Anthropic's IPO paperwork contains one number that will dominate due-diligence conversations: nearly 50% of the company's sales flow through Amazon and Google. That single concentration figure, reported by Yahoo Finance from the filing, reframes how investors should read the AI lab's otherwise strong commercial momentum.

The mechanism behind the number is straightforward. Anthropic sells access to its Claude models in large part through cloud marketplaces and partnerships — and Amazon Web Services and Google Cloud sit at the center of that distribution. Both companies are also investors in Anthropic. That dual role, distribution channel and shareholder at once, is precisely the kind of structural dependency that IPO underwriters and later public-market analysts flag first.

Why concentration at this level matters is a question of downside math. If roughly half of booked revenue depends on two counterparties, any change in commercial terms, reseller priorities, or competitive positioning at either company transmits directly into Anthropic's top line. A renegotiated revenue share, a deprioritized marketplace placement, or a rival model promoted more aggressively inside the same clouds would each move results without Anthropic losing a single end customer. The filing itself frames this as a disclosed risk, which is standard practice — but the size of the exposure, approaching half of sales, sits well above what most enterprise software companies report for their largest channel partners.

There is a second-order effect worth separating from the measured fact. The near-50% figure is what the filing discloses; what it implies about bargaining power is assertion, not data. When a vendor derives half its revenue from two buyers, those buyers hold leverage in any future commercial renegotiation. Whether Amazon or Google choose to exercise that leverage is unknowable from the filing alone, and investors should treat scenarios built on it as hypothesis rather than forecast.

The concentration also cuts differently across deal stages and partner roles. For enterprise buyers procuring Claude through AWS or Google Cloud marketplaces, the arrangement simplifies procurement — spend lands inside existing cloud commitments. For Anthropic's direct sales motion, the question becomes which segments it can reach without the intermediaries, and at what margin. The filing's risk language suggests the company itself views marketplace dependence as a variable to manage over time, not a permanent structure.

Context sharpens the stakes. Anthropic competes in a market where rivals OpenAI and Google distribute through their own channels and Microsoft's ecosystem. Distribution, not just model quality, increasingly determines which vendors land large enterprise contracts. A startup that secures hyperscaler placement gains reach; the cost of that reach is exactly the dependency the filing now quantifies.

For sales and revenue leaders watching the AI vendor landscape, the number carries a practical lesson. Channel partnerships that accelerate early revenue can harden into structural concentration, and the moment to measure that trade-off is before an IPO, a board review, or a renewal cycle forces the issue. Anthropic's filing puts a hard figure — close to half of sales — on what that trade-off looks like at scale.

How investors price that risk, and whether Anthropic diversifies its distribution before listing, will shape the reception of the offering once the IPO process moves forward.