AMD CEO Lisa Su: AI Reaches Inflection Point for Meaningful Work
In brief
- AMD CEO Lisa Su says AI reached inflection point for meaningful work in Yahoo Finance interview
- AMD-Anthropic partnership includes up to 2 gigawatts of Instinct MI450 GPU deployment
- AMD investing $5 billion equity stake in Anthropic as part of strategic deal
- First gigawatt of GPU deployment expected operational by first half of 2027
- Global inference compute projected to surpass training compute in 2026
The Anthropic Partnership
AMD announced a partnership with Anthropic involving the deployment of up to 2 gigawatts of AMD's latest Instinct MI450 GPUs, integrated into Helios rack systems. The first gigawatt of that deployment is expected to be operational by the first half of 2027. AMD is also making an equity investment of up to $5 billion in Anthropic.
This move completes a trifecta of AI relationships for the chipmaker. AMD announced a partnership with OpenAI in October 2025, involving 6 gigawatts of GPU capacity. It followed that with a Meta partnership in February 2026. AMD has reportedly committed around 20% of its equity across these AI partnerships.
Training vs. Inference: The Shift Underway
Su's inflection-point thesis hinges on a fundamental shift in AI compute workloads. Training is when you build an AI model, which requires enormous, concentrated bursts of GPU power running for months. Inference is when you actually use that model, every time someone asks ChatGPT a question, every time an AI agent processes a document.
2026 is projected to be the first year that global inference compute surpasses training compute. That crossover signals sustained, distributed demand for GPU capacity—not just episodic spikes when a new model launches.
The Ecosystem Becomes Collaborative
Su described a shift in how the AI industry operates. The broader AI ecosystem is increasingly intertwined, with hardware providers, model developers, and cloud operators working more collaboratively than competitively. These partnerships with OpenAI, Meta, and Anthropic underscore that reality. Each deal locks in long-term GPU commitments and aligns incentives across the stack.
Su doesn't traffic in hype. Years of methodical GPU business building have earned her credibility when she signals a turning point. The Anthropic announcement, paired with the inference-compute inflection, suggests the market is moving from proof-of-concept to production deployment at scale.
