Anthropic Overtakes OpenAI in Business AI Adoption (2026)

For the first time, Anthropic has passed OpenAI in US business AI adoption — but only by one measure. Ramp's latest AI Index, drawn from corporate-card and bill-pay spend at 70,000+ companies, puts Anthropic at 41% of businesses against OpenAI's 39.5%. A separate IDC enterprise survey says the opposite. The gap between those two numbers is the real story.

TL;DR

• In the latest Ramp AI Index (data through May 2026, published June 9), Anthropic reached 41% of US businesses and OpenAI slipped to 39.5% — a 1.5-point lead for Anthropic.¹
• Ramp first reported the crossover in its May 13 release, when Anthropic hit 34.4% and OpenAI 32.3% under the old methodology; Ramp called it the first time Anthropic had passed OpenAI.²
• In Ramp's current series, Anthropic's share climbed from 10.6% (Dec 2024) to 41.0% (May 2026) — a near-fourfold rise — while OpenAI moved from 33.0% to 39.5%.³
• The Ramp Index measures spend, not seats, and skews toward US mid-market, growth-stage, tech-forward firms — so it is a signal, not a census.⁴
• A different lens disagrees: IDC's March 2026 enterprise survey put OpenAI at ~42% of organizations and Anthropic far lower. Methodology decides the winner.⁵

What You'll Learn

• What the Ramp AI Index actually counts, and why it shows Anthropic ahead
• The month-by-month climb that produced the crossover
• Why the "overtook" headline needs an asterisk after Ramp's methodology change
• How IDC's survey reaches the opposite conclusion — and why both can be right
• The three headwinds Ramp itself flagged for Anthropic
• What the heaviest AI spenders are actually paying per employee

What the Ramp AI Index Measures

The Ramp AI Index is not a survey. It is built from aggregated, anonymized corporate-card and bill-pay transactions across more than 70,000 US businesses on Ramp's spend platform. A firm counts as "adopting" a vendor if it has a positive payment to that vendor's AI product in a given month, with products identified from merchant names and line-item receipt detail.¹⁴

That design is the Index's strength and its limit. Because it watches money move rather than asking people what they use, it catches purchases the moment a card is charged — a startup buying API credits, a mid-market team expensing Claude, an SMB on ChatGPT Team. But Ramp is candid that its customer base skews US-based, mid-market, growth-stage, and tech-forward, and that most US firms are smaller than its typical customer. It is a fast, grounded signal of where business AI dollars flow — not a representative census of every American company.⁴

The Index is authored by Ramp's lead economist, Ara Kharazian, whose spend analyses have been cited by the New York Times, Wall Street Journal, and Financial Times.¹

The 18-Month Climb

Pulling Ramp's vendor-adoption series month by month shows this was a steady takeover, not a single jump. The table below tracks the share of businesses paying each vendor, from the end of 2024 through May 2026.³

Two patterns stand out. OpenAI's business reach has been essentially flat for a year — holding in a narrow band of roughly 39–41% since mid-2025. Anthropic's, by contrast, nearly quadrupled over the same window. The lines didn't cross because OpenAI fell; they crossed because Anthropic kept climbing while OpenAI stalled. Ramp's own framing is blunter: over the year leading into the crossover, Anthropic quadrupled its business adoption while OpenAI grew just 0.3%.²

For context, no other vendor is close in this dataset. In May 2026, Google sat at 6.1%, xAI at 3.1%, and DeepSeek at 0.3% of businesses — vendor shares overlap because many firms pay more than one.³

Why "Overtook" Needs an Asterisk

There is a wrinkle worth being precise about. Ramp announced the crossover on May 13, 2026, reporting April data: Anthropic at 34.4% (up 3.8%) versus OpenAI at 32.3% (down 2.9%), and calling it the first time Anthropic had passed OpenAI.² Then on June 9, 2026, Ramp shipped a reworked version of the Index with "methodological updates to better capture enterprise spend on OpenAI and Anthropic," restating the levels to 41% and 39.5%.¹

So the exact percentages depend on which release you read, and the older 34.4%/32.3% figures and the newer 41%/39.5% figures are not directly comparable. What survives the revision is the direction: in both the original and the revised series, Anthropic is ahead, and Ramp reports the lead at 1.5 points as of May 2026. Treat the crossover as a spring-2026 event confirmed across two methodologies — not a single, fixed-to-the-decimal number.

