skillindustry-news 3 min read

OpenAI Anthropic Gap Holds as Ramp Shows Q3 Catch-Up

TechCrunch reported on 20 August 2026 that Ramp data showed Anthropic at nearly 44% versus OpenAI at nearly 40% in July among paying business users, with OpenAI growing faster in Q3 to date.

PC

PromptCrates Editorial

Staff Writer

0 0
OpenAI Anthropic Gap Holds as Ramp Shows Q3 Catch-Up

TechCrunch reported on 20 August 2026 that Ramp data shows OpenAI gaining on Anthropic among paying business users. Anthropic led 41 percent to 39 percent in May and nearly 44 percent to nearly 40 percent in July, but OpenAI is growing faster in the third quarter to date.

What the OpenAI Anthropic Ramp snapshot shows

The May print is the one that flipped the usual story. Among Ramp paying business users, Anthropic hit 41 percent and OpenAI sat at 39 percent. As of July, Anthropic was nearly 44 percent and OpenAI was nearly 40 percent. Anthropic still leads that slice. The gap did not close in July. The news is the slope after July.

Ramp economist Ara Kharazian is the named voice on the Q3-to-date line: OpenAI is growing faster in this segment so far this quarter. Ramp did not share actual dollars. Percentages only. Anyone who turns those shares into a revenue race is inventing a number the source refused to give.

The same dataset says the paying-for-AI share among these Ramp customers topped 50 percent in March and reached nearly 56 percent by July. That is the market-growing-under-both-labs fact. Share fights can move while the pie gets larger. An Anthropic lead in July can sit on top of a broader rise in paid AI seats.

What this is not

Ramp is one spend graph. It is not the whole enterprise market. TechCrunch is careful that the file covers Ramp’s paying business users, not every company that buys model tokens. Do not write “Anthropic has 44 percent of enterprise AI.” Write “nearly 44 percent among Ramp paying business users as of July.”

No dollar figures were shared. A funding story can name a round. This story cannot. If a slide needs a money axis, leave it blank or mark it unavailable.

Gateway and card data also skew toward the kinds of companies that use Ramp. That is a sample, not a census. Treat it the way you treat a routing-layer rumor: useful color, labeled as the outlet and the dataset, never as a closed market.

How to write the skill so a share flip does not rewrite the stack

A skill should pick a model for the job, not for last month’s Ramp bar. If Sol is the better coding pass this week, say so in the route field. If a Claude model is the better legal pass, say that. Q3-to-date growth is not a failover policy.

Keep two routes anyway. Share leadership in this slice already sat with Anthropic in May and still did in July. Kharazian’s Q3 note says it can move again. Volatility is the finding. Stickiness is the question the data does not answer. Hold the model field as a parameter until a quarter actually ends.

If you brief a client, say Ramp, paying business users, percentages only, no dollars. The citation sentence is Ramp via TechCrunch, 20 August 2026, Ara Kharazian. May: Anthropic 41 percent, OpenAI 39 percent. July: Anthropic nearly 44 percent, OpenAI nearly 40 percent. OpenAI growing faster in Q3 to date. No dollars shared. Paying-for-AI share topped 50 percent in March and nearly 56 percent by July.

Sources

OpenAIAnthropicRampbusiness usersmarket shareAra Kharazian

Related articles