Anthropic / Wikimedia Commons (Public domain)
Anthropic now earns $2.08 for every $1 it spends on compute
The Claude maker's compute efficiency has jumped from $0.40 per dollar in early 2025 as it reportedly prepares for a public listing
Anthropic generated $2.08 in revenue for every dollar it spent on compute in the second quarter of 2026. In the first quarter of 2025, that figure was $0.40.
That works out to more than a fivefold improvement in about a year.
The numbers behind the efficiency jump
The broader first-half figures show how fast the business scaled. Anthropic reported $16.3 billion in revenue for H1 2026, compared with $1.1 billion in H1 2025.
Over the same six months, the company spent nearly $9 billion on compute and infrastructure. That puts its first-half efficiency at roughly $1.70 to $1.81 in revenue per compute dollar.
The second-quarter reading of $2.08 sits above that half-year range. That suggests the efficiency gains accelerated as the year went on, rather than arriving all at once.
Growth kept going into the next quarter. Q3 2026 revenue reached $16.8 billion, up approximately 45% from Q2.
Profitable, depending on how you count
The headline efficiency figure comes with a large asterisk on the bottom line. Anthropic reported an $80.7 billion net loss for H1 2026.
Most of that loss came from non-cash accounting charges tied to convertible instruments.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Strip those charges out and the picture changes. Adjusted figures show a second consecutive quarter of profitability, with operating margins of 11% in Q3 2026.
The company also recorded positive free cash flow in Q3.
Background: cloud bills instead of concrete
Anthropic, best known for its Claude family of AI models, has taken a different route on infrastructure than some rivals. Rather than pouring capital into its own data centers, it has leaned on cloud partnerships.
Those partnerships are enormous. The company has secured long-term cloud commitments estimated at between $450 billion and $600 billion with Google Cloud, AWS, and Microsoft.
Approximately $450 billion of that is in non-cancelable deals spanning the next 7 to 10 years.
What this means for the IPO and the AI trade
Anthropic filed a draft S-1 in October 2026 and is gearing up for a potential IPO in November 2026.
The company is reportedly targeting a valuation as high as $2 trillion.
Anthropic’s revenue is concentrated among a small number of major customers. Those customer concentration risks interact badly with fixed cloud commitments. A company with hundreds of billions in non-cancelable obligations has limited room to absorb a sudden revenue dip.
The key items to watch ahead of the listing are straightforward. Can revenue per compute dollar keep rising? Does adjusted profitability hold through another quarter? And will the S-1 shed more light on how dependent Anthropic is on its largest clients?