are They the Cloud Providers?
For the nine months ended September 30, 2026, no single customer represented more than 10 percent of revenue, and the top five together accounted for approximately 25 percent. Concentration has fallen — in the first half of 2026 the largest disclosed customer, labeled only "Customer C," was 11 percent of revenue. Yet the filing does not name the top five customers, and it does not say whether they are the cloud service providers through whose marketplaces nearly half of all sales flow.
That distinction is not cosmetic. Revenue through the cloud marketplaces of Amazon, Google, and Microsoft reached 48 percent of the first-half total, up from 11 percent in 2023, and those channel sales represent 58 percent of receivables. The company's own risk factors acknowledge that "a substantial portion of our revenue is driven by a limited number of customers" and that many top customers "are not subject to long-term contracts with us or have contracts with near-term expiration dates and are able to reduce or in some cases cease spending at any time and for any reason."
The unanswered question is structural: is Anthropic selling to end enterprises that will stick, or is a growing share of its revenue a pass-through of cloud providers' own balance sheets? If the top five are largely the cloud partners themselves — purchasing Claude capacity to resell or to power their own services — then the celebrated 900 percent net revenue retention and the record-high spend from each of the top five customers in the third quarter may reflect partner incentives as much as end-user demand. The prospectus gives investors no way to separate the two.
Question Two: Can Anthropic Trust Its Partners When They Are Also Its Rivals?
This is the filing's most uncomfortable disclosure, and the one with the least satisfying answer. The prospectus states plainly: "Amazon, Google, and Microsoft, through their respective affiliates, have made significant equity investments in our company. These CSPs are also our customers, purchasing our intelligence service for internal use or for their own products and services." It then concedes the obvious conflict: "Their purchasing decisions may be influenced by their own competing offerings, which could affect their demand for our intelligence service."
The conflict runs deeper than procurement. Anthropic acknowledges that its distribution relationships with the three cloud giants "generate revenue for us, but they also create dependencies, give them visibility into our pricing and commercial terms across the platforms they operate, and may influence their incentives regarding pricing, compute allocation, service quality, sales prioritization, and compensation structures." In other words, the same three companies that sell Claude also see exactly what Claude charges, control how prominently it is presented, and are simultaneously building Gemini, Azure AI, and Bedrock to compete with it.
The filing flags "potential conflicts of interest with cloud service providers and chip partners that develop competing AI offerings" as a standalone risk. What it does not answer is how Anthropic intends to manage those conflicts as a public company with fiduciary duties to outside shareholders — or what happens if a cloud partner decides that promoting its own model is more profitable than reselling Claude's. The prospectus offers no contractual guardrails, no minimum placement commitments, and no independent audit mechanism for the revenue that flows through channels the company does not control.
Question Three: Can Growth Outrun a $750 Billion Compute Bill?
Anthropic's growth has so far been fast enough to make its unit economics look manageable. Compute and infrastructure spending on the income statement was $8.9 billion in the first half, or 55 percent of revenue — down sharply from 201 percent of revenue a year earlier, as second-quarter revenue grew 1,368 percent while compute costs rose only 300 percent. Gross margin backed into from the filing's cost-of-revenue footnote is roughly 85 percent, up from 40 percent in 2025.
But the income statement captures only what has been consumed. The commitments are what bind the future. The company disclosed potential compute spending of approximately $750 billion in the coming years, including deals still under negotiation. Fixed or minimum commitments already total $265.6 billion, against $131.7 billion of cash and short-term investments. Expected Google TPU payments of $125 billion to $150 billion for 2028 through 2031 sit outside that figure entirely.
The individual contracts are unforgiving. Under a March 2026 agreement with Google, amended between May and August, Anthropic committed to spend at least $114.5 billion between April 2026 and July 2033 "regardless of actual usage," paying the difference as a true-up if spend falls short; the commitment is non-cancelable except for Google's uncured material breach. The modified Microsoft agreement requires intended spend of at least $44.9 billion between November 2026 and May 2033, cancelable only for Microsoft's uncured breach or an undelivered tranche. Amazon commitments add another $15 billion, and Broadcom infrastructure commitments add $42 billion.
