Why Intel won't abandon its GPUs

There are rumors… And there are realities. For about two years and the hectic launch of its first dedicated graphics cards “Alchemist” from Intel, a background noise has continued to rustle: Intel will soon close its division in charge of GPU development. However, by dragging his gaiters CES in Las Vegasit was easy to conclude the opposite. Note that Intel has never accelerated so much in this area.

To be convinced, it was enough to see the first “Battlemage” graphics cards, the B580 and B570. Entry-level gaming GPUs launched in December 2024 and which had a good response. Intel doesn't yet have what it takes to really shake up Nvidia and AMD in the mid- and high-end range, but despite some limitations, the second generation of Intel's gaming GPUs is proving more attractive than the previous generation. A good point.

But what directly called things to order was less the gaming chip than the graphic announcement from CES: a graphics chip for… Cars.

Arc for the automobile is coming

On the sidelines of its Core Ultra 200H and Ultra 200 HXIntel mainly took advantage of CES to make automotive announcements. The first is the famous ACU U310 chip which we have already told you about. A processor for controlling the energy of electric car motors that certain manufacturers such as Stellantis are starting to integrate into vehicles.

But the other announcement that concerns us here is the launch of the first generation of Intel graphics cards for automobiles. Called Intel ARC B for automotive – the B being for “Battlemage”, the second generation of ARC chips which is at the heart of the Core Ultra 200S and the B580 and B570 graphics cards – this chip is planned for the end of 2025.

Intel is offering a solution here that will support the car's internal displays, from GPS to entertainment applications, from films to video games – we speak of an infotainment system in the jargon. And quite logically, this chip will also be able to run AI, such as voice recognition and other small local LLMs.

The GPU, a technology in several forms

You have to be careful what you say. If in the minds of ordinary people, a GPU is a physical card integrating a processor in charge of displaying and calculating video games, the reality is much more complex. In truth, a GPU is an architecture (a logical block designated in the form of intellectual property called IP in the jargon) which can take several forms: a dedicated chip, of course. But also a piece of chip in a computer processor (an SoC in reality). But it can therefore also be a chip or a logical element integrated into a car, a digital cash register or any other device. And if you want to confuse the issue even further, intensive computing chips and data centers are always called GPUs (graphic processing units) even though they are most of the time devoid of graphics computing extensions!

So could Intel stop its dedicated graphics cards in the near future? Possible, but not too likely. The giant took a long time to deliver a successful IP, so it would be a waste of money. Then, its second generation, far from being perfect, offers a better quality/price ratio than the first iteration and begins to make a name for itself. Finally, this architecture is integrated into its mobile and desktop chips. But it is also available in professional cards, whether office or data center. That's for the IT part.

The growing auto chip market

As we have seen, Intel is capitalizing on its development by using part of the graphics logic bricks for industrial chips, intended for the automobile. However, if the PC market is progressing little and remains very competitive, the auto chip market is booming. According to the latest report from TechInsightsdriven by electrification and digitalization, the automotive semiconductor market could be worth $117 billion in 2023.

Even if Intel only manages to capture 8% of this market, that still represents a jackpot of almost $10 billion. With an estimated $52 billion in revenue for 2024 (only the first three quarters have been released, averaging $13 billion per quarter), an extra $10 billion would be a boon for Intel. Better yet: pushed around on CPUs both in PCs and in calculators, Intel is also and above all almost absent from these markets in terms of GPUs. The giant can only gain market share by positioning itself, as it does, as a third way against an ultra-dominant Nvidia and a struggling AMD.

Will Intel's GPU strategy succeed? Hard to say. But one thing is certain: Intel can completely abandon its gaming graphics cards in the event of a total fiasco over the next two generations. But given the importance of GPUs in many markets such as automobiles, intensive computing, AI, so-called “edge” chips or even PC processors, the closure of its GPU division seems entirely improbable. The Cassandras will have to change the disk.