Anthropic has crossed into operating profit for the first time, posting nearly a billion dollars on that metric since the first quarter of 2025, but the company has warned investors the green ink is temporary. Seaport Research Partners senior analyst Jay Goldberg told Yahoo! Finance on October 30, 2026 that Anthropic expects to slip back into losses within a quarter or two as it changes how it acquires computing power.
The shift is structural, not cyclical. Anthropic is moving away from renting compute by the hour from cloud providers, chiefly Amazon, toward owning its own infrastructure. Goldberg framed the change bluntly: the company sees proprietary compute as a competitive asset and is prepared to spend heavily to get it.
From Leased Compute to Owned Compute
The current profitability rests on a cost curve that is about to bend the other way. Anthropic's computing bill has been falling quickly, a tailwind from cheaper inference and cheaper leased capacity. That trend reverses as the company starts buying its own hardware, Goldberg said.
The strategic logic is straightforward. Frontier model developers that depend entirely on rented capacity cede a layer of control to their suppliers. Owning the compute stack reduces that dependency and, in Anthropic's calculus, strengthens its competitive position. The tradeoff is a return to heavy capital expenditure and, by the company's own disclosure, a return to losses.
The Circular Dependency With Amazon and Google
The more complicated story is Anthropic's relationship with the hyperscalers. According to Goldberg, roughly one fifth of the cloud backlog at both Google Cloud and AWS is attributable to Anthropic. A large share of Amazon's AI revenue flows from Anthropic as a customer.
Anthropic is simultaneously a major customer, a channel partner, and a competitor. Large enterprises often purchase their Anthropic usage through a cloud provider, so the hyperscalers remain distribution infrastructure even as Anthropic builds toward independence. Goldberg noted that Anthropic's own S-1 filing describes the relationship as multifaceted and complex, which he called an understatement.
With Google the competitive tension is explicit: Google Cloud sells Anthropic models while Google also develops competing frontier models. Amazon holds a large stake in Anthropic while AWS competes for the same enterprise AI workloads. Anthropic must keep leasing from both while it builds an exit path from both.
Market Turmoil and the Value Distribution Question
The profit disclosure landed during a bruising week for tech equities. Alphabet had its worst trading day on record, and investors are questioning where AI value will actually accrue across the stack.
Goldberg rejected the zero-sum framing that frontier labs will capture everything while everyone else loses. He also cautioned against reading Anthropic's prospects off this week's market action, citing interest rates and geopolitical conflict as confounding variables. The company, he said, still has a lot to prove as a standalone public asset.
Context From the IPO Filing
The profit milestone comes against a backdrop of enormous accounting losses. As The Washington Post reported, Anthropic disclosed its financials publicly for the first time ahead of a listing that could be the largest ever, revealing rapid revenue growth alongside staggering losses. CNBC detailed that the company posted nearly $81 billion in first-half losses, driven by accounting charges tied to its rising valuation, while carrying $6.2 billion in debt at the end of June.
Revenue growth is the offsetting signal. Inc reported that quarterly revenue rose from $11.6 billion in Q2 to $16.8 billion in Q3, a pace that explains investor tolerance for the loss figures.
What It Means for the Compute Layer
For anyone mapping the AI supply chain, the signal is clear: the frontier labs are verticalizing. Leasing was the phase-one model, appropriate for experimentation and rapid scaling. Ownership is the phase-two model, appropriate for consolidation and margin defense.
The consequences propagate. Hyperscalers lose their single largest AI tenant over time, which compresses the cloud backlog narrative that has supported their valuations. Anthropic absorbs a heavier balance sheet in exchange for sovereignty over its own training and inference capacity. The market will now price that trade explicitly.
FAQ
Is Anthropic profitable?
Yes, on an operating basis. Anthropic has posted nearly a billion dollars in operating profit since Q1 2025, but it has explicitly warned that losses will return within a quarter or two as it shifts to buying its own compute.
Why is Anthropic building its own compute infrastructure?
The company views owned compute as a competitive asset, reducing its dependence on hourly leases from Amazon and Google and giving it control over its training and inference capacity.
How much of AWS and Google Cloud AI revenue depends on Anthropic?
According to Seaport analyst Jay Goldberg, about one fifth of the cloud backlog at both AWS and Google Cloud is attributable to Anthropic.