OpenAI is growing faster than Anthropic among American businesses in the third quarter, new data from Ramp shows, narrowing a market-share gap that opened in May. Anthropic still leads at nearly 44% of Ramp's tracked AI spend as of July, against nearly 40% for OpenAI, but the ChatGPT maker's Q3-to-date growth rate is now outpacing its rival. The data covers more than 70,000 American businesses that route billions of dollars through Ramp's bill pay and corporate card products.
The reversal in trajectory matters because Anthropic's lead is recent. OpenAI was the runaway leader with both consumers and businesses until May, when Anthropic hit 41% share to OpenAI's 39% on Ramp — the first time the ChatGPT maker had lost the top spot. OpenAI never regained the lead through the summer, but the July-to-August window suggests the gap is closing again rather than widening.
Ramp's customer base skews toward tech and Silicon Valley firms, since the corporate card is popular with startups, and the data excludes large enterprises that use spend-management tools from providers like American Express. Ramp declined to disclose absolute dollar figures, sharing only percentage shares. That makes the data a directional signal rather than a full market picture, but it is one of the few near-real-time reads available before either OpenAI or Anthropic approaches an IPO and opens its books.
Key facts
- 01Anthropic held nearly 44% of Ramp business AI spend in July to OpenAI's nearly 40%, down from a 41% to 39% split in May.
- 02OpenAI is growing faster than Anthropic in Q3 to date among Ramp's 70,000 tracked American businesses.
- 03The share of Ramp customers paying for AI climbed from over 50% in March to nearly 56% by July.
- 04Anthropic's Fable 5 tier faces a 30-day data retention requirement imposed by regulators, dampening adoption.
- 05Ramp data excludes large enterprises served by American Express and similar tools, skewing toward tech-heavy customers.
Ramp economist Ara Kharazian tied OpenAI's renewed momentum to its latest release. Developers are gravitating toward the new model, according to Ramp's read of spending patterns, and that developer pull is showing up in corporate card charges rather than just consumer subscriptions.
“GPT-5.6 Sol is really good, increasingly the choice for developers”— Ara Kharazian, Ramp economist
On the Anthropic side, the story is more complicated. Fable 5, the company's higher-end model tier, has underperformed on Ramp's data. Kharazian pointed to two factors: the price point and a 30-day data retention requirement that Anthropic disclosed to Fable users earlier this year, imposed by regulators. That warning drew user pushback when it landed.
The Fable 5 framing may oversimplify. Anthropic built the tier for a narrower set of enterprise use cases than a general-purpose chatbot, and comparing its adoption directly to OpenAI's flagship model overlooks that scoping. Anthropic's Claude products more broadly continue to hold the top share on Ramp, and the company's growth story remains intact even if the newest tier has landed softly.
“Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators”— Ara Kharazian, Ramp economist
The broader Ramp signal is that the AI market itself is still expanding fast enough that both companies can grow revenue while trading market share. The share of Ramp customers paying for any AI product topped 50% in March and reached nearly 56% by July. That is a rising tide, not a zero-sum fight, and it explains why both OpenAI and Anthropic can post strong topline numbers even as their relative positions swing month to month.
The volatility itself is the most useful takeaway for investors. Businesses appear willing to flip vendors as each lab ships a new frontier model, which cuts against the assumption that enterprise AI spending is sticky the way legacy software contracts are. A developer team that standardized on Claude in the spring is apparently comfortable switching to GPT-5.6 Sol in the fall. Multi-model routing tools like OpenRouter, which Stripe is reportedly acquiring for $7B or more, exist precisely because that swapping is now cheap.
There is still a month left in Q3, which as Kharazian noted is roughly 30 AI years at the current release cadence. Anthropic could ship a Claude update that reverses the trend before quarter-end, and OpenAI's lead in growth rate could evaporate as quickly as it appeared. The next data cut from Ramp will be more instructive than this one.
For OpenAI and Anthropic, the Ramp numbers reinforce a strategic reality both companies already understand: model quality wins share on a release-by-release basis, and neither lab can rest on a prior generation for more than a quarter. For the enterprises writing the checks, the churn is a feature, not a bug — the moment any single provider builds durable pricing power, the multi-model tooling gets sharper and the switching cost falls further. That dynamic keeps margins compressed at the model layer even as the total pie grows, which is why both labs are pushing hard into products, agents, and vertical acquisitions rather than betting the business on API revenue alone.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
