• 3 min read
Anthropic leads AI spending as revenue tops $11.5 billion
Anthropic led Vercel’s July AI spending with 65.1% of share, while reporting preliminary second-quarter revenue above $11.5 billion.

Source: Techradar
Anthropic’s Claude models generated the largest share of spending routed through Vercel’s AI Gateway in July 2026, even though the company’s tokens cost substantially more than those from rival AI labs. Separately, Anthropic told prospective investors that its second-quarter revenue exceeded $11.5 billion, according to The Next Web.
As first reported by Techradar.
The two figures measure different things. Vercel’s data covers transactions flowing through its API-routing service, while the quarterly figure is Anthropic’s own preliminary financial disclosure. Together, they point to strong demand for Anthropic’s models as the company prepares for a potential public listing.
Anthropic leads Vercel’s July AI spending
Vercel’s AI Gateway data shows Anthropic taking 65.1% of total spending in July, while accounting for 30% of token volume. The company has held more than 60% of the gateway’s spending share every month since Vercel began publishing the analysis in December 2025.
That share came despite Anthropic’s average price per token being 4.4 times higher than the overall average across the AI providers analyzed. Vercel defines price per token as total spending divided by total token volume.
Overall spending across the analyzed AI industry rose 37% in July, while the average price per token fell 13.6%. Anthropic’s average token price also declined during the month, which TechRadar attributes to the return of public access for Claude Fable 5 after a temporary suspension.
The data also shows Chinese AI company DeepSeek overtaking Google to become the second-largest provider by token volume. Vercel’s report said:
“We said an open-weight lab would soon be second by volume. In July, DeepSeek surpassed Google to take that place.”
Anthropic’s revenue jumps ahead of a possible IPO
Anthropic’s preliminary second-quarter revenue exceeded $11.5 billion, up from $787 million in the same quarter of 2025, according to documents seen by Bloomberg and reported by The Next Web. Revenue also more than doubled from $4.73 billion in the first quarter, putting first-half revenue at roughly $16.2 billion.
The company also recorded positive adjusted operating income in the quarter. The figures may still be revised, and Anthropic declined to comment.
Four quarters at the reported quarterly pace would produce about $46 billion in revenue, close to the $47 billion annualized run rate Anthropic disclosed in May. That comparison is more concrete than the run-rate figures commonly used to value AI companies.
OpenAI’s reported figure of more than $40 billion is an annualized run rate rather than quarterly revenue, and the two companies may not calculate those figures in the same way. Our earlier coverage tracked OpenAI’s enterprise revenue reaching a $40 billion annualized run rate, making the distinction especially important when comparing the companies.
Anthropic has filed confidentially for a listing and is working with Morgan Stanley, Goldman Sachs, and JPMorgan. Its backers reportedly expect a valuation of $2 trillion in October, as covered in our report on the proposed IPO valuation. An autumn debut would come before a potential OpenAI listing and before DeepSeek, which is preparing to file, according to The Next Web.
Listings have raised $256.4 billion in 2026, excluding blank-cheque vehicles—the highest total since 2021. Anthropic’s combination of gateway dominance, rapid quarterly growth, and positive adjusted operating income gives investors more than a usage story to assess, although the company has not yet publicly confirmed the final listing timetable or valuation.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.


