AMD invests $5 billion in Anthropic and deploys 2 GW of GPUs for Claude

What happened?

On July 22, 2026, AMD and Anthropic announced a far-reaching strategic partnership. According to The Next Web, AMD commits to investing up to $5 billion in Anthropic, one of the most prominent artificial intelligence startups and creator of the Claude model. As part of the deal, AMD will deploy up to 2 gigawatts (GW) of its Instinct MI450 Series GPUs in Helios rackscale solutions to run Claude. The first gigawatt deployment will begin in the first half of 2027. Anthropic already uses AMD hardware in its current operations, indicating a prior relationship that is now deepening.

Why is this important?

This alliance represents a direct challenge to NVIDIA's dominance in the AI GPU market, which currently controls over 80% of the segment, according to Mercury Research estimates. The $5 billion investment is one of the largest ever made by a chip manufacturer in an AI startup, surpassing previous deals like Microsoft's $10 billion investment in OpenAI in 2023, though the latter included venture capital and Azure cloud access. The deployment of 2 GW of computing capacity is comparable to the power of several nuclear plants; for example, a typical nuclear plant generates around 1 GW. This underscores the massive scale required by generative AI: training models like GPT-4 alone consumed approximately 50 GWh, and large-scale inference demands even more resources.

For Anthropic, this deal guarantees preferential access to cutting-edge hardware at a time of global GPU shortage, exacerbated by high demand from companies like OpenAI, Google, and Meta. For AMD, it is an opportunity to prove that its Instinct GPUs can compete with NVIDIA's H100/B200 in AI workloads, and to secure a flagship customer that validates its ROCm ecosystem, the equivalent of NVIDIA's CUDA. While ROCm has improved, it still suffers from lower maturity and library support compared to CUDA, which has limited AMD's adoption in AI.

Historical context and comparisons

Historically, AMD has been a secondary player in the AI GPU market, overshadowed by NVIDIA thanks to its CUDA ecosystem and early bet on deep learning. However, the growing demand for large language models (LLMs) like Claude has led companies to seek alternatives to diversify their supply chain and reduce dependence on a single vendor. This AMD move echoes Microsoft's $10 billion investment in OpenAI in 2023, but with a greater focus on infrastructure than venture capital. Microsoft has also developed its own AI chips (Maia), but has not yet deployed them at scale.

Another precedent is the alliance between Google and Anthropic in 2023, where Google invested $2 billion and provided access to its TPUs (Tensor Processing Units). However, the AMD deal is larger in investment and committed computing capacity. Moreover, unlike Google, AMD does not directly compete with Anthropic in the AI model market, reducing conflicts of interest.

In terms of infrastructure, the 2 GW deployment is comparable to the data center expansion plans of hyperscalers like AWS, Azure, and Google Cloud, which are investing tens of billions in new centers. However, most of those centers use NVIDIA GPUs, while this deal bets on AMD.

Consequences for the market and users

• For tech companies: Competition among chip manufacturers intensifies, which could lower GPU prices and accelerate innovation. Other AI startups may seek similar deals with AMD or Intel, which has also launched its Ponte Vecchio and Falcon Shores GPUs. For example, Stability AI has already tested AMD hardware. Additionally, the alliance could pressure NVIDIA to offer more competitive pricing or accelerate the launch of its upcoming Rubin (2026) and Vera (2027) architectures.
• For Claude users: Claude's performance and capacity could significantly improve thanks to greater computing power, enabling training of larger models and reducing inference latency. However, it could also increase Anthropic's dependence on a single hardware vendor, though the company states it will continue using NVIDIA GPUs in parallel. This is crucial to avoid a bottleneck in case of AMD production delays.
• For investors: AMD's stock could benefit if the partnership proves the viability of its GPUs in AI, while NVIDIA faces pressure to maintain its market share. However, the market reacted cautiously: AMD shares rose 2% in the hours after the announcement, while NVIDIA shares fell 1%, according to Bloomberg data. Long-term success will depend on technical execution and AMD's ability to scale production, which has historically been a challenge (e.g., delays in the MI300X series).

What should readers know?

It is important to note that the $5 billion investment is a maximum commitment, not necessarily disbursed immediately. It is structured as a combination of cash and in-kind payments (GPUs), similar to previous AMD deals with other customers. The 2 GW deployment is ambitious and depends on AMD's ability to manufacture and deliver chips on time. AMD uses TSMC's 3nm node, which faces high demand from Apple, Intel, and NVIDIA, so capacity allocation will be critical. Additionally, Anthropic continues to use NVIDIA GPUs in parallel, so there is no full exclusivity. This deal is strategic long-term, and its effects will only be seen in 2027 when the first gigawatt comes online.

Another point to consider is energy consumption: 2 GW of GPUs would consume a massive amount of electricity, posing sustainability challenges. AMD has stated that the MI450 GPUs will be more efficient than previous ones, but has not given concrete figures. Anthropic, for its part, has committed to using renewable energy for its data centers, but it is unclear whether the AMD deal includes that condition.

In summary, the AMD-Anthropic alliance is a milestone that could redefine the balance of power in the AI chip industry, but its success is not guaranteed. It will depend on technical execution, production capacity, and market evolution. Meanwhile, NVIDIA will not stand idly by: it has already announced its Rubin architecture for 2026 and is investing in its own data center network. The GPU war is just beginning.

“The AMD-Anthropic alliance is a milestone that could redefine the balance of power in the AI chip industry,” notes an analyst at TheVortiq. “However, success will depend on technical execution and AMD's ability to scale production.”