Demand for Claude surged 80x: why Anthropic is seeking supercomputers

Demand for Anthropic services in the first quarter of 2025 grew 80-fold. This forced the company to urgently seek additional computing capacity, including considering offers from major high-performance systems vendors.

How Claude Demand Is Growing

Recent months have shown that the AI services market is far from saturation. Claude is rapidly gaining trust in the corporate sector — from finance and consulting to software development and legal analysis. If at the beginning of the year Anthropic could handle orders with existing computing resources, by the end of the quarter GPU consumption had grown 80 times. This is not a forecast — it is a fact acknowledged by the company's leadership.

Several factors have driven this growth. Claude demonstrates strong results on specialized tasks, the company has attracted major corporate clients, including the public sector. The cost of service through Anthropic's API proved competitive with other solutions. Together, this created a wave of demand that no one expected at this scale.

Anthropic's Strategy Fell Behind Reality

Unlike OpenAI, which aggressively purchased GPUs and secured capacity contracts in its early stages, Anthropic chose a conservative path. The company invested gradually, based on current demand. From a financial perspective this was prudent, but in the face of an unexpected boom, it created a critical bottleneck. When demand grew 80-fold, it turned out that existing capacity was insufficient. The company cannot quickly launch new servers — this requires time to purchase chips, deploy infrastructure, and negotiate with providers. The GPU shortage is acute: modern accelerators (NVIDIA H100, H200) are in short supply globally.

Seeking Solutions

Anthropus is actively negotiating with high-performance systems vendors. Several strategies are on the table:

• Long-term contracts with cloud providers (AWS, Google Cloud) — fast, but expensive
• Purchasing supercomputer systems for own data centers — requires capital investment, but more economical long-term
• Leasing excess capacity from companies that have already invested in GPUs
• Establishing partnerships for joint infrastructure deployment

Each path involves tradeoffs. Cloud scales quickly but creates dependence on a single provider. Own infrastructure takes time but provides independence. Leasing depends on market availability.

What This Means

Anthropic's story demonstrates real demand that has swept through the AI market. The company bet on quality and caution in investments — this won customer trust, but now requires urgent scaling. Computing capacity has become a real strategic asset, not a formality. An 80-fold surge in demand speaks not of a bubble, but of a healthy market ready to grow further.