OpenAI Is Narrowing Anthropic's Lead in Business AI. Here's What Agencies Should Do

Published August 20, 2026By ABD Legacy LLC
AI tools OpenAI vs Anthropic enterprise AI business AI spending

New Ramp spending data shows Anthropic still ahead — 43.5% of paying businesses vs OpenAI's 39.7% — but OpenAI is growing faster in Q3. For agencies, the takeaway isn't "pick a winner." It's how to build for both.

The headline numbers

Anthropic still leads OpenAI among U.S. businesses paying for AI, but the gap is closing. According to the Ramp AI Index, based on spending data from more than 70,000 U.S. businesses, 43.5% paid for Anthropic subscriptions or tokens in July 2026, up 1.1 percentage points from the month before. OpenAI reached 39.7%, up just 0.23 points.

The interesting part isn't who's ahead — it's the trend. Ramp economist Ara Kharazian says OpenAI is now expanding at a faster rate than Anthropic in Q3 to date. That's the first sign of a reversal since Anthropic overtook OpenAI back in May, when it hit 41% to OpenAI's 39%. With a month of the quarter still to go, the direction matters more than the snapshot.

Why the lead is shifting: the model story

The numbers behind the shift come down to two flagship models. Anthropic's Fable 5, released in July, has had a slow start with business customers. It accounted for only 6% of the tokens businesses bought from Anthropic in its first month, and 11.4% of the dollars spent on Anthropic models. OpenAI's GPT-5.6 Sol, by contrast, made up 25% of OpenAI tokens and 23% of spend.

Price is a big part of the story. Fable 5 costs roughly $10 per million tokens — about twice the price of GPT-5.6 Sol. Kharazian was blunt on X: "GPT-5.6 Sol is really good, increasingly the choice for developers," while "Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators."

That last part is worth pausing on for anyone building AI for clients: a more expensive model can lose to a cheaper one that's "still highly performant," especially when extra compliance strings are attached. Data retention requirements didn't just affect Anthropic's costs — they affected which model businesses chose.

Context: the money behind the race

The Ramp data is one window; the companies' own numbers tell a similar story of momentum. Anthropic posted $11.6 billion in Q2 2026 revenue, surpassing OpenAI's $6.7 billion for the first time. OpenAI recorded a $12.3 billion operating loss in Q2, while Anthropic posted a small adjusted operating profit. But OpenAI's CFO, Sarah Friar, said July annualized revenue already exceeded the full Q2 total, crediting GPT-5.6, ChatGPT Work, and Codex. For more on where that momentum shows up, see how OpenAI's enterprise revenue overtook consumer revenue for the first time.

So the picture is: Anthropic has the lead and better unit economics right now. OpenAI has the momentum and the cheaper flagship. Both are enormous businesses. Neither is going anywhere.

What this means for agencies

This race matters to agencies for three practical reasons:

  1. Don't bet the practice on one platform. If you've built your agency's entire stack on Anthropic — or on OpenAI — you're exposed to a market that clearly shifts quarter to quarter. The leading position changed hands in May and may change again. Build multi-model from the start: abstract the model layer so client work can run on whichever model wins the next round, or both.
  2. Model choice is now a client conversation about price and compliance, not just quality. Fable 5's slow uptake shows that a better- or equally-capable model at half the price wins business, and that data retention requirements are a real purchasing factor. When you recommend a model to a client, be ready to talk cost per million tokens, data retention, and what happens to their data — not just benchmark scores. For clients in regulated industries, the compliance story can decide the sale.
  3. The "which AI is winning" question is a selling opportunity. Business owners read headlines like this and worry they're backing the wrong horse. That's exactly the conversation your agency can own: help clients understand the landscape, pick models that fit their budget and compliance needs, and stay portable. Positioning yourself as the neutral expert who monitors these shifts — rather than the vendor pushing one platform — is a durable service line.

A wider trend worth watching

The race isn't just OpenAI versus Anthropic. The share of Ramp-tracked businesses paying for AI rose to nearly 56% by July, up from just over 50% in March — the market itself keeps expanding. Meanwhile, 6.1% of AI-using businesses on Ramp used model-serving platforms providing access to open-source models in July, up from 4.5% in January. First-time AI buyers still choose the American labs almost exclusively, but heavy existing spenders are starting to add open-source options. For agencies, that's another reason to keep client architectures portable — the next competitor isn't a lab, it's a cheaper open-source model.

The caveats to keep in mind

Fairness requires the context Ramp itself publishes. The index draws on corporate-card and bill-pay data from more than 70,000 U.S. businesses, but the sample skews toward the tech industry and excludes large enterprises that use spend-management tools from providers like American Express. Ramp shares percentages, not dollar totals, and the model-level Fable 5 figures come from a subset of customers using Ramp's token-spend-management product — a sample that skews even more tech-heavy. And one month of Q3 remains, which Kharazian himself notes is a long time in AI. Treat these numbers as a strong directional signal, not a final verdict.

What to do next

If your agency helps clients choose and deploy AI, this is the moment to formalize that advisory: a model-selection framework, a data-retention checklist, and a default stance of multi-model portability. The agencies that win the next two years won't be the ones that picked the winning lab — they'll be the ones that could switch when the lead changed. If you're evaluating partners to help build on the right platforms, see how vetted AI agencies approach model choice and client AI strategy — browse top AI agencies for small business.

Build on both, not one.

Browse Vetted AI Agencies →

Related reading

Sources

Accuracy note: All figures come from the Ramp AI Index (August 2026), based on corporate-card and bill-pay data from more than 70,000 U.S. businesses; the sample skews tech-heavy and excludes large enterprises using providers like American Express, and model-level Fable 5 figures come from a subset using Ramp's token-spend-management product. Ramp shares percentages, not dollar totals. Anthropic/OpenAI revenue and operating figures are as reported by the companies for Q2 2026. Treat the share numbers as directional, not a final verdict.