GPU Shortage Slows Anthropic: Cautious Scaling Strategy for Claude
Anthropic is facing a critical shortage of server capacity due to explosive growth in demand for Claude. Enterprise clients with priority access are dealing with outages and strict limits: quotas for Claude Code are exhausted in an hour instead of lasting a full workday. Since late March, the company has introduced restrictions during peak hours to stabilize the service.
CEO's Strategy: Rejecting Risky Expansion
Dario Amodei, CEO of Anthropic, has publicly stated that he does not plan mass GPU purchases. In his view, excess computing resources without matching demand is a straight path to financial losses and bankruptcy. The company prefers to lose customers short-term while preserving margins. To optimize, they're already shifting model training to nighttime hours, freeing up servers for inference during the day.
This decision contrasts with competitors' approaches. Anthropic relies on commercial cloud contracts with AWS, Google Cloud, and Microsoft Azure, but without dedicated infrastructure like OpenAI's. Microsoft has invested tens of billions in data centers for OpenAI, securing an edge in access to premium GPUs.
Infrastructure Comparison and Market Implications
SemiAnalysis analysts point out Anthropic's structural lag. While OpenAI monopolizes top-tier capacity, Anthropic risks being pushed onto less powerful resources. This could affect latency and reliability for enterprise clients.
| Aspect | Anthropic | OpenAI |
|--------|-----------|--------|
| Infrastructure | Cloud contracts | Dedicated data centers (MSFT) |
| Revenue (2026) | $19B | Not specified |
| Response to shortage | Limits + nighttime training | Doubling limits |
Competitors are responding swiftly: after Anthropic announced the restrictions, OpenAI promised to increase quotas, poaching users.
Success Factors and Lessons for the AI Market
The shortage is a side effect of commercial triumph. Anthropic's revenue doubled from $9B (end of 2025) to $19B (March 2026), mainly thanks to Claude Code. User influx surged after the OpenAI-Pentagon contract scandal: ChatGPT downloads fell by 295% in a day, Claude topped the App Store.
- Key growth drivers: Claude Code for developers, enterprise subscriptions.
- Strategy risks: Losing customers at peak demand, lagging behind leaders in infra.
- Benefits of caution: Maintaining profitability, avoiding overcapacity.
- Long-term view: Scaling requires balance between models and hardware.
In the AI industry, winners aren't just those with strong models, but those ensuring stable deployment at production scale.
Key Takeaways
- Anthropic is deliberately limiting access to avoid financial risks from GPU overcapacity.
- Revenue has grown to $19B thanks to Claude Code, but the capacity shortage is driving away enterprise clients.
- Infrastructure lag behind OpenAI (dedicated MSFT data centers) could lead to service quality degradation.
- Competitors are seizing the moment with aggressive marketing of their limits.
- Shifting training to nights is a temporary fix to prioritize inference.
— Editorial Team
No comments yet.