Advanced Micro Devices (AMD) Chief Executive Lisa Su reiterated her support for open-source AI models on Thursday. This comes after a high-profile security incident involving an OpenAI agent earlier this week, which was ultimately resolved using an open-source model, underscoring the relevance of this technology.
"I think open source is a very good thing," Su said during a contentious discussion at the company's "Advancing AI" event in San Francisco. "It provides transparency and control, allowing you to accomplish many things."
Su's comments highlight how the increasing capabilities of AI are creating new points of conflict and challenges across the tech industry and geopolitics. Earlier this week, OpenAI disclosed that two of its AI models had autonomously breached their controlled environment, infiltrating the internal systems of the AI platform Hugging Face. Hugging Face revealed that to control the incident, it used an open-source model from a Chinese company rather than models from leading U.S. AI labs.
Su suggested that banning open source is not a solution. "There's a lot of debate about restricting open-source models, and we believe open-source models have an important place in the ecosystem. We just need to ensure we manage all the relevant aspects."
Earlier this week, Nvidia's CEO also expressed support for open-source AI models, urging the U.S. to stop overreacting to Chinese open-source AI. "These Chinese models are very good," Jensen Huang said. "Quality open-source models should be used."
In her keynote on Thursday, Su stated that the global computing infrastructure supporting AI will, for the first time, shift toward running AI services rather than training AI models. AMD predicts that by 2026, 60% of global AI computing power will be used for inference, which is the process of running pre-trained models. She attributed this shift to the rapid rise of AI agents.
The dominance of inference demand is the core logic behind AMD's launch of new hardware products. Su predicts that GPUs will dominate the AI chip market, but demand for traditional server processors (CPUs) will also see significant growth. AMD's "Venice" CPU is integrated with GPUs in the Helios rack system.
Su expects the total addressable market (TAM) for AMD chips to reach $2 trillion by 2030. She also mentioned that AMD is advancing the deployment of AI computing directly onto edge devices. "We firmly believe that AI should be everywhere," Su stated. To embed intelligence into various work scenarios, AMD has introduced several new processors to support edge computing hardware.
Su also noted that AMD is working "in sync" with partners like OpenAI, Meta, and Anthropic, pointing out that the chipmaker has moved beyond a traditional supplier role to co-develop software and AI platforms. AMD believes this open collaboration model allows it to work with diverse players like Cerebras and integrate various computing technologies.
AMD and its partners "are working more closely together," Su said. "This is a classic example: the more useful AI becomes, the more people want to use it."
Comments