Uber burned through its entire 2026 AI budget in just four months. The culprit: Claude Code. Engineers were racking up $500 to $2,000 per month in API costs alone.
The company behind that Claude — Anthropic — just surpassed OpenAI in enterprise AI adoption for the first time. The data comes from Ramp, a fintech firm that tracks real payment data from more than 50,000 U.S. businesses.
What's behind the numbers?
As of April 2026, 34.4% of U.S. businesses are paying for Anthropic — surpassing OpenAI (32.3%) for the first time. Ramp's survey isn't based on self-reported answers. It counts companies that actually wrote a check for AI, making it more reliable than most industry surveys.
The speed of this shift is striking. Anthropic's enterprise adoption went from 0.03% in June 2023 to 34.44% in April 2026 — in just three years. Over the past year alone, Anthropic quadrupled its business adoption while OpenAI grew just 0.3%.
OpenAI peaked around 36.5% in mid-2025 and has been declining since. On OpenRouter's developer-community leaderboard, Anthropic overtook OpenAI back in December 2025.
The engine driving this shift is Claude Code. Recent analysis shows 4% of all public GitHub commits worldwide are being authored by Claude Code — double the figure from just one month prior. Anthropic's strategy was intentional: win early-adopter engineers and technical teams first, then let them spread it across their organizations. Ramp economist Ara Kharazian described it as: "start with a very technical customer base, focus on their needs, succeed in execution, and then start broadening out".
What does this mean for your team?
"Anthropic is #1 now — should we switch?" That's the wrong question. Ramp's chief economist Kharazian highlighted three headwinds facing Anthropic even as it takes the lead. And these risks matter to OpenAI users just as much.
| Cost-unaware adoption | Cost-optimized adoption | |
|---|---|---|
| Model selection | Always the latest, most expensive model | Route by task complexity |
| Usage tracking | Left to individual discretion | Team-level dashboard monitoring |
| Budget planning | Annual fixed budget | Monthly usage pattern tracking |
| Result | Like Uber: 1 year of budget in 4 months | Maximum ROI within budget |
The Uber case is the clearest warning. Their CTO revealed the company spent its entire 2026 AI budget in just four months, mostly on Claude Code and Cursor. Engineers incurred $500–$2,000 per month in API costs. Claude adoption jumped from 32% to 84% of Uber engineers in months. The faster your company-wide AI rollout, the faster the budget shock arrives.
Three structural risks Kharazian identified:
- The token economics trap
Anthropic makes more money when businesses buy more tokens — incentivizing them to push users toward expensive models even when cheaper ones suffice. Audit whether your team is running simple tasks on premium models unnecessarily. - Quality and reliability risk
Recent weeks saw Claude outages, rate limits, and rising user dissatisfaction. Anthropic responded with a new compute deal with SpaceX, but for mission-critical workloads, a multi-model strategy remains the safe play. - 3x token cost for image prompts
A recent Anthropic model update tripled token costs for prompts that include images. Teams doing heavy image analysis should monitor spend closely right now.
The key insight
The real story isn't that Anthropic passed OpenAI. It's that the product that made it happen — Claude Code — is already blowing up enterprise AI budgets. Unchecked rollout can cost $500–$2,000 per engineer per month.
The essentials: how to manage AI costs before they manage you
- Set model routing rules
Not every task needs the latest Sonnet. Build a simple guide: "Does this require complex reasoning?" Yes → premium model. No → Haiku or a cost-efficient alternative. This alone can cut 60–80% of AI spend. - Make usage visible at the team level
Personal accounts for Cursor or Claude Code make total spend invisible. Tools like Langfuse or Helicone let you centralize API monitoring — seeing who's using what and how much, at a glance. - Implement prompt caching
If your team reuses the same system prompts or long context windows, Anthropic's Prompt Caching can cut costs by up to 90%. Essential for teams making direct API calls. - Set monthly AI spend limits
To avoid the Uber scenario, set per-person and per-team monthly caps. Treat AI API budgets like AWS or Azure spend — with hard thresholds and alerts before they blow up. - Quarterly ROI reviews
Is AI actually improving productivity — or just generating spend? Build a quarterly review to measure output per dollar. Deployment matters, but so does measurement.
Go deeper
Ramp AI Index Monthly measurement of business AI adoption in the U.S., based on real payment data from 50,000+ companies. The most reliable B2B AI trend tracker available. ramp.com
Ramp Economics Lab — Anthropic beats OpenAI Kharazian's original May AI Index analysis with detailed breakdown of the three headwinds. econlab.substack.com
VentureBeat deep dive Detailed analysis of Anthropic's rise, Claude Code growth data, and the full Uber budget case study. venturebeat.com
Anthropic Prompt Caching docs Official docs on reducing API costs with caching. Essential reading for teams making frequent API calls with repeated context. docs.anthropic.com
OpenRouter Rankings Developer-community AI model leaderboard — useful cross-reference for Ramp data trends. openrouter.ai