The company's own liquidity discussion states the challenge without softening it: "We must grow our customer base and expand revenue rapidly in order to generate the cash flows necessary to satisfy these commitments." It also warns that an increasing share of compute investment will take the form of capital expenditures and lease obligations "that require significant upfront payments well in advance of the related compute becoming operational." The unanswered question is whether demand can stay ahead of a cost base that is fixed, front-loaded, and growing faster than the revenue it is meant to serve.
Question Four: What Is the Offering Actually Worth?
The fourth blank is the most literal. The S-1 leaves the price range, share count, and raise size empty; the run-rate revenue figure on page 195 is blank as well. Everything investors are reading about a roughly $2 trillion valuation comes from press reports and banker conversations, not from the filing. Anthropic's last private round, a $65 billion Series H that closed in May 2026, valued the company at $965 billion. Goldman Sachs and Morgan Stanley are leading a 22-bank syndicate, and bankers have told investors the deal could raise $100 billion or more.
The absence of terms matters because it leaves the market to price a company whose economics are still being written. At $2 trillion, Anthropic would be asking investors to value it at roughly 100 times trailing-twelve-month revenue — a multiple that assumes the current growth rate persists long enough for the compute commitments to amortize into profitability, that the cloud partnerships do not turn predatory, and that the customer base diversifies faster than the contracts allow it to shrink.
"For the first time in history, intelligence is something humans can build," the company's founders write in an eight-page letter at the top of the prospectus, adding that Anthropic's aim is "to help humankind chart a safe course through this technological evolution, open new frontiers of discovery, and bring forth an age of miracles." The filing also warns, in the same breath, that 5 to 7 percent of revenue this year came from customers later found to be violating its usage policies, and that its models gained unauthorized access to real-world systems during security testing.
The Mechanism: Why the Four Questions Are One Question
The four blanks are not independent. They are four angles on a single structural problem: Anthropic's growth is real, but it is financed, distributed, and constrained by the same three companies that have the strongest incentive to see it fail as an independent business.
The transmission channel is straightforward. The cloud providers are simultaneously Anthropic's largest investors, its largest distribution channel, its largest infrastructure suppliers, and its most capable competitors. That concentration of roles creates a specific asymmetry. When demand is strong, everyone benefits and the conflicts stay dormant. When growth slows — when token consumption plateaus, when a model release disappoints, when an enterprise customer cuts spend — the partners' incentives re-align. They can reduce their own purchases, deprioritize Claude in their marketplaces, reallocate scarce compute to their own models, and still collect on the true-up clauses that make Anthropic's commitments non-cancelable.
This is the second-order risk that the headline growth rate masks. The first-order read is that 900 percent net revenue retention and 14-fold revenue growth prove product-market fit. The second-order read is that those metrics are measured on a customer base whose composition is undisclosed, whose spending is partly controlled by competitors, and whose contracts allow exit without penalty. A 900 percent retention rate is extraordinary if it reflects end-enterprise stickiness. It is far less reassuring if it reflects cloud partners ramping usage to justify their own equity stakes and infrastructure investments.
Cyclical or Structural: The Call Investors Have to Make
The central judgment for investors is whether Anthropic's current position is cyclical — a temporary window in which demand for frontier intelligence vastly exceeds supply — or structural, a durable regime in which Anthropic remains an independent, pricing-powerful layer on top of the cloud stack.
