To set the stage: I’ve heard the recent news about layoffs with Intel. Before that I read from their new CEO “On training, I think it is too late for us”. Lastly there has been some offhand comments (from LTT) that they’re preparing to sell the company.
Yet while I have no doubt that they are behind; their revenue is about 55 billion since 2023, down from the high of 78-80ish Billion during the pandemic, but about the same as the plateau leading up to the pandemic 2015-2019.
Maybe i’m naive about the way businesses work; but if your still profitable, and you know you need to “catch up” why lay off people and close sites? Maybe that works for a consumer goods company; if your overhead is too high and your not making a profit: slim down.
However for a company where RND is really where the value is, like Intel, it just doesn’t seem to make sense; your not going to get better designs and processes by reducing your experienced staff and letting them go work for the competition. Maybe some restructuring, (in the engineering sense not the euphemism for layoffs).
Intel’s core product became big primarily due to decisions they had nothing to do with.
Roll back the clock to IBM developing their first PC. They usually developed hardware in house, but they were late to the game and needed something fast. They choose to use off the shelf hardware, with the BIOS being the one thing that’s proprietary.
For the CPU, they could have gone with Motorola or MOS or Texas Instruments. They choose Intel. Why? Because Intel fulfilled a memory contract on another project, so sure, use their 8088.
Compaq then reverse engineers the BIOS and the whole thing pases legal muster. Now anybody can make a compatible, but they have to use the same CPU that IBM did.
Microsoft does its thing with DOS and Windows. Everyone is writing software against that. Now everyone starts getting locked in.
After that, all Intel has to do is keep x86 going well enough that nobody wants to make the effort to switch. Yes, AMD and Cyrix are out there, but at this point, they’re both the cheap alternative that isn’t as good.
They fucked up the 64-bit transition. AMD did the version we all use now.
Intel’s entire success is based on making good on a memory contract to IBM decades ago. That’s it. It hinges primarily on decisions they did not make themselves. Weren’t even in the room at the time.
Also they did that netburst stuff that was a bad idea, and let AMD beat them to the punch on integrated memory controller… So many scenarios where Intel flubbed the “big change”.
However they have frequently been good on the little stuff. For a long time they had a fab advantage (and have messed that up for a decade now). There are decent at dotting the "I"s and crossing the ts, including broader software and firmware enablement.
Nowadays the only software people care about is CUDA and AMD actually has cash to take care of a lot of little things. Intel kept trying to make side efforts happen, like having a rack of memory and processors freely associated in ways no one ever asked for, phase change memory which while cool, was an awkward in-between NAND and SDRAM, various FPGA efforts which they never really figured out the actual point, putting an infiniband variant in the processor package in an awkward way that just made things worse, and a ton of weird accelerator architectures they abandoned within a couple years each time…
Yes, and that really puts the Raptor Lake failures in perspective. It undermined this whole argument that Intel stuff is rock solid even if it’s not the best in other ways. Xeon stuff was affected, too, and given how lucrative the server market is, that may have been the biggest mistake Intel has made.