Intel's CEO scraps forecast of selling $500M+ worth of Gaudi AI chips in 2024, blaming slower uptake on Gaudi 2 to Gaudi 3 transition and software ease of use
Intel's (INTC.O) upbeat revenue projections on Thursday masked a sore point for the embattled company: chips touted …
Context & Ripple Effects
Intel positioned Gaudi3 as its 2024 entry against Nvidia's H100 and AMD's MI300X in its initial Gaudi3 launch plan. The withdrawn sales target shows that bringing the product to market has not translated cleanly into customer adoption.
The setback lands after Intel's Data Center and AI unit posted modest growth in its first-quarter 2024 results, making Gaudi's commercial ramp an important test of whether Intel can convert AI demand into accelerator revenue.
First-order effects
- Intel no longer expects Gaudi AI-chip sales to clear $500 million in 2024, reducing the near-term contribution expected from its accelerator business.
- Customers evaluating Gaudi face a product-transition and software-usability hurdle, which can delay deployments or limit adoption of the newer Gaudi3 platform.
Second-order effects
- Intel must improve software usability and manage the Gaudi2-to-Gaudi3 handoff more effectively to turn its announced hardware into orders; rivals benefit when alternative accelerators impose less migration risk.
- The miss separates broad AI-data-center demand from demand for any particular accelerator: execution around software and platform transitions can constrain sales even when the market is expanding.
Third-order effects
- If this pattern persists, AI-accelerator competition will be decided less by chip specifications alone and more by software maturity, deployment continuity, and customers' switching costs.
- For Intel, the episode raises the execution bar for establishing Gaudi as a durable alternative in AI infrastructure; a recovery depends on resolving the identified adoption frictions rather than demand alone.
The trend: The AI infrastructure race is shifting from headline hardware launches toward the harder work of software ecosystems and reliable customer migrations.