Hacker News Reader

Comment by Traster | Hacker News Reader

I don't think it's fair to say that Intel 'sat' on their monopoly. From McAfee, through Altera to (whatever that stupid company that had the drones was), Intel was desparately trying to leverage their monopoly to expand in to new markets. What I think is unique about Intel when compared to their rightful competitors (FAANG + M/S + AMD + Nvidia) is that Intel don't pay for talent. Intel have been cost orientated for atleast a decade now, when 14nm cost spiralled. The problem wasn't that Intel was a monopoly so much as it's just that Intel has been living it's life in the world of real companies, not silicon valley. Intel is viewed/valued/run as a manufacturing company. So they need to hit the eternal treadmill of performance whlist spending like they're just doing average work. It's difficult to do that when Google is just going to 10x the salary of any of your 10x staff.

View on HN

Intel is not wrong about being a manufacturing company though. Their true competition is TSMC, whose formula to success is having an army of process engineers working three shifts figuring out every little detail on an island that has seen stagnant wages for more than a decade. Intel's mistake was to believe they should diversify and nearly every acquisition outside of their core competence turned out to be a disaster.

Turned out to be a disaster?

Intel's acquisitions have been pre-existing disasters, also-rans and underdogs looking for their big break, often hoping that getting access to Intel's fabs could make their next generation of chips more competitive. Many—if not most—of their acquisitions over the past decade only really make sense if they were intending to leverage Intel's supposed core competencies to not just get Intel into new markets, but grab a lot more market share in those new segments. And then being part of Intel ended up not really helping those companies, because Intel couldn't deliver on their core competencies.

Not true for NVidia. Companies ultimately sell stories, not products. Tesla could be as much as manufacturing company as Ford but they are not. Stories are very different.

You could say, running the company with MBAs, instead of BSEEs, when their competitors were doing the opposite.

Putting the money they spent on acquisitions, into manufacturing technology, process innovation, etc. would have kept them right up front.

As for paying for talent, they certainly paid enough dollars for talent, but could have directed those same dollars at fewer resources with the talent they actually needed. That means paying top dollar for the right people, and skipping the rest.

If they let others poach, that is because those competitors were smarter at valuing talent while Intel either slept or fell victim to broken HR policies.

Intel was shipping 4-core desktop/laptop CPUs for years when they could have had 10-12 cores. They were definitely sandbagging until AMD forced their hand.

Is this true? I thought most of the reasons why AMD can push higher core counts is largely caused by very recent manufacturing innovations with chiplet/infinity fabric.

Namely that it enabled the manufacturing process to glue together slightly broken CCXes to enable high core count lower binned parts.

Intel has had no trouble moving their mainstream product lines from 2-4 cores up to 6-10 cores, without copying AMD's chiplet playbook, or using their own more ambitious chiplet interconnects, or even delivering significant improvements to their CPU fabrication process. They just stretched out a ring bus a bit further to add in more of the same CPU cores they'd been shipping for years, with very incremental improvements/fixes. Aside from some of the clock speed increases and some of the integrated GPU updates, all the big improvements they've made to their consumer 14nm chips since 2015's Skylake are things they could have delivered in 2016 if there had been competitive pressure.

Not to mention the 2 core 15 watt laptop CPUs.

Which most developers have no idea how to use them for, beyond running more OS processes.

Or more like Developers have no incentive to optimise the usage for it purely because so little people have it.

Go to any developers meeting and make an informal query about how comfortable most are with these kind of subjects, I bet most would fail a parallel programming Pub quiz.

I certainly might, and all applications that I have written commercially since 2000 make use of parallel programming and concurrency.

Intel's US employees are mostly in Arizona and Oregon, which are much lower cost of living locations than the Bay Area and Seattle.

I would be curious to know which companies pay more for process engineers than Intel in the US. My best guess would be Samsung?

From Glassdoor for Process Engineer (not senior): Intel average is 124k, Apple average is 143k

Apple doesn't hire outside of Cupertino, and the cost of living there is easily double Arizona and probably 50% more than Portland. I'd rather make 125K in Arizona than 143K in Cupertino that's for sure.

