An area where Intel (INTC -3.67%) seriously dropped the ball was artificial intelligence (AI). Nvidia essentially owns the AI accelerator chip market, and AMD it’s at least competitive with its Instinct line of data center graphics processing units (GPUs). Nvidia generates tens of billions of dollars in revenue every quarter from its data center GPUs, and AMD makes a few billion a year. Meanwhile, Intel did not reach its goal of selling $500 million of AI accelerators by 2024.
The boat is missing
It shouldn’t be like this, but mistakes made long ago have hindered Intel’s efforts in AI. The company was working on a discrete GPU project, codenamed Larabee, in 2008, and its architecture would have been well suited for the massively parallel computing tasks needed to train AI models. Larabee was overruled, however, setting the stage for Intel to be blindsided by the AI revolution and increased demand for GPU accelerators in recent years.
Intel’s current line of AI accelerators comes from Habana Labs, an AI chip company Intel acquired in 2019. The Gaudi family of AI chips are not traditional GPUs, although they are similar in nature.
The latest Gaudi 3 chip offers solid performance, and Intel has been aggressive on pricing. Gaudi could be a success story for Intel, but an immature software ecosystem is hurting sales. Developers who have been working with GPUs for years have no experience with Gaudi’s distinct architecture, something that has been impossible for Intel to overcome.
Software immaturity is starting to become a theme for Intel in the graphics and AI areas. In 2023, the company tried again in the discrete GPU market with its Arc Alchemist gaming graphics cards, but they failed to sell well due to faulty software drivers and a myriad of game-breaking bugs . The company has kept up with it, greatly improving its software over time, and its second-generation graphics cards are getting much better.
Don’t expect great things in 2025
If Intel’s experience in the graphics card business is any indication, it will take time for the AI software ecosystem to improve enough for Gaudi to be a major success. Another wrinkle is that Intel’s AI accelerator path is complicated. Under the Max brand, Intel already sells data center GPUs, which power the Aurora supercomputer at Argonne National Laboratory.
Neither the Max family of GPUs nor the Gaudi family of AI accelerators will be expanded. Rather, the next generation Falcon Shores will be a traditional GPU that will integrate some of Gaudi’s unique features.
Falcon Shores is expected to launch in late 2025, although there are no guarantees given Intel’s current turmoil. Gaudi 3 sales should be up this year compared to last year, but given the software issues, sales are unlikely to be anywhere near what AMD is raking in with its data center GPUs .
Intel has had some successes, although they have been few and far between. A major win was a deal IBM to put Gaudi 3 chips into IBM’s cloud data centers and integrate them into IBM’s watsonx AI platform. That should help boost Gaudi’s sales this year, although the size of the deal was not disclosed.
A greater opportunity of AI
Intel may eventually find success selling data center GPUs that double as AI accelerators, but it probably won’t happen this year. In the long run, manufacturing AI accelerators for other companies may be a more lucrative endeavor.
Intel is betting big on leveraging its decades of manufacturing expertise and tens of billions of dollars in manufacturing investment to build a viable foundry business. The company is on the cusp of completing its initial process path, with the Intel 18A process scheduled to go into volume production next year. Intel will make some of its central processing units (CPUs) on Intel 18A, and has already won a few high-profile customers.
Those customers include Microsoftwhich will use Intel 18A to produce an undisclosed chip, and Amazonthat’s up to Intel 18A for an AI fabric chip. If Intel can prove to potential customers that Intel 18A is the real deal by successfully making its own chips and making chips as part of these initial partnerships, more AI-related chips could be in the works for Intel foundries in 2026 and beyond.
Long story short, Intel’s AI accelerator revenue probably won’t be as impressive in 2025 as the company works through software issues and solidifies its roadmap. However, looking further ahead, the foundry business may ultimately be a bigger opportunity for Intel to raise some AI spending.