Как хиреют мастодонты

[vak]

"Intel lays off 15% of its staff and its stock tumbles as it struggles to turn around its ailing business"

Есть у меня на работе коллега John Oliver. Он некоторое время назад в Интеле трудился. Делится воспоминаниями на LinkedIn:

I can't resist my $0.02 worth of input. I'm probably going to regret this. In the very early teens, there were six of us mere engineers in the embedded space at Intel who REALLY tried hard to get Intel management to listen to us about the (for Intel) "crazy new" technology called AI (Artificial Intelligence). Intel management saw AI architecture as competition to their dominant lead with Intel Architecture (IA) and rejected our recommendations. I understand why. Years earlier when Intel acquired StrongARM technology I asked Gelsinger (a General Manager at the time) why we/Intel were not pursuing the benefits of RISC architecture. Intel Embedded could have REALLY used something like RISC in the embedded space. Gelsinger's response was that they (Intel management) had witnessed too many companies going under because they adopted multiple architectures and could not sustain the burden of propping them all up. It was Gelsinger that killed the StrongARM at Intel. Months later my engineering colleagues and I decided to try again, but with a new tactic. We positioned AI as an adjunct/complementary accelerator to IA ... you know, like a GPU. But it was too little too late, Intel management saw us coming and again rejected our technical counsel. I can't help but feel somewhat responsible, for it was my job as a techie to influence and guide the business leaders to do the right thing; and I failed. Nonetheless, I find it ironic that one of those early business management key influencers has since happily retired from Intel and is currently sitting on the Board of Directors at ARM Holdings. Ha! I never had access to senior Intel management, but I did interact with a bunch of Intel middle management. Funny, everyone I talked to had read the words of Peter Drucker and the book "Innovators Dilemma" by Clayton Christensen. All the signs were there ... and at great personal peril, I highlighted them. Yet, the pressures of the moment compelled Intel managers to focus on immediate deliverables and did so at their peril. Not REALLY at their peril, because ALL of them have since retired from Intel and are no longer around to blame. The whole thing breaks my heart as I witness good friends who have devoted their entire careers to Intel being pushed out the door via yet another big layoff. The Intel legacy is OVER ... deal with it (unless they productize their Neuromorphic technology, then watch out).

https://www.linkedin.com/posts/delton-john-oliver_artificialintelligence-technology-intel-activity-7226996257762787328-H2GG

Comments

[juan_gandhi]

Yep.

And is not it magic that IBM (brain-dead, in my humble opinions) still survives.

[dijifi]

Just a gentle reminder: Gelsinger was kicked out of Intel in 2009 after the Larrabee disaster that he was in charge of, and became the CEO of VMware. He came back to preside over the dumpster fire at Intel in 2021.

[sergegers]

А я недопонял, чувак считает, что бизнес Intel погибнет из за AI? Думаю, его AMD выебла без смазки. Вот где нужно держать оборону и контраковать.

[kondybas]

Да не в АМД дело, а в АРМе, верней, в RISC, преимущества которого біли совсем неочевидні на заре технологий.

Примерно так же у коммутации каналов при масштабировании цена растет єкспоненциально, а у коммутации пакетов - логарифмически. На начальном єтапе коммутация каналов чуть ли не дармовая, но по мере масштабирования и усложнения ВНЕЗАПНО оказівается, что с пакетами все гораздо более иначе. Но телеком уже завяз в каналах и просто не умеет міслить иначе.

[sergegers]

Это вообще странно звучит. В Intel под CISC фурычит RISK, и затраты на трансляцию составляют меньше процента, и никакого экспоненциального роста ни по какому параметру нет.

[ichthuss]

При коммутации каналов все же полиномиально, а не экспоненциально.

[dedekha]

25+ лет назад несколько лет работал контрактором в стартапе который помимо неплохой по тем временам идеи что делать с пакетами на краю сети решил быть впереди планеты всей и сделать все на strong-arm. Это была не главная но все-таки существенная причина неудачи этого начинания - баги в периферийных чипах заставили использовать все процессоры на 50% частоты и все карты имели проблемы с пропускной способностью + плюс приходилось на коленке разбираться с экзотическими по тем временам версиями gcc, netbsd и gdb.