Anthropic will deploy up to 2 gigawatts of AMD MI450 GPUs to train and run Claude models under a $5 billion investment agreement. AMD is investing capital in Anthropic while securing a major customer for its AI chips.
Anthropic and AMD announced a significant partnership where AMD will invest up to $5 billion in the AI safety company. In return, Anthropic commits to deploying up to 2 gigawatts of MI450 GPUs for training and inference operations powering its Claude language models.
The deal marks another major win for AMD in its effort to challenge Nvidia's dominance in AI chip supply. AMD has secured comparable agreements with Meta and OpenAI in recent months, establishing itself as a credible alternative to Nvidia for large-scale AI infrastructure.
2 gigawatts represents substantial computational capacity. For context, this deployment would support large-scale model training and inference operations required for Claude's development and consumer access.
The MI450 is AMD's latest data center GPU designed for AI workloads. The chips compete directly with Nvidia's H100 and H200 processors that have dominated AI infrastructure buildouts.
However, industry observers have raised questions about the structure of these agreements. Critics point out that AMD's investment into AI companies that immediately commit to purchasing AMD chips creates circular cash flows. The distinction between genuine product adoption and investment-driven purchasing remains unclear.
For Anthropic, the deal provides substantial GPU capacity needed to scale Claude's training and deployment without equivalent capital expenditure. The company has grown rapidly since launching Claude but faces significant computational demands as it expands model capabilities and user access.
AMD's strategy targets the consolidation of AI infrastructure suppliers. With Nvidia facing supply constraints and competition concerns, AMD positions itself as a viable path for companies building large language models and AI services.
The agreement reflects the critical importance of GPU supply in the AI sector. Access to sufficient computational resources has become a key constraint for AI companies scaling their operations and model development.
Broadcom acknowledged it over-emphasized VMware Cloud Foundation, damaging relationships with small and medium businesses. The company is now prioritizing trust and simpler solutions to win back the segment.
Anthropic's $2 trillion initial public offering places its unconventional governance structure under public scrutiny. The AI company's use of external trustees to enforce its mission represents a novel test case for balancing commercial success with stated values.
Google is discontinuing Gmail's "Send as" feature for non-Google email addresses like @yahoo.com and @outlook.com. Users will no longer be able to compose and send messages from external email accounts through Gmail's interface.
Elon Musk has secured a court order preventing others from using the "Twitter" name, though the terms allow continued use of "tweet" and the bird logo. The ruling permits Musk to label X as "formerly Twitter" in app stores.