Anthropic Captures 70%+ of New Enterprise AI Deals vs OpenAI, Per Ramp Data

73% — Anthropic Just Took the Enterprise First-Time-Buyer Market

Here's the deal: on May 4, Semafor cited Ramp's payments data to land a single line — "73% of businesses buying AI tools for the first time are choosing Anthropic." That's a 70:30 lead over OpenAI, up from 50:50 about 10 weeks earlier, and from a 60:40 OpenAI lead in December. The same week, both labs announced PE-backed enterprise JVs: Anthropic's $1.5B with Blackstone, Goldman, and Hellman & Friedman; OpenAI's $10B Deployment Company with TPG, Brookfield, and Bain. Not coincidental — both labs are racing in enterprise services. Anthropic taking 73% of new deals signals a structural shift: the SaaS and AI application industry is converging toward Claude as the default LLM over the next 12-24 months.

The Players — Ramp, Anthropic, OpenAI, Enterprise IT Decision-Makers

Ramp is a U.S. corporate spend management platform with payments data from 350K+ U.S. SMBs and mid-market companies. Their Ramp AI Index is the single most reliable public-facing market-share indicator for enterprise AI in the U.S. Payments-data based, so it captures actual revenue flow rather than self-reported intent.

Anthropic is led by Dario Amodei. Founded in 2021 by 7 ex-OpenAI engineers. Hit $10B ARR in 2024, $50B in 2025 (5×), and an estimated $80-100B run-rate as of Q1 2026. Claude and Claude Code are the core products, with Claude Code alone generating $20-30B ARR. The 73% share reflects how Claude's coding and analytical-reasoning advantages are translating into actual market choices.

OpenAI is led by Sam Altman. ChatGPT (Nov 2022) made OpenAI the global AI baseline. ARR hit $10B in 2025, ~$130-150B run-rate in Q1 2026 — still 1.5× Anthropic's absolute revenue. But on new enterprise deals, OpenAI is at 30% — and ChatGPT consumer makes up 75% of revenue, leaving the deep-enterprise B2B layer comparatively weaker.

Enterprise IT decision-makers (CTO, CIO, VP Eng) cite three reasons for choosing Claude: superior coding performance (SWE-bench, HumanEval), Constitutional AI safety narrative (matters for finance, healthcare, legal), and 200K-1M token context windows for enterprise document workflows.

Per Semafor, businesses choosing AI tools for the first time picked Anthropic at 73% as of May 4 — up from 50:50 ten weeks earlier and 60:40 OpenAI in December.

The 10-Week Reversal Curve

Snapshot	OpenAI	Anthropic	Net Move
2025-12	60%	40%	OpenAI +20pt
2026-02 (~10 wks ago)	50%	50%	Tied
2026-05-04	27%	73%	Anthropic +46pt

A 46-point net swing in 10 weeks is unusually fast for SaaS. It's the convergence of multiple drivers, not one variable.

Driver 1: Claude Opus 5 launch (mid-April 2026). Anthropic shipped Opus 5, hitting 90% on SWE-bench Verified — clearly ahead of GPT-5.4 (85%) and Gemini 3 (82%). For coding workloads, that's the moment Claude default solidifies.

Driver 2: Claude Code adoption + cost-efficiency. Claude Code became the default coding agent, and GitHub Copilot, Cursor, and Cody now route significant traffic to Claude as a backend. Enterprise teams self-hosting or routing API directly naturally pick Anthropic.

Driver 3: OpenAI's consumer concentration. With 75% of OpenAI revenue from ChatGPT Plus/Team/Enterprise (consumer/SMB-skewed), the deep enterprise layer (large-company IT) has been Anthropic-favorable. The 73% statistic is on new enterprise deals; OpenAI still dominates the 150M+ ChatGPT Plus consumer base.

Driver 4: Same-week PE JVs. Anthropic's $1.5B with Blackstone/Goldman/H&F vs. OpenAI's $10B Deployment Company. The 6.7× capital ratio for OpenAI signals OpenAI needs more capital deployment to push back on the share reversal.

Who Wins — Anthropic, OpenAI, PE Sponsors, Enterprise Industry

Anthropic wins three ways. Enterprise default = ARR ramp acceleration: 73% new-deal share could push the ARR curve from $80B → $200-250B over the next 12-24 months. Valuation re-rating: informal valuation moves from $35-45B toward $70-90B. Compounded narrative: SpaceX compute deal + 73% enterprise + Constitutional AI safety positioning combine into the new global AI standard story.

OpenAI faces mixed outcomes. ChatGPT consumer revenue ($100B+) still solid. New enterprise share dropped to 30%, weighing on valuation and capital strategy. The $10B Deployment Company JV is OpenAI's response — using PE capital to push enterprise share back. If 50% share isn't recovered in 12-18 months, "OpenAI dominance" narrative permanently weakens.

PE sponsors (Blackstone, Goldman, H&F / TPG, Brookfield, Bain) get into "AI implementation services" — the next-gen consulting market. Application of AI to accounting, legal, banking, healthcare, manufacturing is forecast at $50B-1T over five years.

Enterprise SaaS industry shifts toward Claude defaults. With 73% of new buys going to Anthropic, downstream SaaS vendors (Salesforce, Microsoft, Workday, ServiceNow) lean toward Claude integration over the next 6-12 months.

