4 hours ago
602
Intel is officially stepping into the GPU war, and it’s not half-assing it. CEO Lip-Bu Tan said Tuesday that the company has hired a new chief architect to start building graphics processing units.
He didn’t say the name, but admitted it took “some persuasion” to get the person to join. Nvidia and AMD already dominate the space, and their chips power everything from large language models to the biggest AI data centers out there.
The demand for GPUs has exploded as more companies rush to build AI infrastructure. Intel wants a piece of that money pile, and it’s finally doing something about it.
Intel’s foundry struggles, Wall Street targets, and memory chip deal
But the timing is rough. Intel has had a bumpy few years. The company fell behind in the AI chip race while others soared.
Even though its latest quarterly results beat expectations, investors were focused on other issues, like manufacturing delays and the lack of a key foundry customer.
Intel’s foundry division is supposed to make chips for outside clients, but right now it mostly makes its own. That’s not what Wall Street wanted to hear.
Last year, the U.S. government, SoftBank, and even Nvidia threw money at Intel, betting on its recovery. And there’s some progress. Daiwa Capital Markets just raised its price target from $41 to $50. MarketBeat says the average target sits at $45.76, though analysts overall still say “Reduce.”
Meanwhile, Nasdaq shared data from Fintel showing the average one-year target at $46.77, up 22.1% from the Jan. 11 estimate. But not everyone’s convinced. Stacy A. Rasgon at Bernstein kept his neutral rating and a lower $36 target.
Intel is also jumping into new memory tech. On Feb. 2, Tokyo-based SAIMEMORY, a SoftBank unit, said it signed a deal with Intel to develop “Z-Angle Memory.” It’s a new kind of chip that’s supposed to work better for AI inference, the part where models actually run in production.
These chips will need to move a huge amount of data quickly, use less power, and have higher capacity. The plan is to start prototyping by March 2028, and maybe sell them commercially by fiscal 2029.
Meanwhile, Lip-Bu didn’t ignore the bigger problem: memory chip shortages. He told the Cisco AI Summit the demand from AI data centers has made things worse.
There’s not enough supply to go around, and that’s let memory makers keep jacking up prices. Lip-Bu called AI the “biggest challenge” for memory and said he expects “no relief until 2028.”
Intel wants to compete with Taiwan Semiconductor Manufacturing Co., which already builds chips for most of the world’s biggest names. But right now, Nvidia’s GPUs are still the go-to choice for AI. AMD isn’t far behind. With this new hire and deals like SAIMEMORY, Intel is finally taking big swings. Now it has to prove they won’t miss.
If you're reading this, you’re already ahead. Stay there with our newsletter.
English (US)