Intel’s hastily installed interim CEO duo revealed that its Falcon Shores GPU architecture will never make it to a data centre near you, as they announced Q4 and full year losses in their first earnings call yesterday.
The news of another AI setback for the ailing CPU giant came as alongside full year figures that showed falls year on year, and Q1 forecasts that came in below market expectations.
Michelle Holthaus and David Zinsner only took up their roles in December, when Intel veteran Pat Gelinsinger retired, amidst reports he had been forced out by the board. The firm, which once dominated Silicon Valley, still has massive weight but is being openly talked of as a takeover target.
Q4 revenues came in at $14.3bn, down 7% on the year, with a $126m net loss, compared to $2.7bn net profit the previous year. Full year revenues were $53.1bn, down from $54.2bn the previous year. That yielded a $18.8bn net loss, compared to $1.7bn net income in 2023.
Restructuring and other charges of $7bn for the full year played a part in that woeful performance – losing 15,000 “friends” doesn’t come cheap. But the company is continuing to struggle with multiple problems, with delayed product launches and ongoing issues with its efforts to build a separate foundry business
It doesn’t help that its critical “datacentre” market is now obsessed with AI, and therefore GPUs. And Intel continues to struggle to catchup on that front.
Holthaus insisted, “The world's data center workloads still primarily run on Intel silicon, and we have a strong ecosystem, especially within enterprise. We are going to leverage these strengths as we work to stabilize our market share in 2025. One of the ways we'll do this is by reengaging the x86 ecosystem.”
She said the firm had to stem a “tide of share loss in the data centre” and would be “fighting for every socket.”
When it comes to the AI data center, "I will start by saying that this is an attractive market for us over time, but I am not happy with where we are today. On the one hand, we have a leading position as the host CPU for AI servers.”
However, hopes of closing the gap around GPUs with its Falcon Shores chip – which combines GPU and x86 in a single package - were shut down.
“Based on industry feedback, we plan to leverage Falcon Shores as an internal test chip only without bringing it to market. This will support our efforts to develop a system-level solution at rack scale with [planned successor] Jaguar Shores to address the AI data center.”
That leaves Intel’s AI accelerator offering centred on Gaudi, the Habana Labs architecture it borged in 2019.
In Q/A, Holthaus added that “one of the things that we've learned from Gaudi is it's not enough to just deliver the silicon. We need to be able to deliver a complete rack scale solution.”
That will come with Jaguar Shores. But it seems unlikely Jaguar Shores will appear until next year. And that is a really extended time frame in the blisteringly fast AI market.
Afterall, earlier this week, Nvidia’s shareprice was just nudging $150. By the end of the week, it was around $124.65, giving it a market cap of $3trillion after the Deepseek inspired minimarketmeltdown.
By comparison, Intel’s market cap is stubbornly bumping around $86bn, just a fraction of what Nvidia lost in a day as the market panicked over the introduction of one new LLM.
