Anthropic crosses OpenAI in U.S. business AI adoption — first time in the Ramp Index

What Ramp’s May 13 Index actually shows

Ramp publishes a monthly AI Index based on corporate-card and invoice-based spending across approximately 50,000 U.S. companies. The May 13, 2026 release reported the following for April 2026 business AI usage:

• Anthropic: 34.4% (up 3.8 percentage points month-over-month)
• OpenAI: 32.3% (down 2.9 percentage points month-over-month)

This is the first time Anthropic has crossed OpenAI in the index since Ramp began tracking AI adoption in 2023. The 12-month trajectory is the more striking number: Anthropic roughly quadrupled its business adoption over the trailing year; OpenAI grew approximately 0.3 percentage points.

The methodology is worth understanding. Ramp doesn’t measure “which AI does your company use” via survey. It measures which AI vendor your company pays. That distinction matters — survey responses capture sentiment; payment data captures committed budget. Companies do not pay AI vendors $10K–$10M/year for fun; the spend follows real, recurring usage. That’s why investors and ops teams watch this index more closely than they watch most “X% of CEOs say…” surveys.

The other significant detail in the May index: the fastest-growing vendor category is AI-inference platforms offering access to cheap open-source models (Together AI, Fireworks, Modal, etc.). The “frontier model duopoly” framing of 2024 doesn’t quite hold in 2026 — open-source inference is the third growth vector competing with Anthropic and OpenAI for the same enterprise wallet share.

The $30 billion context

The Ramp index sits inside a broader Anthropic financial story. On April 7, 2026, Anthropic disclosed a $30 billion annualized run-rate revenue, up from approximately $9 billion at the end of 2025. The trajectory from January 2025 ($1 billion run-rate) to April 2026 ($30 billion) is a 30x jump in 15 months — among the steepest revenue curves any software company has posted publicly.

A few specifics from the disclosure that matter:

• 1,000+ enterprise customers paying $1M+ annually, a figure that doubled in recent months
• Revenue mix is “almost entirely” enterprise API consumption + developer tooling (Claude Code and the API business)
• Internally, roughly 90% of code at Anthropic is now written by Claude Code — the company eats its own dogfood at a depth that compounds the development pace

The April SpaceX/Colossus 1 capacity deal, the Wall Street agent templates push (JPMorgan, Goldman, Citi, AIG, Visa), the Microsoft 365 GA integration, and the $40 billion Google investment all support the same trajectory: Anthropic is winning the enterprise procurement contest faster than most observers projected six months ago.

What Claude for Small Business adds this month

In parallel with the index publication, Anthropic shipped Claude for Small Business — a packaged offering bringing Claude into the apps small businesses actually live in:

• Accounting: QuickBooks, with workflows for invoicing, month-end close, and payroll prep
• Payments: PayPal, with transaction analysis and fraud-flagging templates
• CRM and marketing: HubSpot (lead qualification, sequence generation), Canva (brand-asset generation)
• Document workflows: Docusign (contract review and clause analysis)
• Productivity: Google Workspace and Microsoft 365 (the integration that landed GA in early May)

The strategy is clear: Anthropic’s growth path from $30B → larger doesn’t run through more model-only spend at large enterprises; it runs through owning the AI layer inside the tools small and mid-sized businesses already use. The SMB segment is where OpenAI has historically led via ChatGPT Plus subscriptions. This is the first time Anthropic has shipped a coherent SMB push, and it lands during the same week the adoption index crossed over.

The three threats that could erase the lead

VentureBeat’s analysis of the Ramp data named three structural risks to Anthropic’s adoption lead, all of which are worth taking seriously:

1. Capacity remains the binding constraint. Anthropic’s April rate-limit doubling helped, but enterprise demand is outpacing supply faster than the compute deals close. If a major enterprise can’t get the Opus 4.7 throughput it needs, it routes to OpenAI by default. The SpaceX Colossus 1 capacity (300+ MW) starts mitigating this in Q2/Q3 2026 but doesn’t close the gap fully.

2. Google ecosystem integration depth. Gemini sits inside every Google Workspace tenant by default, and Google is pushing Gemini Intelligence to move across apps autonomously (the agent layer announced in early May 2026). Workspace customers experiencing “Gemini just works in Gmail and Docs” don’t switch to Anthropic for marginal quality gains. The Microsoft 365 GA gives Anthropic parity inside the Microsoft ecosystem; Google’s footprint is still its own.

3. OpenAI’s enterprise sales motion is finally engaged. Through 2024-2025, OpenAI’s sales motion was lighter-touch, biased toward developer self-serve. In 2026, OpenAI has built a real enterprise field team, GPT-5.5 has closed most of the writing-quality gap that drove Claude adoption, and the recent GPT-5.5 Instant default upgrade reduced hallucinations on high-stakes prompts by ~52%. The product gap that justified switching is genuinely narrower than it was 12 months ago.

The pragmatic read on these three threats: each is real, none is fatal alone, but combined they could narrow the gap by Q4 2026. The index might not stay tilted this way through the rest of the year.

What this means for Pick Right readers

If you’re a ChatGPT user thinking about Claude: the structural case is stronger than it was six months ago. Enterprises with real budget are voting for Claude at the margin. For writing quality and coding, this matches Pick Right’s standing recommendation. See the head-to-head Claude vs ChatGPT comparison for the use-case-by-use-case breakdown.

If you’re an Anthropic API customer: the rate-limit doubling that landed May 6 is the most direct impact for you. Capacity should continue improving as the SpaceX Colossus 1 GPUs come online through Q2.

If you run a small business considering AI tooling: the Claude for Small Business rollout makes Anthropic genuinely accessible without enterprise procurement friction. If you already use QuickBooks, PayPal, or HubSpot, the integration is now native. For broader context on AI tools for small operations, see the small business AI guide.

If you’re tracking the AI market more broadly: the Ramp index is the cleanest publicly available measure of where committed budget is landing. May 13’s reading is a real shift — not a survey-quality “vibe” reading. Watch the June and July reports to see whether Anthropic’s lead consolidates or whether the three threats start narrowing it.

The pragmatic read

The market structure of 2026 AI is more interesting than “frontier model duopoly” framing suggests. Anthropic genuinely leads on enterprise procurement now — that’s measurable, not aspirational. OpenAI’s product breadth and ecosystem (Custom GPTs, image generation, voice) still wins for consumer breadth and for organizations whose work spans many modalities. Google sits behind both on raw chat usage but ahead on integration depth inside Workspace. Open-source inference is growing faster than any of the three but from a smaller base.

For a publication that tries to give recommendations rather than affiliate-driven rankings: the Ramp data is one of the better signals to weight, because it captures what enterprises actually pay for rather than what they say in surveys. The May 13 reading shifts the structural picture — and confirms that the “Claude is better for serious work” recommendation Pick Right has stood on for two years is increasingly the consensus position at the procurement layer.

For broader background, see the Claude review, the ChatGPT review, and the head-to-head Claude vs ChatGPT comparison.

Questions or corrections? Email Pick Right. Want the full list? See all news.