Former Intel CEO Pat Gelsinger criticized current AI hardware, particularly GPUs, for their power inefficiency and computational limitations. He argued that future AI advancements will depend more on improving overall infrastructure, including compute, memory, networking, energy, and economics, rather than solely on deploying more accelerators. OpenAI's head of compute, Sachin Katti, echoed these concerns, highlighting bottlenecks across the entire supply chain, from data center construction and energy availability to chip manufacturing and system integration. AI
IMPACT Highlights that future AI progress may be constrained by infrastructure and efficiency, shifting focus from raw compute power to broader system-level improvements.
RANK_REASON The cluster consists of commentary from a former CEO and an industry participant about AI hardware limitations and infrastructure needs, rather than a new product release or research finding.
Read on Data Center Knowledge →
- graphics processing unit
- High Bandwidth Memory
- Intel
- OpenAI
- Pat Gelsinger
- Playground Global
- Sachin Katti
- The New Yorker
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
