Anthropic is exploring whether to design its own AI chips, according to Tech-Economic Times, as the company and other AI developers respond to a shortage of AI chips needed to power and develop more advanced systems. The exploration is in early stages, and the company has not committed to a specific design or formed a dedicated team, according to the outlet. Anthropic’s spokesperson declined to comment.
Demand for Claude and the compute constraint
Demand for Anthropic’s Claude model accelerated in 2026, with the startup’s run-rate revenue now surpassing $30 billion, up from about $9 billion at the end of 2025, Anthropic said earlier this week. This growth underscores why chip availability is a strategic concern: the company uses a range of chips, including tensor processing units (TPUs) designed by Alphabet’s Google and Amazon’s chips, to develop and run Claude.
Chip availability directly affects training and deployment capacity. A shortage can translate into slower scaling of training runs, constrained inference capacity, or forced prioritization of workloads. The source frames the shortage as affecting both the development and operation of more advanced AI systems, suggesting the bottleneck spans both training and ongoing deployment.
Custom chips remain under consideration
According to three sources cited by Tech-Economic Times, Anthropic may still decide to only purchase AI chips rather than design any. Two people with knowledge of the matter and one person briefed on Anthropic’s plans said the company has yet to commit to a specific design or put together a dedicated team to work on the project.
The distinction between buying and designing chips is technically significant. Purchasing chips keeps a company aligned with vendor roadmaps and manufacturing schedules, while designing chips requires investment in engineering, verification, and manufacturing readiness. If Anthropic proceeds with custom chip design, it would require additional organizational and engineering work before any hardware becomes available.
Recent infrastructure commitments
Earlier this week, Anthropic signed a long-term deal with Google and Broadcom, which helps design TPUs. That deal builds on the company’s commitment to invest $50 billion in strengthening US computing infrastructure. These actions represent concrete steps to address hardware constraints through partnerships and infrastructure investment.
The economics of chip design
Designing an advanced AI chip can cost roughly half a billion dollars, according to industry sources cited by the outlet. This cost reflects the need to employ skilled engineers and ensure the manufacturing process has no defects. The substantial capital requirement highlights why the decision is not simply an engineering question but involves weighing upfront expenses against the option of purchasing chips from existing vendors.
The source does not provide internal cost estimates from Anthropic, nor does it state whether Anthropic’s exploration includes a timeline for prototypes or production. The most defensible reading is that the company is evaluating whether the economics and operational leverage of custom silicon outweigh the uncertainty and capital intensity.
Industry-wide chip design efforts
Anthropic’s discussions mirror similar efforts underway at large tech companies seeking to design their own AI chips. Meta and OpenAI are also pursuing comparable initiatives. This suggests a broader industry pattern: as AI models scale and demand rises, hardware strategy becomes part of competitive positioning, not just a procurement detail.
The source does not claim these companies have reached the same stage as Anthropic, but it does place Anthropic’s exploration within a wider set of responses to chip supply constraints and compute scaling demands.
What comes next
Anthropic’s strategy remains uncertain. The company may decide to design chips, or it may ultimately remain focused on purchasing chips from vendors. That uncertainty is likely to be a key variable for supply planning across the ecosystem, particularly for partners involved in TPU infrastructure.
For AI developers and platform teams, the central takeaway is that compute strategy is becoming a recurring consideration as demand rises and supply remains constrained. Anthropic’s exploration, alongside reports of similar efforts at Meta and OpenAI, suggests that companies may increasingly evaluate whether their next scaling phase requires silicon involvement—or whether partnerships and infrastructure investment are sufficient.
Source: Tech-Economic Times