Global AI computing demand is severely short of supply: AMD's Lisa Su predicts a 100-fold increase needed

Ahead of the upcoming CES event, AMD CEO Lisa Su publicly stated that current computing power is still insufficient to meet the demands of AI development. She shared that the industry has only just begun the AI revolution, despite impressive progress in recent years.

Long-term vision for computing needs

Lisa Su pointed out a major challenge: to achieve feasible AI goals, global computing capacity must increase dramatically. Specifically, over the next 5 years, computing scale will need to grow 100 times, from the current Zetta level to 10 YottaFLOPS. This represents a computing explosion compared to 2022, when the scale was only 10,000 times smaller.

Product strategy and technological breakthroughs

To address this demand, AMD announced a series of new AI processing chips. The MI450 chip line and Helios desktop are introduced as significant steps forward. Furthermore, Lisa Su revealed that the MI500 processor line will debut in 2027, with performance increasing 1,000 times compared to the MI300 launched in 2023.

Key factors: Architecture and manufacturing process

This breakthrough depends not only on system architecture design but also on innovations in manufacturing technology. This is an important insight from Lisa Su, indicating that AMD understands AI will gradually become the foundational infrastructure for the entire digital economy in the future.