|
I liked Pat and I think he should've been given a longer leash, but I feel the commenters here have been looking at the situation through rose tinted lenses, probably because Pat's PR very effectively presented him as an engineer and not a bean counter. In reality, his tenure was filled with disasters and not much measureable upside: 1) while expected, the process node issues haven't been resolved. Instead Intel had to rely on TSMC to push out a semi-competive mobile platform. 18a and other processes have been delayed and 20a outright cancelled under the guise of rapid 18a progress (how rapid can the progress be, when 18a was just delayed again?). And this is just what we know externally, the board must have much better info anyway. A far cry from the promised "5 nodes in 4 years" and more than remeniscent of the situation in the decade+ before. 2) Pat single handedly directly harmed Intel's bottom line on almost day 1 by trash talking Taiwan and partner TSMC (which was supposed to be their lifeline to gain access to competitive process nodes) in a tactless attempt to attract US government support. This resulted in TSMC cancelling the massive ~40% wafer discount for Intel's chips. 3) in line with his Larabee lineage, Pat invested significant resources into Gaudi AI accelerators and Arc discrete GPUs. Both uncompetitive, released borderline unusable with pathetic software support and on the verge of cancellation from day 1, which killed what little enthusiasm they might attract. Even the much smaller (employee # wise) AMD has been able to push out serviceable AI/GPU software and competitive DC cards for training and has been riding the AI wave on Nvidia's coattails; Intel just looks completely lost. 4) arrow lake, released years into his tenure, is characteristaly delayed, overpriced and uncompetitive. Software issues pushed the needle for the launch towards disaster. 5) raptor lake instability/degradation scandal and months long lies and embarrassing messaging from Intel's side making people literally avoid buying Intel CPUs 6) Epyc Turin just annihilated the most recent Xeon gen, making Intel further uncompetitive for DC/HPC on the CPU side for at least another year, but probably more. 7) lack of any timely decisive and proactive moves which Intel desperately needs. I can't identify a single area where Pat pushed hard and fought against the inertia. He might've been more honest about Intel's financial woes (through his unnecessary constant presence on business news channels), but if the situation was so bad, why did he do nothing but complain? Why did it take so long for him to cut the dividend or lay people off? If he cut the dividend on day 1 and made Arc his Manhattan project, he would be much more palatable, even if the project missed the AI wave. Instead, he seemed to spend time clamoring for political attention while projecting the exact image of Intel in 2010s: excuses, delays, and underwhelming products. In totality, whatever Pat's been doing hasn't resulted in any real, tangible results. If anything, Intel's disasters have accelerated in frequency and scale under his helm and they've been losing ground at an accelerated rate everywhere, possible unrecoverably (mobile, graphics, desktop, DC/HPC).
Of course, turning Intel around is a massive undertaking and as always, there have been glimpses of promise and innovation. But it doesn't feel fair to cherry pick the rare positives from the sea of negatives and assign them to Pat. Still, it's completely possible that the results of his work were just not seen yet and he should've been given more time - at this point, we'll never know. |
Becausefrom what I've read on forums, by all accounts from the insiders, the technology behind 18A is incredible (gate all around transistors, backside power delivery, 3d Chip stacking...) and that in a year or two ,it should put Intel manufacturing a little ahead of TSMC.
Although, have also read that the one problem is that even 18A won't be profitable for a couple years. Yeah, that the problem with Gelsinger's original plan is that 3 years wasn't enough time to get to profitability, but 5-8 (but that it still is a good plan).