The enterprise AI leaderboard just flipped. Fintech firm Ramp released its May 2026 AI Index this week showing Anthropic's Claude now leads paid business adoption at 34.4%—ahead of OpenAI's ChatGPT at 32.3%. Even more striking: Anthropic is winning roughly 70% of head-to-head enterprise contests for new AI service contracts.
This isn't noise in the data. Anthropic's enterprise adoption has surged 4x since 2025 while OpenAI's growth has plateaued. For CTOs evaluating AI vendors and CFOs tracking AI spend, this shift signals a fundamental change in what enterprises value when they're writing six- and seven-figure AI contracts.
The Ramp AI Index Data: How the Numbers Moved
Ramp tracks corporate card and invoice payments from over 50,000 U.S. businesses, making this a real spend signal rather than survey data or vendor-reported metrics. In April 2026, Anthropic rose 3.8 percentage points to capture 34.4% of enterprise AI adoption, while OpenAI fell 2.9 points to 32.3%.
The year-over-year trend tells a clearer story. Anthropic's adoption has quadrupled since early 2025, driven primarily by Claude Code (their AI-assisted programming tool) expanding beyond technical teams into finance, legal, and research workflows. OpenAI remains dominant in consumer usage and still holds major enterprise deals—Microsoft, Salesforce, and others—but the velocity has shifted.
Overall enterprise AI usage continues climbing, now at 50.6% of Ramp's tracked businesses, up from 42% six months ago. The market is growing fast, but Anthropic is capturing a disproportionate share of that growth.
Why this matters for vendor selection: When enterprises pit Claude against ChatGPT in proof-of-concept trials, Anthropic is winning 70% of those contests according to industry reports. That win rate suggests Claude is delivering something enterprises need that ChatGPT isn't—and it's showing up in contract decisions, not just product demos.
Why Anthropic Is Winning Enterprise Deals
The 70% win rate isn't about model performance benchmarks or API latency. It's about how Claude fits into enterprise workflows where compliance, governance, and cross-functional use cases matter more than raw intelligence scores.
Vertical-specific products: Anthropic launched Claude for Financial Services and Claude for Legal in recent weeks, both designed to integrate directly into Microsoft 365 and other enterprise platforms. Legal teams use Claude to triage incoming matter work, flag contract requests, and surface compliance issues. Finance teams use it for reconciling books, payroll processing, and spotting trends in financial data.
These aren't generic chatbot use cases. They're department-specific workflows with measurable outcomes. A legal team that cuts contract review time from 6 hours to 90 minutes can calculate exact ROI. A finance team that automates monthly close processes can quantify headcount savings.
Claude Code's cross-functional expansion: What started as a developer tool (Claude Code for AI-assisted programming) has become Anthropic's Trojan horse into non-technical departments. Engineering teams adopted Claude Code first, then recommended it to finance for data analysis, to legal for document review, and to research teams for literature synthesis.
This organic expansion pattern—technical teams vouching for Claude's reliability, then advocating for broader deployment—mirrors how Slack and GitHub spread through enterprises a decade ago. The difference is speed: Claude Code went from engineering-only to cross-functional in under 12 months.
Enterprise governance and data controls: Anthropic's Constitutional AI framework gives enterprises more visibility into how Claude makes decisions. For regulated industries (financial services, healthcare, legal), this interpretability matters when auditors ask "how did the AI reach that conclusion?" OpenAI has strong governance features too, but Anthropic has positioned Constitutional AI as a core differentiator for risk-averse buyers.
Customer prompts and responses are not used for model training by default in Claude Enterprise. Data retention, access logs, and audit trails meet stringent compliance requirements. For a Fortune 500 CISO evaluating AI vendors, these aren't nice-to-have features—they're deal requirements.
OpenAI's Response: "Code Red" and the Spud Model
OpenAI didn't sit idle while Anthropic gained ground. In February 2026, OpenAI President Fidji Simo reportedly told staff the company was in "code red" over Anthropic's enterprise surge. That alert triggered several strategic shifts visible in recent announcements.
The Spud model for professional work: OpenAI is developing a new AI model codenamed "Spud," designed specifically for "high-value professional work." According to leaks from OpenAI's roadmap presentations, Spud will offer stronger reasoning capabilities, better understanding of user intent, and more reliable output for tasks like legal analysis, financial modeling, and strategic planning.
This is OpenAI's direct counter to Claude's success in finance and legal verticals. The message: ChatGPT can do high-stakes professional work just as reliably as Claude. Whether enterprises buy that argument depends on execution—and how quickly OpenAI can ship Spud to production.
Mobilizing consulting partners: OpenAI has deepened partnerships with PwC, Bain, BCG, and Accenture to support enterprise implementations. These consulting firms bring credibility and on-the-ground support that pure software vendors can't match. For a Global 2000 CIO planning a company-wide AI rollout, having Bain managing the deployment reduces perceived risk.
Anthropic is countering with its own consulting venture—a joint services company backed by Blackstone, Hellman & Friedman, and Goldman Sachs—aimed at mid-sized enterprises that need hands-on implementation help. Both vendors recognize that selling AI to enterprises isn't just about APIs; it's about change management, workflow redesign, and organizational buy-in.
Compute infrastructure as a moat: OpenAI continues to tout its massive compute advantage—billions invested in NVIDIA GPUs and custom AI infrastructure—as a competitive edge. The argument: when you need to run complex multi-agent systems or process 100,000-page document sets, OpenAI's infrastructure can handle it at scale.
