News

Will AXG survive Gelsinger’s axe?

Five years of no results in the midst of dumping businesses

Jon Peddie

 

Intel’s financial report revealed two hard-to-ignore situations. In a quarter where the company reported a loss combined with dropping margins and sales, Pat Gelsinger also jettisoned several non-essential business units that contributed to the drain.

Of the groups Gelsinger got rid of was Optane (started in 2017, never made a profit), sold McAfee (bought in 2010, never made a profit), and shut down the drone group (started in 2015, never made a profit). Last year, Intel sold off its NAND business to Hynix, giving up its only Chinese fab, and sold off its money-losing sports group; and this year, the company shut down its Russian operations. Since Gelsinger’s return, Intel has dumped six businesses, saving $1.5 billion in operating costs and loses.

Gelsinger is not afraid to make tough decisions and kill pet projects if they don’t produce—even projects he may personally like.

So, is the company’s dGPU group next? Started in 2016 the dGPU group snatched showboater Raja Koduri away from AMD with great fanfare in 2017— it looked like Intel was all in. And they were. The hiring continued unabated up till 2021 when they got their latest prize, Tom Petersen from Nvidia. Almost anyone who could spell GPU could get a job at Intel.

Since Q1’21 when Intel started reporting on its dGPU group, known as AXG or accelerated graphics, the company has lost a staggering $2.1 billion and has very little to show for it.

In fact, the company has actually invested more than that. We estimate the number is closer to $3.5 billion (or more, depending on how Intel chooses to account for those investments, and the setup costs at TSMC).

The logic of why Intel would invest in a dGPU operation is obvious—the GPU is in multiple segments from supercomputers to Chromebooks and represents a major socket Intel doesn’t fill with a dedicated chip. At that same time, Intel was buying every AI semi-company it could find—dGPUs were the darlings of AI training. Since then, some of Intel’s other AI investments have shown more promise. And its efforts at building a supercomputer dGPU have slipped; the company has announced a second-generation part, Rialto Bridge before the much-touted Ponte Vecchio was even fabbed.

Intel’s results with dGPUs in the consumer space have been an embarrassment with just a quiet release from an obscure Korean notebook maker and no AIBs available until lately. The various leaked results have been less than stellar, and the availability date keeps slipping.

The company has done a great job on the PR front with Koduri dropping hints on Twitter, in-depth presentation of the generalized design at Intel’s Architecture Day and other events, and guest appearances at popular tech Web sites. The drumroll never stopped, even to the point of talking about code improvements in a Linux driver—Linux? That’s Intel’s first choice?

So, the rumor mill has been hinting that the party is over and that AXG would be the next group to be jettisoned. That rumor was denied by Koduri in one of his tweets, saying, “We are very much committed to our road map. We are ramping Alchemist and will continue to improve the experience.”

But once again, it’s a someday statement and not even an official statement, just a tweet.

Should Intel dump its AXG group? Probably. The company started the project six years ago. Since then, AMD and Nvidia have brought out three generations of new and stunningly powerful dGPUs, and more are in the pipeline. Four new companies have started up in China, and two new ones announced in the US. Intel is now facing a much stronger AMD and Nvidia, plus six start-ups—the rules of engagement have dramatically changed while Intel sunk money into projects it can’t seem to get off the ground.

Not many CEOs would put up with that, especially while repairing their company from previous misguided investments. Gelsinger was brought in to clean things up and get back to the company’s core strengths. The dGPU program is noble in its concept, intriguing in its alleged design, and an adventure too great for even Intel, especially in these days of recovery.

The best thing Intel could do at this juncture is to find a partner and sell off the group. It could even be dressed up as a strategic move, just as they did going to TSMC to build the dGPU in the first place. The company can’t continue to carry an enormous payroll, pay a competitive fab for wafers, and then ask governments to subsidize its investments in new fabs that can’t even build the parts they are presumably designing. Not only is that a bewildering investment strategy, but it’s also an embarrassment.

At this point, being one of the blind men feeling the Intel elephant, it’s a 50–50 guess whether Intel will wind things down and get out. If they don’t, the company is facing years of losses as it tries to punch its way into an unfriendly and unforgiving market.

Perhaps the clouds will lift by the end of this quarter.