Anthropic explores Microsoft's Maia 200 AI chip to reduce Nvidia dependence amid compute struggles

Anthropic Exploring Microsoft's AI Chips to Ease Compute Pressures

Anthropic is in active discussions with Microsoft to adopt the tech giant's Maia 200 AI chip, marking a potential shift in the AI startup's hardware strategy as it grapples with escalating compute capacity demands 1

1

. The talks come as Anthropic CEO Dario Amodei recently acknowledged the company has had "difficulties with compute," driven by surging popularity of the Claude AI assistant and Claude Code programming tool 4

4

. While no deal has been finalized, the discussions signal Anthropic's intent to diversify beyond its heavy reliance on Nvidia hardware, which has traditionally powered its AI model training and deployment 1

1

.

Microsoft-Anthropic Partnership Deepens Beyond Financial Ties

The potential AI chips agreement builds on an already substantial Microsoft-Anthropic partnership. In November, Microsoft invested $5 billion in Anthropic, while the AI startup committed to spending $30 billion on Azure infrastructure 1

1

. Reports indicate Anthropic has been increasing its Azure server rentals, reflecting the company's expanding cloud-based computing needs as AI model development scales 2

2

. Microsoft has also integrated Claude AI models into Office 365 applications including Word, Excel, Outlook, and PowerPoint, with developers finding Anthropic's latest models outperformed OpenAI technology in certain tasks like Excel financial functions and PowerPoint slide generation 2

2

. This represents a notable expansion of Microsoft's AI strategy beyond its long-standing relationship with OpenAI, in which it has invested more than $13 billion 2

2

.

Microsoft Pushes Custom AI Chips to Challenge Cloud Rivals

For Microsoft, a deal to supply Microsoft's custom AI chips to Anthropic would represent a significant win as the company attempts to close the gap with cloud rivals Amazon and Google in the special-purpose AI silicon market 1

1

. Microsoft unveiled its second-generation Maia AI chip in January, but has yet to make it widely available through its Azure cloud platform 3

3

. The discussions highlight Microsoft's broader effort to position its in-house AI hardware as a competitive option within the rapidly growing AI infrastructure market 2

2

. Securing Anthropic as a customer could validate Microsoft's chip development efforts and provide crucial momentum as it competes against Amazon Web Services and Google, both of which already offer their own AI chips to customers 4

4

.

Anthropic Pursues Multi-Cloud Strategy to Reduce Dependence on Nvidia

While the Microsoft talks progress, Anthropic continues to pursue a diversified approach to secure computing resources and reduce dependence on Nvidia processors. The AI startup has signed a 10-year agreement with Amazon Web Services centered on Trainium chips, reportedly worth more than $100 billion, and also plans to utilize Google-developed TPU chips 3

3

4

4

. Anthropic also relies on cloud services from Amazon and Google alongside its Azure infrastructure commitments 1

1

. The company's computing expenses remain substantial, with SpaceX recently revealing that Anthropic will pay $1.25 billion per month through May 2029 for computing capacity 4

4

. This multi-vendor AI strategy reflects the competitive AI landscape, where securing adequate compute capacity has become as critical as algorithmic innovation. Market analyst Jim Cramer called Microsoft's willingness to pay Amazon-backed Anthropic for technology it believes superior "big," highlighting shifting alliances as companies navigate the rapidly evolving AI infrastructure ecosystem 2

2

.