Anthropic models took 70.1% of spend on proprietary AI models routed through Databricks’ Unity Gateway in the week of 2 November, by our sums of the model-level figures in the Databricks AI Index, which Databricks published on 16 November. OpenAI’s share was 27.4%, up from 10.1% in the week of 20 July.
Counted in tokens rather than dollars, Anthropic had 61.5% of proprietary-model volume and OpenAI 37.0%. Our number: Anthropic’s spend share sits 8.6 points above its token share, while OpenAI’s sits 9.6 points below it, by our arithmetic.
Business Pill · Concentration explained
An evergreen Business Engineer explainer on concentration, when a few big pieces make the total. It teaches the general idea only and does not discuss the Databricks AI Index.
The key insight: As we read the Databricks AI Index, Anthropic’s position is larger in dollars than in tokens: by our sums it took 70.1% of list-price spend on proprietary models through Unity Gateway in the week of 2 November but 61.5% of tokens, while OpenAI took 37.0% of tokens and 27.4% of spend.
- Open-Weight AI Hits 31% of Enterprise Tokens; DeepSeek Rises: open-weight models went from 7.3% to 30.9% of tokens to Databricks-hosted models on Unity Gateway between the weeks of 20 July and 2 November, and DeepSeek from 0% to 28.2% of open-weight tokens, by our sums.
- Median AI User’s Weekly Spend Rose 6x Since July: Databricks: the median user’s weekly spend went from $0.40 to $2.49 and P99 from $511.44 to $699.74, 6.2x against about 1.4x by our arithmetic.
- 71% of Firms Use 2+ AI Model Providers; OpenAI Among Adds: the share of active Unity Gateway companies that had used two or more model providers went from 36% to 71.1%; Databricks CEO Ali Ghodsi posted 75%.
What Databricks Published
Databricks co-founder Patrick Wendell wrote on X that the index shows “trends from enterprises who manage AI with our Unity Gateway” and that the data “covers >10,000 companies with broad representation (Digital Natives, Fortune 500, Global 2000, etc…)”. The index page says it is based on “aggregated, anonymized data from thousands of enterprises using Unity Gateway”.
The spend chart prices tokens “at each provider’s published list price, with cache reads at the cache-read rate”, and the page notes that this is “not what Databricks bills”. Both the token and spend charts include traffic routed to external model providers. The page plots weekly values from the week of 20 July to the week of 2 November.

What Was Posted on X
Patrick Wendell, a Databricks co-founder, opened a six-post thread at 17:15 UTC on 16 November announcing the Databricks AI Index, which he says covers more than 10,000 companies that manage AI through Databricks’ Unity Gateway. This piece uses the index page’s own chart values for every share it reports; the provider totals are our sums of the page’s model-level figures.
In the second post Wendell says 75% of companies now use two or more model providers and names OpenAI’s GPT models and open-source models as the most common new ones. The page’s chart shows 71.1% for the week of 2 November; this piece reports both figures.
Anthropic’s Share, Model by Model
Opus 5.5 alone took 35.8% of proprietary-model tokens and 34.7% of spend in the week of 2 November, the page’s figures show. It first appeared in the week of 21 September at 7.2% of tokens. Sonnet 5.5 had 11.5% of tokens in the latest week.
Anthropic’s spend share fell from 85.6% in the week of 20 July to 70.1%, a drop of 15.5 points, and its token share from 80.0% to 61.5%, by our sums. Opus 4.8 alone had 49.8% of tokens in the week of 20 July and 1.3% in the week of 2 November.

OpenAI’s Climb
Over the same 15 weeks OpenAI’s token share rose about 2.3x, from 16.4% to 37.0%, and its spend share about 2.7x, by our arithmetic. GPT 6.1 Sol was its largest model by tokens in the week of 2 November, at 14.4%. Gemini and Grok together had 1.4% of tokens and 2.7% of spend.
Wendell wrote that the most common new providers “are OpenAI (GPT models) and OSS models”. His thread says 75% of companies now have two or more providers; the page’s chart shows 71.1% of active companies using two or more in the week of 2 November, up from 36% in the week of 20 July, and 47.9% using three or more.
Our earlier pieces covered Menlo Ventures’ enterprise AI spend figures, Anthropic’s share price against its run rate, Claude models in Microsoft 365 Copilot and a Berkeley paper on agent harness cost.
The Structural Read
As we read the model lines, Anthropic’s share now rests heavily on one model: Opus 5.5 took 35.8% of proprietary tokens and 34.7% of spend in the week of 2 November, six weeks after it first appeared in the data, at 7.2% of tokens in the week of 21 September.
As we read the trend, OpenAI’s gains came as companies added providers: the page shows 71.1% of active companies on two or more providers in the week of 2 November, and Wendell names OpenAI’s GPT models among the most common new ones.
Patrick Wendell, Databricks co-founder, on X
“Enterprise AI consumption is changing quite a bit”
Three Implications
FOR BUYERS The page prices spend at list price and says it is not what Databricks bills, so negotiated discounts sit outside these shares.
FOR MODEL MAKERS Anthropic’s spend share fell 15.5 points in 15 weeks by our sums, while OpenAI’s rose from 10.1% to 27.4%.
RELATED Menlo Ventures’ enterprise AI spend figures give a survey view of the same market.
The Business Engineer Lens
This story maps onto the Business Engineer framework Tokenomics: The Economics of AI.
The framework argues that the token “is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read it, the index shows why the two counts differ: the same token carries a different list price at each lab, so Anthropic’s share of spend runs above its share of tokens. The framework predates the index and does not discuss it.
What Is Not Established
The figures describe companies that manage AI through Unity Gateway, as the page and Wendell’s thread say, so they show that customer base rather than every enterprise. Spend shares are at list prices, which the page says are not what Databricks bills, and the provider totals here are our sums of the page’s model-level values.
We read the index page in full, including the values behind each chart, and Wendell’s six-post thread. We did not contact Databricks, Anthropic or OpenAI for this piece.
The Bottom Line
Anthropic models took 70.1% of list-price spend and 61.5% of tokens on proprietary models routed through Unity Gateway in the week of 2 November, by our sums. OpenAI’s spend share rose from 10.1% to 27.4% in 15 weeks.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. All figures come from the Databricks AI Index page at developers.databricks.com/ai-index, which we read in full on 16 November 2026, including the exact values its charts plot, and from Databricks co-founder Patrick Wendell’s six-post thread on X of the same day. The page marks these charts as updated on 2 November 2026 and covers traffic through Databricks’ Unity Gateway; spend shares are at providers’ list prices, not Databricks’ billing.
Provider totals, the gaps between token and spend shares and the growth multiples are our arithmetic from the page’s model-level values. Databricks, Anthropic and OpenAI were not contacted for this piece.
Sources: Databricks AI Index (page updated 2 Nov 2026, published 16 Nov 2026) · Patrick Wendell on X, 16 Nov 2026 (post 1 of 6) · Patrick Wendell on X, 16 Nov 2026 (post 2 of 6)









