AMD Teams Up With Meta, OpenAI and Anthropic to Optimize AI Models, CEO Says the Whole Ecosystem Benefits

AMD CEO Lisa Su said on the company's second-quarter earnings call that AMD is working directly with leading AI companies Meta, OpenAI, Anthropic, and others to optimize their AI models on ROCm, the company's open AI software platform, with the improvements benefiting the entire AMD ecosystem. According to Benzinga, the work extends beyond individual customers and improves software performance for all developers deploying AI workloads on AMD hardware.

Su said the collaborations are making it significantly faster and easier for developers to bring AI models into production at scale. AMD added that more than 3 million AI models now run out of the box on its platform, open-source contributions to ROCm have increased more than tenfold over the past year, and new open-source AI models are supported on its AI chips as soon as they are released, so developers can deploy the latest models without waiting for software updates.

The disclosure came alongside a Q2 earnings beat. AMD posted second-quarter revenue of $11.54 billion, topping analyst estimates of $11.28 billion, with adjusted earnings of $1.66 per share beating expectations of $1.61 per share. The company guided third-quarter revenue to approximately $13 billion, plus or minus $300 million, above Wall Street's $12.52 billion forecast.

In terms of price action, AMD closed 7% higher on Tuesday at $518.58 and fell 8.30% in early pre-market trading on Wednesday. Benzinga's Edge rankings put AMD's momentum score in the 98th percentile and its growth score in the 97th percentile.

The deepening ties with leading model makers arrive as the open-source AI ecosystem heats up. In July, Nvidia CEO Jensen Huang said the world needs both frontier closed models and frontier open models, endorsing an industry letter backing open models as a driver of innovation and AI sovereignty. AMD's ROCm collaborations with Meta, OpenAI, and Anthropic are widely read as its latest push against Nvidia's CUDA software moat.

For developers, the practical signal is that the barrier to deploying AI models on AMD hardware is falling, and the cycle from model optimization to production is shrinking. The questions to watch are whether ROCm improvements translate into sustained data-center order growth for AMD, and whether more open-source models begin shipping with default AMD support.