Gelsinger: AI compute efficiency must jump 10,000x amid power and memory limits

Pat Gelsinger said at a Playground Global event in Taiwan that AI compute efficiency must improve 10,000-fold to overcome constraints in power, memory, and fab capacity. The VC firm brought portfolio companies working on memory, power management, optical communications, and advanced packaging to showcase joint breakthroughs.

• Gelsinger set a 10,000x target for AI compute efficiency gains
• Constraints cited: power consumption, memory, and fab capacity
• Playground Global brought memory, power, and packaging firms to Taiwan