AMD to Invest Up to $5 Billion in Anthropic in Massive AI Infrastructure Partnership
AMD has committed to invest up to $5 billion in Anthropic as part of a major infrastructure agreement that could involve tens of billions of dollars in artificial intelligence computing systems.
Under the agreement, Anthropic plans to deploy up to two gigawatts of AI computing capacity powered by AMD’s next-generation Instinct MI450-series accelerators. Deployment of the first gigawatt is expected to begin during the first half of 2027.
AMD’s investment will be released based on Anthropic meeting specific infrastructure deployment milestones. However, neither company has publicly disclosed the exact conditions, investment schedule, purchase obligations, or cancellation provisions attached to the agreement.
The deal combines two separate components: a strategic equity investment from AMD and a large-scale hardware commitment from Anthropic.
AMD Links Investment to AI Chip Deployment
The Anthropic partnership reflects AMD’s growing strategy of combining large infrastructure contracts with financial incentives for major AI companies.
AMD previously announced an agreement to provide OpenAI with systems supporting up to six gigawatts of computing capacity. That transaction included warrants that could allow the ChatGPT developer to acquire approximately 10 percent of AMD, provided that deployment, commercial, and share-price targets are achieved.
AMD also reached an agreement with Meta involving up to six gigawatts of graphics processing unit capacity. The Meta deal reportedly included performance-based warrants connected to shipment and purchasing targets.
The structure of the Anthropic agreement is different. AMD is committing to a direct equity investment in Anthropic, while its agreements with OpenAI and Meta provide those companies with the right to acquire AMD shares after meeting certain conditions.
These arrangements highlight the increasingly interconnected relationship between AI developers, chip manufacturers, cloud providers, and data-centre operators.
Anthropic to Deploy AMD Helios Systems
Anthropic will use AMD Helios rack-scale systems equipped with:
- Instinct MI455X graphics processors
- EPYC “Venice” server processors
- Pensando networking technology
- AMD’s ROCm software platform
The deployment expands Anthropic’s existing use of AMD’s earlier MI355X accelerators.
Rather than purchasing only standalone GPUs, Anthropic will receive a complete computing platform covering accelerators, processors, networking equipment, and software.
“We are thrilled to deepen our partnership with Anthropic and deploy AMD Helios at gigawatt scale,” AMD Chair and Chief Executive Officer Lisa Su said.
Su said the collaboration combines Anthropic’s work in frontier artificial intelligence with AMD’s high-performance computing capabilities.
AMD hopes the partnership will establish Helios as a major platform for future AI infrastructure deployments.
Two Gigawatts Could Require Tens of Billions of Dollars
Anthropic’s planned two-gigawatt deployment represents approximately 2,000 megawatts of power capacity.
AMD executives have previously said that developing one gigawatt of AI computing infrastructure can cost tens of billions of dollars, depending on the chips, servers, cooling systems, networking technology, electrical infrastructure, and data-centre facilities involved.
At full scale, the Anthropic deployment would become one of AMD’s largest announced AI infrastructure projects.
AMD has now announced prospective gigawatt-scale commitments involving:
- Up to six gigawatts from OpenAI
- Up to six gigawatts from Meta
- Up to two gigawatts from Anthropic
Some of Anthropic’s AMD systems are expected to be installed in its own data centres. Other systems may be hosted by cloud providers and specialist AI infrastructure companies.
AMD and Anthropic are also looking for operators capable of hosting the planned systems. Su explained that gigawatt-scale infrastructure requires between 12 and 24 months of preparation before deployment can begin.
Anthropic Expands Its Multi-Chip AI Network
Anthropic is not relying on a single chip supplier.
The company’s AI infrastructure network already includes:
- Nvidia GPUs
- Amazon Trainium processors
- Google Tensor Processing Units
- AMD Instinct accelerators
Amazon remains Anthropic’s primary cloud and AI training partner. Anthropic has secured up to five gigawatts of capacity from Amazon and reportedly uses more than one million Trainium2 processors. Trainium3 deployments are also planned during 2026.
Anthropic has also secured next-generation TPU capacity through Google and Broadcom, with deployments expected to begin in 2027.
Claude, Anthropic’s AI assistant, is currently available through Amazon Web Services, Google Cloud, and Microsoft’s Azure-based Foundry platform.
“Running across a diversified range of hardware lets us map the right workloads to the right hardware,” Anthropic co-founder and Chief Compute Officer Tom Brown said.
Brown explained that access to computing resources remains essential to improving Claude and meeting growing customer demand.
“By partnering with AMD across the stack, we are securing the capacity we need and optimising it for training and serving Claude,” he said.
Claude to Help Improve AMD’s AI Platform
The agreement also includes a multi-year engineering programme involving AMD and Anthropic.
The companies will work together to optimise Claude for AMD Instinct accelerators and improve compatibility with AMD’s ROCm software ecosystem.
AMD also plans to deploy Claude across its engineering and product-development teams.
This means Anthropic will not only become a major customer of AMD infrastructure. It will also participate in improving the software environment used to operate AI workloads on AMD hardware.
The arrangement could help AMD address one of its biggest challenges in competing with Nvidia: building a mature and widely adopted software ecosystem for AI developers.
The Global Battle for AI Computing Power
The AMD-Anthropic agreement demonstrates how access to computing infrastructure has become one of the most important factors in the global AI race.
Developing advanced AI models now requires enormous quantities of processing power, electricity, networking equipment, cooling systems, and data-centre capacity.
Instead of depending exclusively on Nvidia, major AI companies are increasingly building diversified computing networks using hardware from AMD, Amazon, Google, and other providers.
For Anthropic, the partnership provides additional infrastructure needed to train and operate Claude.
For AMD, it strengthens the company’s position as a serious alternative to Nvidia in the rapidly expanding market for AI accelerators and integrated data-centre systems.
AMD shares reportedly increased by 2.4 percent following the announcement. The company’s stock has more than doubled since the beginning of the year, outperforming the broader Philadelphia Semiconductor Index.
The partnership signals that the next stage of artificial intelligence competition will not be determined by software alone. It will also depend on who can secure the chips, electricity, data centres, and infrastructure required to operate AI at an unprecedented scale.