Past Parallels — Wins and Losses

AWS enterprise share expansion (2010-2015): AWS established enterprise default before Azure/GCP launched and held 35-40% cloud share for a decade. Anthropic's new-deal default could establish a similar long-term lead.

Salesforce vs. Siebel CRM (2003-2008): SaaS-CRM Salesforce reversed Siebel's on-premise lead in five years. AI may follow the same shape with Anthropic catching OpenAI.

Slack vs. Microsoft Teams (2017-2024): Slack lost the messaging default to Teams via Office 365 bundling over seven years. OpenAI could potentially reclaim share through Office 365/Azure bundling — bear-case scenario.

Apple Maps vs. Google Maps (2012-2014): Apple Maps lost share back to Google in six months. Default leads aren't permanent. If Anthropic's Opus 6 ramp slips and OpenAI's GPT-6 is competitive, share could reverse again.

Counter-Plays — OpenAI, Google, Microsoft, New Entrants

OpenAI counters two ways. Accelerated GPT-6 ramp (Q4 2026 launch target) — push SWE-bench above 92% to retake the coding default. OpenAI Deployment Company $10B — packaged "implementation services + price cuts" delivered directly to new enterprise buyers. PE capital fueling 12-18 months of share recovery attempts.

Google DeepMind counters with Gemini 3 (Q3 2026 expected) — possibly ahead of GPT-6, with coding parity targeted. Google Workspace and GCP bundling open channel-distribution advantages neither Anthropic nor OpenAI have.

Microsoft plays both sides. Azure hosts OpenAI models, but Microsoft Office Copilot offers Anthropic Claude integration as an option. Vendor-neutral strategy maximizes leverage. Microsoft's own Phi line provides a third option.

New entrants (xAI, Mistral, DeepSeek, MiniMax, Reflection) benefit from "OpenAI/Anthropic duopoly → real multi-vendor market." xAI's Grok 4/5 takes federal share via Pentagon channels; Mistral and DeepSeek compete in EU and emerging markets. The 12-24 month equilibrium could be Anthropic 50-60% / OpenAI 25-30% / others 10-20%.

What Changes — Devs, Founders, Investors, End Users

Devs: "Claude default" era accelerates. Cursor, Windsurf, Copilot, Cody adopting Claude as backend default shifts coding workflows. Copilot's Claude share could rise to 60-70%.

Founders: Claude API gets prioritized in new SaaS integrations. ~70% of new SaaS over the next 12 months treats Claude as primary, OpenAI as secondary — flipping a 24-month-old default.

Investors: Anthropic re-rate (cited above). OpenAI ceiling questioned at $500B+ valuation. IPO timing/pricing affected.

End users: limited direct impact. Indirectly, ChatGPT Plus pricing pressure rises, possibly leading to consumer price cuts or expanded free tiers. Claude.ai consumer base could ramp toward 100M users in 12 months.

Stakes

• Wins: Dario Amodei (Anthropic CEO) — 73% new enterprise share + ARR $80B → $200B+ trajectory; Blackstone/Goldman/H&F — Anthropic PE JV gets into next-gen consulting; Claude Code users + integrating SaaS — default position strengthens.
• Loses: Sam Altman (OpenAI) — new enterprise dropped to 30%; Microsoft Azure-OpenAI single-stack — Anthropic multi-cloud diversification weakens lock-in; ChatGPT consumer-dominance narrative — clearly doesn't translate to deep B2B.
• Watching: Salesforce, Workday, ServiceNow — Claude vs. GPT default integration calls; Korean Naver, Kakao, Lunit — Claude integration vs. own-LLM ramp; Pentagon IL6/IL7 — how OpenAI/xAI fill the Anthropic-excluded space.

The Skeptics — "Ramp Data Skews to U.S. SMBs, Not Truly Representative"

Benedict Evans-style market analysts argue Ramp data is U.S. SMB-centric and misses true large-enterprise (Salesforce, Workday-class) decision-making. Where the largest decisions land matters more than what shows up in payments aggregation.

Patrick McKenzie-style SaaS analysts point to "churn vs. new acquisition" math. Even if 73% of new buys go to Anthropic, low OpenAI churn (~5%) means total market share could re-converge toward 50:50 over 12 months. The "73% new" headline doesn't translate cleanly to "OpenAI defeated."

Two skeptic lines: (1) Ramp data representativeness, (2) churn stability dampens total-market swing. Both push back against an "Anthropic 73% = permanent win" reading.

TL;DR

• New enterprise AI deals: Anthropic 73% — reversed from 50:50 to 70:30 in ~10 weeks (Ramp payments data).
• Same week PE JVs: Anthropic $1.5B (Blackstone/Goldman/H&F) and OpenAI $10B (TPG/Brookfield/Bain).
• Drivers: Claude Opus 5 coding lead + Claude Code default + Constitutional AI safety positioning.

References

• OpenAI, Anthropic ramp up enterprise push — Semafor
• Anthropic Gains On OpenAI Amid Rising Adoption Among Enterprises — PYMNTS
• Anthropic and OpenAI launching joint ventures for enterprise AI services — TechCrunch