AMD Bets Up to $5 Billion on Anthropic in Chip-for-Equity Deal

The News

Advanced Micro Devices said on Wednesday that it will invest up to $5 billion in Anthropic, the maker of the Claude family of models, while helping the lab expand its raw computing power. It is one of the largest single commitments AMD has ever tied to a customer.

Under the partnership, Anthropic will deploy up to 2 gigawatts of AMD's Instinct MI450 GPUs, running on the chipmaker's newly unveiled Helios rack-scale system. The two firms plan to switch on the first gigawatt in the first half of 2027, with the remainder following as capacity is built out. The arrangement was first reported by The Wall Street Journal before AMD confirmed it publicly.

The deal stacks on top of a run of data-centre agreements Anthropic has signed over the past year as it races to secure the silicon needed to train and serve ever larger models. For AMD, landing a marquee model lab as both an investment target and a buyer is a statement of intent in a market Nvidia has dominated.

Why It Matters

The structure matters as much as the sum. A chipmaker is putting equity into the very company it will sell hardware to, so a slice of the $5 billion effectively flows back as GPU orders. That circular quality has become the defining financial signature of this cycle. AMD struck a comparable multi-gigawatt supply pact with OpenAI last year, and Nvidia has poured tens of billions into the labs that consume its chips.

For buyers, the logic is diversification. Nvidia's accelerators remain scarce and expensive, and every serious lab now wants a credible second source. An MI450 deployment at gigawatt scale is the strongest signal yet that AMD's roadmap can carry frontier training workloads, not just inference.

The last time the industry saw commitments of this shape - Microsoft's early, compute-heavy backing of OpenAI - it reset how investors valued the entire stack. When the supplier, the financier and the customer collapse into one relationship, capital stops being the bottleneck and power delivery becomes it. Two gigawatts is roughly the draw of a large Indian city, which is why energy, not money, is fast becoming the real constraint.

Indian Angle

For India, a second heavyweight GPU vendor with real capacity is quietly good news. Home-grown model builders such as Sarvam AI and Ola's Krutrim have spoken openly about compute scarcity and the cost of importing Nvidia hardware in a weak-rupee environment. A viable AMD supply line widens their options and gives procurement teams leverage on price.

It also lands directly on the IndiaAI Mission, MeitY's programme to subsidise access to a common pool of GPUs for start-ups and researchers. Empanelled providers building that pool have leaned heavily on a single vendor's roadmap. A proven MI450-class alternative could ease allocation queues and soften the per-hour compute rates that Indian founders quote in their unit economics.

The deeper lesson is about capital intensity. India's data-centre operators - Reliance, Yotta, CtrlS, Sify - are raising billions to build AI-ready capacity, but none can write a $5 billion cheque to a model lab. As global chipmakers and labs fuse into self-funding blocs, Indian players will increasingly compete not on ideas but on access to power and silicon, the two things the AMD-Anthropic deal was really about.

FAQ

How much is AMD investing and in what form?

AMD has committed up to $5 billion in Anthropic. The investment runs alongside a hardware agreement under which Anthropic will deploy up to 2 gigawatts of AMD Instinct MI450 GPUs, so the money and the chip orders are tightly linked rather than separate transactions.

When does the compute come online?

The companies plan to deploy the first gigawatt of MI450 capacity in the first half of 2027, on AMD's Helios rack-scale system. The remaining capacity, up to the full 2 gigawatts, follows as data-centre build-out continues.

Why does this matter for Indian AI start-ups?

Indian labs such as Sarvam and Krutrim face GPU scarcity and high import costs. A credible AMD alternative to Nvidia at gigawatt scale can widen supply, improve pricing leverage and support programmes like the IndiaAI Mission's shared compute pool.

Where can I read the original announcement?

The deal was first reported by The Wall Street Journal and confirmed by AMD, with detailed coverage published by The Verge, linked below.

This story was reported by The Verge. Read the full original coverage at The Verge.