Intel's leaders have stopped pretending – and it's about time Pat Gelsinger's tenure as Intel's chief executive was epitomized by his unwavering optimism and ambitious plan to return the ailing chipmaker to its former glory. His successor has no such delusions of grandeur. For Lip-Bu Tan, Intel may as well be a gut job, and he's setting expectations accordingly. In a leaked recording …
They relied on x86, DOS, Windows and PCs. They had i960, DRAM, ARM and other tech & products but didn't develop them. The x86 was a deeply flawed architecture rescued by AMD, yet Intel listened to HP and did the doomed Itanium. I don't know what the problem was with Optane. For over a decade Intel integrated graphics was rubbish unless you only did simple wordprocessing and spreadsheets.
It was only the PC, over-priced CPUs and production engineering that made them a market leader. In reality it was dead-end company by the end of the 20th C. Compare RCA, once the mightiest company spawning HMV and JVC and gone by 1986. Marconi was once a part owner of RCA but the US government kicked him out as a foreigner.
Lets not forget that Intel lost a lawsuit for stealing parts of the DEC design. The tipoff was they bragged about it in the Wall Street Journal.
"When Intel introduced its Pentium Pro chip in November 1995, Mr. Palmer said he was surprised by the new chip's substantial increase in performance. His suspicions grew last August, he said, when Andrew S. Grove, Intel's chief executive, and Craig Barrett, its chief operating officer, seemed to admit in an article in The Wall Street Journal that Intel took its chip designs from others."
https://www.nytimes.com/1997/05/14/business/suit-by-digital-says-intel-stole-pentium-design.html
Did Intel listen to HP? or was it HP listening to Intel?(perhaps a bit of both..) I only recently learned that the i860 was a VLIW processor like Itanium. I was aware of i860s existence but never knew anything about it(before my time, never looked into it, other than seemingly it was the first CPU that WinNT was designed for). By contrast I think I remember my Mylex RAID controller around 2000 having an i960 which I know was a totally different kind of chip despite having a very similar name(wasn't until not too long ago that I learned they were so different), was the last/only time I ran RAID 5 at home(with 5x9G SCSI drives I think), everything RAID1 or 10 since.
It's what happens when the company puts profits first...
I often wonder what the world would look like now had Intel taken the StrongArm that they inherited and run with that as a basis for their future. For one thing they'd probably dominate the mobile market... But no. Pride is mighty powerful, and also mighty dangerous...
They wouldn't have dominated with their follow on to the StrongARM. The XScale was the worst ARM core ever, they clearly did not understand the simplicity of the existing design and ended up with an utter dogs breakfast. They clocked it up from 200MHz to 600MHz, but sacrificed instructions per clock with lots of ridiculous pipeline interlocks which have never been in any other ARM design before or since. Tellingly the XScale erata was longer than the entire reference manual for the ARM7.
Optane was just "fast cache" in an awkward position in this stack
L1 cache, cycles to reach
L2 cache, dozen or so cycles to reach
L3 cache, hundred or so cycles to reach
RAM: thousands of cycles to reach (1600 MHZ vs 3 GHz)
oPtAnE: several thousands of cycles to reach
SSD: ... A dozen thousand seconds
A completely new computing design could take advantage of that, but with many deep hardware changes to, say, the bus lines between CPU and the optane socket. It's and the 50-90MB cache-on-die with the Ryzen chip goes fast: it fits well between RAM and cpu cache, *because* it was a redesign of the cpu-on-a-stick and hard-drive-on-another-stick networking in the motherboard
You're *almost* right on point, yet so far off … Intel has no commitment to anything, BUT their x86 alone. Though even that has been fading badly since.
Intel is notoriously in the very same position ever since two whole decades, as they *still* see their age-old x86-architecture as the ONLY lone solution for every possible problem ever arising. Period.
Intel not only completely missed the boat on everything mobile, they fully intentionally ignored the ocean-liner's very bugle and alarm-horn, when they wilfully gave Apple the cold shoulder on a deal for their iPhone-SoC in 2006/2007 and shrugged off about ARM's potential — A monumentally stoopid move of Intel.
– Intel refused Apple their ARM-based chip (which was, what Apple was asking for), for deliberately NOT supporting any other architecture (read: like ARM for instance) and anything BUT their own x86. Intel was only willing to offer Apple a Atom-based x86-design instead (which Apple refused, of course), then Apple in turn told Intel to go pound sand (for Intel's high sand-castle under construction).