The Other Dataset Says the Opposite

Here is the part most headlines skip. A different measurement reaches a different conclusion. IDC's FERS survey from March 2026 found OpenAI used by roughly 42% of organizations and Google by about 38%, while only 19% of enterprises reported using Anthropic's models extensively and 25% were still evaluating them.⁵

Both can be true at once, because they count different things. Ramp watches discrete card and invoice spend, which captures bottoms-up, developer-and-team purchases the moment they happen — exactly where Claude has surged, especially in coding. IDC surveys organizations, which leans toward formal, top-down enterprise standardization and larger accounts where OpenAI's enterprise distribution runs deep. Spend-based data sees Anthropic's grassroots momentum first; survey-based data reflects entrenched enterprise relationships. The honest read is that Anthropic leads on bottoms-up business spend while OpenAI still leads on broad organizational penetration — and anyone citing a single "who's winning" number is choosing a methodology, not settling a fact.

Three Headwinds Ramp Flagged for Anthropic

Notably, the lead came with a warning label from Ramp itself. Kharazian explicitly cautioned against reading the crossover as proof Anthropic is the "definitive leader," and listed three risks.²

First, incentive misalignment: Anthropic earns more when businesses consume more tokens, which nudges users toward pricier models even when cheaper ones would do — a real concern as firms hit AI budgets. Second, reliability complaints: through spring 2026, users reported outages and rate limits (Ramp points to April 2026 reporting in Fortune); Anthropic reset usage limits in April and signed a SpaceX data-center deal to ease compute — part of the wider AI-infrastructure build-out behind SpaceX's blockbuster IPO. Third, cost pressure from model changes that can raise per-prompt token costs. Underneath all three sits the cheap-model threat: some of the fastest-growing vendors on Ramp are inference platforms serving low-cost open-weight models, and rivals like Moonshot's open-weight Kimi K2.7-Code undercut frontier labs on price with minimal switching cost.²

What "AI-Pilled" Companies Actually Spend

The reworked June Index shifts focus from whether firms use AI to how intensely. Ramp's answer is a wide spread. The top 1% of firms by AI spend per employee pay about $7,449 per employee per month; the top 10% pay $611; the median firm pays just $11.38 — roughly one enterprise ChatGPT or Claude seat.¹

Even the heaviest spenders aren't matching the "spend as much on AI as on an engineer's salary" advice making the rounds — the top 1% still pay less than half a typical engineer's monthly salary on AI. Spend is rising fast, though: the top 1% grew per-employee AI spend 14.1% in a single month. And these leaders don't lock in — advanced adopters deliberately run multiple frontier models plus open-source access platforms, which is precisely why a vendor's "lead" can swing more than 20 points in roughly a year and a half. That same low-friction, multi-vendor reality is now reaching consumers, too, as iOS 27 lets users bring third-party models like Claude into Apple Intelligence.¹

The Bottom Line

Anthropic overtaking OpenAI is a real milestone with a precise scope: bottoms-up business spend, as measured by one fast-moving index, in spring 2026. The trajectory is unambiguous — Anthropic nearly quadrupled its business reach while OpenAI held flat — and that momentum, concentrated in coding and developer teams, is the genuine signal. But the headline hides three asterisks: Ramp changed its own ruler mid-stream, IDC's survey still shows OpenAI leading on organizational penetration, and Ramp itself listed three ways Anthropic's lead could evaporate. The useful takeaway isn't "Anthropic won." It's that business AI adoption is now volatile enough to flip a 22-point lead in roughly a year and a half — which means the next reversal may already be underway.