Anthropic Aims for 5GW AI Compute Capacity by 2026, Doubling to 10GW by 2027

Anthropic, a prominent AI company known for its Claude AI models, has announced its ambitious plan to secure 5 gigawatts of AI compute capacity by the end of 2026. This initiative, reported by the New York Times, is part of a broader infrastructure development involving major tech firms such as Amazon, Google, and Broadcom.

This expansion aligns with Anthropic’s strategy to bolster its AI capabilities amid increasing competition in the sector. Market participants are interpreting this move as a positive indicator of Anthropic’s potential to deliver top-performing AI models by late 2026.

Future Expansion Plans

In addition to the initial target, Anthropic has reportedly informed investors of its intention to double its AI compute capacity to approximately 10 gigawatts by 2027. This ambitious goal reflects the company’s commitment to scaling its infrastructure to support advanced AI model training and deployment.

Key takeaways from this development include:

• Anthropic’s planned expansion is seen as a strategic move to enhance its competitive position in AI development.
• Market pricing indicates increased confidence in Anthropic’s ability to produce a leading AI model by October 2026.
• Partnerships with major tech companies suggest a significant commitment to scaling AI capabilities.

Market Observations

Market observers will be closely monitoring Anthropic’s progress in rolling out its planned compute capacity and any further partnerships with industry giants. Developments in AI model performance benchmarks and competitive releases from other major players, such as Google and Meta, will also influence market perceptions of Anthropic’s position in the AI landscape.

New Developments in Anthropic's AI Compute Capacity

• Anthropic aims to achieve approximately 10 gigawatts of compute capacity by the end of 2027, a significant increase from the 1.4 gigawatts available at the close of 2025.
• The company has secured a partnership with Google and Broadcom for about 3.5 GW of next-generation TPU capacity, expected to roll out starting in 2027.
• An agreement with AWS includes provisions for up to 5 GW of new Trainium capacity, Amazon's custom AI training chip.
• A collaboration with AMD focuses on MI450 GPUs, contributing up to 2 GW, with an initial 1 GW expected to be operational by the first half of 2027.
• Anthropic has signed a lease for a 2.16 GW data center campus in Australia, designed for inference workloads and set to launch in 2027.
• The company has also secured a deal with Nscale for 460 megawatts of capacity, reportedly worth $45 billion over six years, with operations starting in late 2027.
• By the end of 2026, Anthropic's compute footprint is projected to reach around 5 GW, with a target of 10 GW by the end of 2027.
• As of April 2026, Anthropic's annualized revenue run-rate exceeded $30 billion, a substantial increase from $9 billion at the end of 2025.
• The company serves over 1,000 customers, each spending more than $1 million annually on its Claude AI models, with projections suggesting annualized revenue could surpass $65 billion.
• Anthropic is diversifying its compute sources by utilizing Google TPUs, AWS Trainium, and AMD's MI450 to optimize for data sovereignty and latency across different markets.

New Insights from Anthropic CEO Dario Amodei

On September 12, 2023, Dario Amodei published an essay titled "We Must Pace the Frontier," emphasizing the urgent need for the AI industry to slow down its development of frontier models due to safety concerns.

Amodei argues that the rapid advancement of AI capabilities has outpaced the necessary safety infrastructure, warning that unchecked AI progress could lead to catastrophic outcomes, including potential damages in the hundreds of billions of dollars.

His essay outlines a three-part plan for managing AI risks: embedding independent evaluators within AI organizations, establishing industry-wide safety standards among democratic nations, and fostering international cooperation on AI governance.

Notable figures in the tech industry, including Sam Altman, Elon Musk, and Demis Hassabis, have expressed support for Amodei's call for increased safety measures and interpretability in AI systems.

Amodei emphasizes that without transparency in AI decision-making processes, safety evaluations remain largely speculative, highlighting the critical need for interpretability research.