TSMC

I don't know if it's the salary delta so much as Intel feels like the modern version of Fairchild, IBM, or Sperry-Rand; business-suit wearing lifers more interested in defending their established market than creating new ones or competing with themselves.*

So as the Andy Grove generation retired out, very competent and talented but ultimately risk-averse replacements came in.

So they failed to make meaningful inroads into mobile and are now getting flanked by the myriad of companies who did as those companies are moving upstream with lower power, lower heat and thus potentially more scalable chips.

Intel also failed to rise to the occasion of "pivot or preserve" and just went to a "preserve at all costs" strategy even when it was clearly not working out for them. Chips got delayed 6 months, then 12, then 18, then got redefined just to get something marginally better out as their true intentions continually got pushed further.

And now, instead of say, sniping some stars from Rockchip or the other amazing up and comers, they did a reorg from within; a very GE/Texas Instrument style move that usually does not work - especially when you have already passed a generational changing of the guards.

But they have plenty of market still captured, a continued strong brand identity in the vast majority of the public, plenty of money to pad a runway and plenty of decent products so there's still significant time to turn their ship around and not continue to ignore every word of every book of Jack Trout, Al Ries, Clayton Christensen, Geoffrey Moore, and Steve Blank. First they need to stop believing "it can't happen here" when it's clearly already happening here.

If they don't, they might become another Bethlehem Steel or Sears; deftly unable to read the cards despite evidence attacking them from all directions. They need to realize their old ways will not carry them into a new world.

Look how Microsoft flipped. As a hardcore Linux user I would have never thought I'd ever be giving cash to Microsoft (via GitHub and Azure), and monthly at that! These things are possible: "When losing ground focus on your potential instead of preserving position"

Even after whatever AMD is going to say on October 8 and whatever Nvidia is going to do with ARM, the chip ball is still in Intel's court for now and it's still their game to lose. I am pretty confident they will figure out this rough patch and a substantial recovery will come (come see my stock positions if you don't believe me). My tune may change in 6 months...

---

* The modern Sperry-Rand, at Unisys, for instance, is STILL maintaining their UNIVAC operating system - why? Good question. Do they think that's still a potentially winning hand? A path for recapturing a market they lost 60 years ago as if say, somehow hot new startups will start choosing OS 2200 running on Univacs? I don't know either. I downloaded and tried their free virtualized x86 version about 6 months ago just to see what it was like and I got a high-touch white-glove sales-treatment of emails and phone calls. It all felt very desperate. (https://en.wikipedia.org/wiki/OS_2200 and https://www.unisys.com/offerings/clearpath-forward/clearpath...). I wish them the best of luck, they'll need it.

>So as the Andy Grove generation retired out

They had Patrick Gelsinger. But he was pushed out. You left with a Board lead by former CFO Andy Bryant, chooses a CEO that fits his taste.

It was all Politics.

> It was all Politics.

And this has trickled down since.

Sure. Apple had the Sculley/Spindler/Amelio era and Microsoft had the Ballmer era. These things are recoverable if they get addressed in time. (Then there's the SGIs, Yahoos and SUNs of the world - coasting for years into oblivion; the Radioshack/A&P Grocery model. I don't think Intel is in that camp though.)

>Apple had the Sculley/Spindler/Amelio era and Microsoft had the Ballmer era

I would actually argue those two had strategy problem rather than execution. And those two had comparatively little politics / power play involved.

Intel has both, and politics.

I have to disagree. The politics and power play of the kicking out of Jobs is famous, kind of the textbook case. Woz even quit over it.

The mantle passing to ballmer came at the objection of many major stock holders but some tiny cabal of gates et al overrode it. For years there were protests and threats that people would sell off millions if they didn't can him. There was substantial slippage and he lost all support and bounced, like Lee Iacocca except without any glory days

> Intel was desparately trying to leverage their monopoly to expand in to new markets.

If your business is in an industry with very rapid innovation, would it be wise to lose focus on your core products and processes?

Reply on news.ycombinator.com