CES 2026, AMD CEO Lisa Su delivers keynote on AI

Key Concepts

• AI as a Transformative Force: Artificial Intelligence is positioned as the most significant technological advancement in decades, poised to impact all sectors and become integral to daily life.
• Compute as the Primary Bottleneck: Exponential AI growth necessitates a massive increase in compute power – from 100 zettaflops to an estimated 10 yottaflops within five years – to overcome the primary constraint on AI progress.
• AMD’s Central Role: AMD is establishing itself as a key provider of the foundational computing infrastructure (CPUs, GPUs, adaptive computing) for AI across cloud, enterprise, edge, and PC applications.
• On-Device AI & Multimodality: A shift towards bringing AI capabilities directly to devices (AIPCs) is underway, with a focus on efficient “liquid models” capable of multimodal processing (text, vision, audio).
• Open Ecosystem & Democratization: AMD champions an open ecosystem – encompassing hardware, software, and solutions – to accelerate AI innovation and make it accessible to all.
• Public-Private Partnerships for AI Leadership: The US government, private sector, and universities are collaborating through initiatives like the Genesis mission to maintain US leadership in AI through R&D, infrastructure development, and AI diplomacy.
• AI Education & Workforce Development: Significant investment is being directed towards AI education at all levels, from K-12 students to professionals, with AMD contributing substantial resources.

---

AI: The Defining Technology & The Compute Challenge

The presentation frames Artificial Intelligence (AI) as the most important technological advancement of the last 50 years, projecting growth from 1 billion to over 5 billion active users. This exponential growth is driving an unprecedented demand for compute power. Current global capacity stands at around 100 zettaflops, but an estimated 10 yottaflops (10,000x the current capacity – 1 yottaflop = 1 followed by 24 zeros) will be needed within the next five years. This compute demand is the primary bottleneck to AI progress. GDP growth itself will be driven by the amount of compute available in a region, according to OpenAI President Greg Brockman.

AMD’s AI Infrastructure & Solutions

AMD is positioning itself as a central force in addressing this compute challenge, providing CPUs, GPUs, and adaptive computing solutions for AI applications across various environments. The company emphasizes the importance of an open ecosystem – hardware, software, and broader solutions – to accelerate innovation. AMD’s compute portfolio is tiered: Helios (bleeding-edge performance), MI440X (enterprise), and MI430X (sovereign AI/scientific computing).

The Helios Rack Scale Platform was unveiled, designed for AI infrastructure at “yacht scale.” It’s a double-wide rack featuring liquid cooling and integrates MI455 GPUs, Epic CPUs, and Pensando networking chips, delivering 2.9 exaflops of performance with approximately 18,000 CDNA 5 GPU compute units and 4,600 Zen 6 CPU cores, weighing around 7,000 pounds.

Advancing Generative AI: On-Device & Multimodal

AMD is developing “liquid models” (LFM) designed for efficient on-device AI processing, reducing computational costs without sacrificing quality. LFM 2.5, a new generation of tiny class models (1.2 billion parameters), offers instruction-following capabilities comparable to larger models like DeepS and Gemini 2.5 Pro. Five instances are available: chat, instruct, Japanese-enhanced, vision-language, and lightweight audio, optimized for AMD Ryzen AI CPUs, GPUs, and NPUs and available on Hugging Face and Leap. LFM 3, a natively multimodal model capable of processing text, vision, and audio, is expected later in the year.

The Ryzen AI Max processor for AIPCs features 16 Zen 5 CPU cores, 40 RDNA 3.5 GPU compute units, and an XDNA 2 NPU delivering up to 50 TOPS of AI performance, with a unified memory architecture supporting up to 128GB. It outperforms the MacBook Pro in AI and content creation and offers competitive pricing compared to NVIDIA’s DGX Spark (1.7x more tokens/second/dollar with GPT OSS models). The Ryzen AI Halo reference platform, launching in Q2, can run models with up to 200 billion parameters.

Spatial Intelligence & Scientific Discovery

World Labs, led by Dr. Fay Lee, is focusing on spatial intelligence in AI, enabling machines to create and understand 3D/4D worlds. Their Marble model generates navigable 3D worlds from minimal input (images), powered by AMD MI325X chips and Rockom.

AI is accelerating scientific discovery through initiatives like the Genesis Mission, integrating national labs, supercomputers, and private sector innovation. Examples include the Lumi supercomputer reducing climate model update times by 85%, ENI utilizing AMD-powered supercomputers for battery and fuel development, and El Capitan modeling virus mutations for antibody treatments. AMD technology is also being used in Blue Origin’s lunar landers for autonomous operation and edge AI.

Government Initiatives & AI Education

The US government is investing heavily in AI through the Genesis mission, prioritizing removing regulatory barriers, bolstering AI infrastructure and energy production, and pursuing AI diplomacy through the American AI Export Program. A key component is AI education, with the Presidential AI Challenge engaging over 5,000 students and 1,000 educators. AMD has committed $150 million to expanding AI education access, collaborating with over 800 educational institutions and offering free online courses to over 150,000 students.

The Hack Club nationwide AI and robotics campaign culminated in a hackathon where Team Armender (Emmy McDonald, Rosanna Gaboan, and Afia Ava) won with their AI robot barista, utilizing a unified vision language model trained on AMD developer cloud with MI300X GPUs and running on an AMD Ryzen AI laptop. Each team member received a $20,000 educational grant.

Conclusion

The presentation underscores the transformative potential of AI, contingent upon overcoming the critical challenge of compute capacity. AMD is positioning itself as a central enabler of this transformation through its comprehensive portfolio of AI infrastructure, its commitment to on-device AI and multimodal capabilities, and its collaborative efforts with government, academia, and industry. The emphasis on an open ecosystem, coupled with significant investment in AI education, signals a commitment to democratizing access to AI and ensuring US leadership in this rapidly evolving field. The overarching message is that AI is not just a technological advancement, but a fundamental force that will reshape the future, and is poised to be “everywhere for everyone.”