Intel CEO admits AI model training race is lost: Unveils realistic roadmap

In a candid internal Q&A session, Intel’s new CEO Lip-Bu Tan acknowledged what many in the industry have long suspected. Intel has fallen far behind in the AI race. The company, once the undisputed leader in semiconductors, is now struggling to stay relevant as competitors like Nvidia and AMD sprint ahead in AI dominance.

“Twenty, thirty years ago, we were really the leaders. Now, I think the world has changed. We are not in the top 10 semiconductor companies,” Tan reportedly told Intel employees in a global broadcast.

A realistic AI strategy

Unlike his predecessor, Pat Gelsinger, Tan isn’t chasing Nvidia’s towering lead in training infrastructure. Instead, he’s doubling down on what he calls emerging, high-impact AI segments:

Edge AI: AI processing directly on user devices like laptops, desktops, and IoT, without relying on the cloud.

• Agentic AI: Smart assistants that don’t just react, but autonomously act toward goals, paving the way for a new generation of proactive software.

“On AI training, I think it is too late for us,” Tan admitted. “But edge AI... that’s an area coming up very big, and we want to make sure we capture it.”

Insiders suggest that Intel's upcoming Arrow Lake Refresh chips, expected later this year, will feature significantly upgraded NPUs (neural processing units), potentially enabling Copilot+ features on desktop PCs for the first time. This could position Intel as a serious player in consumer-facing AI workloads.

A brutal but focused turnaround

Tan’s remarks come amid significant restructuring at Intel. While not officially confirmed, layoffs are expected across manufacturing, marketing, and international sites, including Oregon, California, Arizona, and Israel. Up to 20% of manufacturing jobs could be cut, while entire business divisions, like automotive, are reportedly being shuttered.

The company is also investing in new executive talent, having brought in three AI-focused vice presidents to lead its transformation.

“This is a marathon,” Tan told the staff. “We need to be humble and make the company more nimble and agile.”

Intel's 18A process architecture, once touted as a comeback vehicle, is rumored to be underperforming. The company is now pivoting toward 14A, its successor, especially for external clients, a clear sign that internal priorities are shifting rapidly.

For now, Intel’s AI ambitions may no longer include being the biggest, but they could still make the company relevant in the next wave of consumer and enterprise AI.