[1]: https://www.tomshardware.com/tech-industry/semiconductors/in...
[2]: https://www.tomshardware.com/tech-industry/intel-ceo-says-it...
When he joined only a few months ago, he set the vision of making Intel a worthy participant in the AI space.
Then just a few months later, he announced "we cannot compete".
What happened in the middle? Recent articles came out about the conflict between him and Frank Yeary, the head of the Intel board. He wanted to acquire a hot AI startup, and Frank opposed it. Two factions were formed in the Board, and they lost a lot of time battling it out. While this was going on, a FAANG came in and bought the startup.
I think his announcement that Intel cannot compete was his way of saying "I cannot do it with the current Intel board."
My read is basically that Intel's board is frustrated they can't part the company out, take their cash, and go home.
I'd also be incredibly frustrated working with a board that seems deadset on actively sabotaging the company for short term gains.
Now whether they fired the right tens of thousands is another matter.
According to Wikipedia, for FY25:
Intel: 102,000 employees
AMD: 28,000 employees
Nvidia: 36,000 employees
I'm pretty sure the latter two have been growing headcount lately, and even then combined they still have fewer employees than Intel.
Now intel isn’t doing as well on the chip design side as amd or nvidia nor as well on the fab side as tsmc but i suspect that’s on leadership thrashing constantly like it is more than anything else.
[1] https://www.tomshardware.com/tech-industry/intel-will-keep-u...
>"On training, I think it is too late for us,"
Not too late for AI, but too late for training meanwhile there's inference opportunity or something like that
With their own fabs, at that
Seems like you'd prefer yet another +1 selling AI oil and promises ...
Same as how they messed up the Core Ultra desktop launch, of their own volition - by setting the prices so high that they can’t even compete with their own 13th and 14th gen chips, not even mentioning Ryzen CPUs that are mostly better in both absolute terms and in the price/perf. A sidegrade isn’t the end of the world but a badly overpriced sidegrade is dead on arrival.
Idk what Intel is doing.
In other countries, any place that sold them at near-MSRP very much was out of stock immediately (not that there was that much of it to begin with, which is my critique), leaving other vendors to raise the prices to around 350 USD and the more greedy ones and scalpers all the way up to 400 USD, I saw those listings myself:
https://www.tomshardware.com/pc-components/gpus/where-to-buy...
https://www.digitaltrends.com/computing/intel-arc-b580-out-o...
Whether anyone actually bought those is another question entirely, but a GPU that's only available to you at 350 USD but should have been sold for 250 USD (and performs well for that price) is just plainly bad value.
It is nice that since the prices have dropped, however the hype around it is already past, so we're in the long tail of the product's lifecycle, especially with things like the CPU driver overhead being identified and getting lots of negative press. They had a chance to move a lot of units at launch thanks to positive reviews from pretty much everyone... and they fumbled it.
It seems like a bad choice at all times. A product with a 45% margin -- or a 5% margin -- is profitable. It's only the products with negative (net) margins that should be cut.
And meanwhile products with lower margins are often critical to achieving economies of scale and network effects. It's essentially the thing that put Intel in its current straits -- they didn't want to sell low-margin chips for phones and embedded devices which gave that market to ARM and TSMC and funded the competition.
Intel did this for memory in the 80s. Memory was still profitable, and could be more so again (see Micron), but it required much investment.
But Intel might not be in this position, and filling the fabs by itself can defiantly be worth it.
But if you don't have the capacity in the new fab, maybe that isn't an issue, so its hard to say from the outside.
It's usually not about the number of people. If you have two projects and both of them are profitable then you can hire more people and do both, even if one of them is more profitable than the other. The exception would be if that many qualified people don't exist in the world, but that's pretty rare and in those cases you should generally divert your focus to training more of them so you don't have such a small bus factor.
Another common mistake here is the sunk cost fallacy. If you have to invest ten billion dollars to be able to do both X and Y and call this five billion in cost for each, and at the end of that one of them has a 5% net margin and the other a 75% net margin, or even if the first one has a -5% net margin, you may not be right to cancel it if you still have to make the full ten billion dollar investment to do the other one. Because it's only -5% by including a five billion dollar cost that you can't actually avoid by canceling that product, and might be +20% if you only include costs that would actually be eliminated by canceling it.
I also don't think its not as rare as you suggest finding people. Depending on your location and industry. It takes time to add people. Good people existing somewhere on the world wasn't enough, specially before remote work.
Also if you can grow your 50% margin business even a little bit faster by focusing on it, over the 5% margin business. It doesn't take that much focus for that to be the better choice. So if to achieve this 5% margin, lots of your best people work it, shifting those to the larger margin business might make sense.
But I agree, if you are a mature company that has the needed infrastructure to keep that product alive then doing so makes sense. Specially because maybe in the future it could be a more important better business.
>"On training, I think it is too late for us,"
Not too late for AI, but too late for training meanwhile there's inference opportunity or something like that