Anthropic's counter: partnerships with AWS (Amazon Bedrock), Google Cloud (Vertex AI), Microsoft Azure ($5 billion investment announced in November 2025), and even a $1.25 billion per month deal with SpaceX for compute capacity through 2029. Anthropic isn't building its own data centers, but it has locked in capacity from every major cloud provider.
What CTOs Need to Know: Vendor Selection in a Two-Horse Race
For technical leaders evaluating AI vendors, the Anthropic-OpenAI competition creates both opportunities and risks. The market is bifurcating into two dominant platforms, with smaller players (Google, Meta, Mistral) competing on price or niche use cases.
Multi-vendor strategies are becoming standard: Most enterprises I talk to are deploying both Claude and ChatGPT, using each where it fits best. Claude for legal and compliance workflows. ChatGPT for customer support and content generation. This hedges vendor lock-in risk and gives teams flexibility to choose the best tool for each use case.
The downside: managing multiple AI vendors increases operational complexity. You need separate governance policies, distinct data access controls, and different integration patterns. For a 50-person platform team, that's manageable. For a 5,000-person organization rolling out AI across 30 departments, it's a scaling nightmare.
Integration depth matters more than model benchmarks: When I ask CIOs why they chose Claude over ChatGPT (or vice versa), the answer is rarely "Claude scored 2 points higher on this benchmark." It's usually "Claude integrates directly into our Microsoft 365 environment and our legal team can use it without leaving Word."
Tight integrations into existing workflows reduce adoption friction. If finance analysts have to switch between Excel, a web browser, and a command-line API to use AI, they won't use it. If AI shows up as a native feature inside Excel, usage jumps 10x.
Expect rapid product iteration: Both vendors are shipping new features weekly. Claude for Financial Services didn't exist six weeks ago. OpenAI's Spud model wasn't on the roadmap three months ago. Any vendor evaluation that takes 6+ months to complete will be evaluating outdated products.
Recommendation: run 30-day pilots with real workflows, not synthetic benchmarks. Give legal, finance, and engineering teams both Claude and ChatGPT. Measure actual task completion rates, user satisfaction, and workflow integration. Make decisions based on how your teams work, not how the vendors demo.
What CFOs Need to Know: Cost, Value, and ROI
For business leaders tracking AI spend, the Anthropic-OpenAI competition is driving down prices while improving capabilities—a rare combination in enterprise software.
Pricing is still token-based, but outcome pricing is coming: Today, both vendors charge per API token (roughly per word processed). A typical enterprise contract might be $50,000-$500,000 per year depending on usage volume. Both vendors are experimenting with outcome-based pricing where they charge for completed tasks rather than tokens consumed.
Example: instead of paying $0.10 per 1,000 tokens, you'd pay $5 per contract reviewed or $20 per financial report generated. This aligns vendor incentives with customer outcomes but requires clearer scoping of what counts as a "completed task."
ROI calculation differs by department: Legal teams see ROI through reduced contract review time (6 hours → 90 minutes = $450 saved per contract at $100/hour billing). Finance teams measure ROI through month-end close acceleration (15 days → 8 days = 7 days of analyst time saved). Engineering teams track ROI through faster feature delivery (30% code completion rate = 12 hours/week saved per developer).
The key: define measurable outcomes before deployment. "AI will make us more efficient" is not an ROI case. "AI will reduce contract review time by 60%, saving $180,000 annually" is.
Watch for hidden costs in multi-vendor deployments: If you deploy both Claude and ChatGPT, you're paying for two platforms, managing two vendor relationships, and training teams on two interfaces. Those coordination costs aren't trivial. For every $100,000 in AI platform spend, budget another $20,000-$40,000 for integration, training, and ongoing management.
What Happens Next: The 2026 Enterprise AI Race
The Ramp AI Index data suggests we're watching a real market leadership transition, not a temporary blip. Anthropic's 4x growth trajectory and 70% win rate in enterprise contests indicate staying power, not just a good quarter.
OpenAI's response—the Spud model, consulting partnerships, and compute investments—shows they're taking the threat seriously. Expect aggressive enterprise-focused product launches from OpenAI over the next 6-9 months.
For enterprises, this competition is good news. Both vendors are innovating faster, pricing more aggressively, and building features that matter for real business workflows. The risk: betting too heavily on one vendor just as the market dynamics shift again.
Actionable next steps for enterprise leaders:
For CTOs: Run 30-day pilots with both Claude and ChatGPT on your three highest-value use cases. Measure task completion rates and user satisfaction. Don't rely on demos or benchmarks.
For CFOs: Require ROI calculations for every AI deployment. If legal can't quantify time savings from Claude, don't deploy it. If engineering can't measure productivity gains from ChatGPT, pause the rollout.
For both: Assume the market will shift again in 12 months. Build vendor-agnostic infrastructure so you can swap platforms without rewriting every integration. The AI model you deploy today probably won't be the leader in 2027.
The enterprise AI market just entered a new phase. Anthropic's 34.4% market share and 70% win rate prove that challengers can displace incumbents—even in a market dominated by a company with OpenAI's brand recognition and Microsoft backing. For enterprise buyers, that means more leverage, better pricing, and faster innovation. Use it.
Continue Reading
Enterprise AI Strategy:
- Agentic AI Scaling: Why 90% of Pilots Fail and How to Fix It
- Claude vs ChatGPT: Enterprise TCO Comparison 2026
- Multi-Vendor AI Strategy: When to Deploy Multiple LLMs
Know someone who'd find this useful?
Forward this to a colleague navigating AI vendor decisions. They can subscribe at beri.net/#newsletter — it's free, twice a week, and I read every reply.
If you were forwarded this, click here to subscribe.
— Rajesh