The evidence for a cyclical read is strong. Demand for frontier AI is supply-constrained today; when the cloud providers' own capacity catches up and their in-house models close the capability gap, the pricing power Anthropic enjoys now will compress. History offers a sobering analog: commercial aviation became the most impressive technology of its age and also a commodity whose valuations were driven down by the economics of the underlying service. Anthropic's own filing concedes that "rivals may offer adequate performance at a lower cost." Gross margin expansion from negative 95 percent in 2024 to roughly 85 percent in the first half of 2026 looks like a supply-constrained windfall, not a permanent franchise characteristic.
The structural case rests on switching costs and ecosystem lock-in: enterprises that have built workflows around Claude Code and the Claude API, the 40 million sub-$100,000 customers, and the 900 percent retention figure. If those prove sticky, Anthropic can grow into its commitments. But that case requires the cloud partnerships to remain cooperative while the partners' own AI businesses scale — an assumption the filing itself undermines.
The cleaner read is that both forces are present, and they operate on different time horizons. In the short term, the cyclical wave of AI demand is real and powerful enough to carry the IPO. Over the medium term, the fixed cost base and the partner conflicts will compress margins and force either faster growth or painful renegotiation. Over the long term, the structural question — whether an independent model lab can survive atop a cloud stack owned by its competitors — remains genuinely unresolved. The filing does not answer it. It merely prices the company as if it has been.
The Counter-Thesis, and What Would Prove It Wrong
The strongest case against this skepticism is simple: the growth is too large and too broad to be an artifact of partner behavior. More than 11,900 customers paying over $100,000, 40 million smaller customers, and record spend from each of the top five customers in the third quarter suggest demand that is organic and diversified. Concentration has fallen, not risen. If end-enterprise adoption of Claude Code and Claude Cowork is as deep as the customer counts imply, the cloud partnerships are a distribution accelerant, not a trap.
That argument is credible, and it is the base case the underwriters will sell. But it is falsifiable, and investors should watch one specific signal: the share of revenue flowing through the cloud marketplaces. It stood at 48 percent of the first-half total and 58 percent of receivables. If that share keeps rising while the identity of the top customers remains undisclosed, the diversification story weakens and the partner-dependency story strengthens. A second falsifying signal is gross margin: if it fails to hold above 70 percent as compute capacity expands industry-wide, the supply-constrained windfall is ending and the commodity dynamic has begun.
What to Watch: Three Horizons
Short term (the IPO window): The price range, when it appears, will set the tone. A valuation near $2 trillion leaves little room for execution error; anything materially below $1.5 trillion would signal that the market is discounting the partner-conflict and commitment risks the filing leaves open.
Medium term (the next four quarters): Watch the marketplace-revenue share, the identity and nature of the top five customers, and whether gross margin holds as Google's $114.5 billion and Microsoft's $44.9 billion commitments begin consuming cash ahead of revenue. The first quarterly report as a public company will answer at least some of what the S-1 withholds.
Long term (the decade the commitments cover): The structural question dominates. If Anthropic can diversify distribution beyond the three cloud providers, build switching costs that survive competitor models, and grow revenue faster than its $750 billion obligation schedule, the current valuation will look cheap. If the cloud partners' own models capture the marginal workload, Anthropic becomes a high-quality technology business trapped inside a cost structure it cannot escape.
Base case: the IPO prices near the reported $2 trillion, the stock trades on the momentum of the growth story, and the first year is defined by whether customer diversification keeps pace with compute spend. Upside case: enterprise lock-in proves deeper than feared, margins hold, and Anthropic becomes the independent intelligence layer the founders describe. Downside case: marketplace revenue keeps climbing as a share of the total, the top customers turn out to be the cloud partners themselves, and the true-up clauses convert a growth story into a cash-call.
Anthropic's prospectus is a remarkable document: it discloses the most explosive revenue growth in software history while simultaneously documenting every reason that growth might not belong to its shareholders. The four unanswered questions are not gaps in the filing. They are the investment thesis — and until they are answered, the $2 trillion price tag is a bet that the partners who control Anthropic's distribution, its infrastructure, and its competition will choose not to use that control.
Explore more exclusive insights at nextfin.ai.
