NewsNTech
The constraint that makes AI chip roadmaps matter is compute efficiency at scale.
How much useful work a piece of silicon delivers against its power draw and cost in a production data center, repeated across millions of inference requests, determines the economics of running large AI workloads.
Hyperscalers that develop custom hardware rather than buy merchant GPUs are betting that ownership of that efficiency curve compounds over years.
Alphabet's new AI chip roadmap surfaced as a significant technology development on Monday, covered by CNBC's Investing Club in its Homestretch segment alongside two separate factors the Club identified as capping the session.
Keep reading