Will Intel End Its Torment By Closing Its Graphics Card Business? It could be a 50/50 decision, says popular semiconductor analyst Jon Peddie (opens in new tab)because the company was looking for a way to get rid of its own weight.
“Should Intel ditch its AXG group? Probably,” Peddie says.
“Intel now faces a much stronger AMD and Nvidia, plus six start-ups – the rules for engagement have changed dramatically, while Intel has been pouring money into projects that seem to be failing.”
It’s true that Intel probably faced a much worse launch than it ever imagined for Intel Arc. Especially the first generation of Alchemist gaming GPUs. The first wave of discrete Alchemist cards has so far failed to reach a global market by 2022, and performance issues have turned Intel’s most recent launch headache into a migraine (opens in new tab).
The group responsible for Intel’s GPU development, the Accelerated Computing Systems and Graphics Group (AXG), reported a loss of $507 million in the past three months, significantly ahead of its first quarter losses of $168 million.
That’s not a huge amount of money for a goliath like Intel, although it’s very bad timing. Intel posted a rare $500 million loss in the quarter.
Combine that with the competitiveness of both Nvidia and AMD with their most recent generations of GPUs, and that they’re both looking for a new generation with new architectures, probably in the next six months, Intel is at risk of falling even further behind in graphics. before it really got going.
Since Gelsinger’s return, Intel has dumped six companies.
Jon Peddie
“…it’s a 50-50 gamble whether Intel will wind up and get out. If they don’t, the company faces years of losses as it tries to forge its way into an unfriendly and unforgiving market.” .”
Intel has invested in many new ventures and technologies over the past decade, including development of GPUs, SSDs, security, memory, automobiles, and even drones. Few, however, have survived the heel while tightening his belt.
Intel began moving away from 3D XPoint memory technology, the core of its Optane business, when it sold its share of the manufacturing facilities needed to make it to its partner Micron. Micron later sold this facility (opens in new tab) to Texas Instruments, citing low demand, and Intel finally ran out of memory (seemingly for good) last month.
“I never want to be in memory, you see, I’m doing everything I can to get out of our memory business in that regard,” Intel CEO Gelsinger told an analyst (opens in new tab) earlier this year.
Intel sold its NAND SSD business to Korean memory and storage giant SK Hynix last year. However, it will not be completely finished until 2025 (opens in new tab).
Intel-owned security company McAfee went public in 2020 and was later sold to an investment firm in 2021. The drone activities have disappeared, the sports activities are also bankrupt.
As for the automotive AI business Mobileye, Intel plans to float that but retain a majority stake.
“Since Gelsinger’s return, Intel has dumped six companies, saving $1.5 billion in operating costs and losses,” said Peddie.
So where is the loss-making AXG group? I’m in the camp that Intel still sees the value in offering its own graphics acceleration. There is clearly a lot of money to be made if you can successfully bring GPUs into the data center en masse.
Intel and Nvidia both offer data center technologies at scale, and both companies fell short of their targets for the previous quarter. Although Intel’s data center business fell 16% (opens in new tab) year over year during that period, Nvidia’s 61% (opens in new tab).
That’s not just due to the adoption of GPUs in the data center – Intel has also struggled to get its Sapphire Rapids server chips out the door and that will lead to much slower growth as companies wait with large-scale upgrades. Still, there’s a clear demand in the data center for accelerated computing — AI and machine learning are a big part of that — that Intel won’t be willing to give up without a fight.
Whether gaming will remain in the roadmap, I’m not so sure. However, I’d like to think it will last, if only because it would be a great disruptor for AMD and Nvidia, which is great for the GPU race and us, the customers.
There are also some big launches waiting in the wings. Ponte Vecchio, Meteor Lake, Falcon Shores… Intel has focused on chiplet-based architectures (which it calls Tiles) in its product portfolio – it won’t be easy to pull the plug on AXG without going all the way back to the chalkboard .
0 Comments