AMD CEO Lisa Su at CES 2026 outlined the company's AI strategy, noting AI users grew from 1M to 1B post-ChatGPT, a pace faster than the internet's rise. AMD projects 5B active AI users by 2030.

To make AI ubiquitous, global computing power needs a 100x increase in coming years. AMD's ROCm software saw up to 5x AI performance gains last year, with Ryzen/Radeon support doubling and downloads up 10x. The new Adrenalin AI Bundle simplifies local deployment for image gen, LLMs, and PyTorch.

Su emphasized AI infrastructure now relies on CPU, GPU, network, and software synergy, not just single-chip power. AMD showcased the massive "Helios" open dual-wide rack platform developed with Meta.

OpenAI's Greg Brockman highlighted the need for "billions of GPUs" for true AGI, far beyond current capacity. Partners like Luma AI and World Labs demoed video gen, lightweight models, and 3D world creation on AMD.

AMD also presented AI applications in healthcare, robotics, and space exploration, including a computing platform with Blue Origin and accelerated drug discovery.

Su called AI the most important tech in 50 years, committing to democratizing "AI for Everyone."

With compute demand soaring and AMD pushing hardware/software integration, is the industry ready to scale infrastructure for 5B AI users?