– Intel deliberately refused Apple to supply them anything ARM, despite sitting on the industry's single-most powerful and having the market's omni-potent ARM-architecture within their own portfolio; DEC's famous StrongARM™ Intel got, which was overtaken from Digital (when bought out from DEC over their lawsuit-settlement due to Intel's IP-theft before on DEC's ALPHA-processors), while StrongARM by then was already renamed to XScale — God forbid that a product at Intel might give off the impression, that Intel could possibly support anything but x86, much less a competitor-architecture!
– Instead, Intel passed up on the chance with Apple, only to then turn around immediately after the iPhone-deal fell through, to demonstratively and well-intended sell everything StrongARM/XScale to Marvell (even INCLUDING every given related personnel!), out of spite in a fit of cold-hearted calculation and determination towards anything x86 – So that no-one ever again could possibly mistake Intel for everything else but a one-stop x86-shop exclusively.
Intel actually HAD everything they needed to offer for Apple (in terms of a potent StrongARM-based ARM-design), yet they demonstratively did NOT wanted to offer Apple anything but their own x86 alone. Also, the sole reason WHY Apple approached Intel (for a ARM-based RISC-design), was that the Apple Newton already was based upon DEC's StrongARM SA-110 (which Intel then was sitting on).
There is a certain irony in the fact that, Intel immediately after turning down Apple (and selling former StrongARM to Marvell), went on blow through BILLIONS of money, only to fight the then very spawned myriad of ARM-vendors (like Samsung, Qualcomm, MediaTek and the rest of ARM-licensees), Intel *itself* created in the first place by refusing Apple, which then ALL got filthy rich on making tiny little ARM-cores, Intel stupidly refused to fab before (For reasons of increased margins, allegedly) – Everyone of these ARM-licensees became absolute IP-power-houses and/or ARM-heavyweights in the market, often directly competing with Intel head-on.
Mistakes were made along the way, I guess …
“Spirits that I've cited, My commands ignore!” — Johann Wolfgang von Goethe – Der Zauberlehrling (The Sorcerer's Apprentice)
So their very x86-blinders are the reason, why they completely dropped the ball on the following smartphone-revolution and utterly failed on everything with regards to the whole mobile market, was them refusing to supply Apple their ARM-based iPhone-SoC – Allegedly for reasons of margins, when in reality that pretence from Otellini (based on profits), was nothing but a cheap shot and lame excuse from Intel, when they didn't really refused based on margins, but because they refused OUT OF PRINCIPLE.
Of course, *publicly* Intel refused allegedly "for reasons of margins", yet what Intel actually wanted, was to bring Apple to use a x86-design, instead of anything ARM, of course – When Steve Jobs insisted on a ARM-design, Intel PR came up with this myth over margins (to save some face) for declining the deal of the century through plausibility …
Care for another prominent example of Intel willingly ignoring everything else, which isn't x86-based?
Picture this: Intel's “Teraflops Research Chip” project aka »Polaris«.
A awesome platform for HPC and General Purpose-computing of every kind, basically a scientist's dream architecture, which every possible compute-workload could be thrown at and applied to with rather ease, to run whatever kind of simulation-games, floating-point bomb-loads or weather-models, and being architecture-agnostic even no matter the given architecture to boot! All in early 2007!
– A architecture-agnostic (non-x86) custom ISA.
– A 2D mesh network-on-chip (NoC).
– 3D-stacked SRAM (Freya) offering up to ~1 TB/s bandwidth.
– Designed to nominally operate at 4GHz, Intel showed even 5.67 GHz!
– Sweetened up with some for once *really* open Parallel Studio-SDK, offering its C/++-based Intel Ct programming API (spelled, ›cute‹) to boot – Basically Intel's own CUDA …
– Delivering no less than 1.0 TFLOPS at 3.16 GHz with just 62 W of power-draw already, it was extremely efficient too.
In a nutshell: The most-promising candidate of Intel's Tera-Scale Computing Research Program …
Offering a performance-metric and power-efficiency, which AMD and Nvidia only could show to offer over a year later in the middle of 2008 …
At only significantly (AMD: 110W) or even vastly (Nvidia:236W) higher numbers of power-draw! So neither AMD nor Nvidia (the compute-king today) could come close to it, at that point in time.
Since a comparable GeForce GTX 280 (the first nVidia-card breaking through the TFLOPS-barrier), needed already ~236W to achieve the very same.
AMD's earlier Radeon HD 4850 (1TFLOPs) and their first of reaching one teraFLOPs of processing-power, had a TDP of 110W while their Radeon HD 4870 (1.2TFLOPs) a TDP of 150W.
Yet still, all graphics-cards reaching the very same single-precision floating-point compute-performance, came out only in 2008 … over a year afterwards of Intel's POLARIS itself.
That was, until [›Gee, who could it possibly be?!‹] himself k!lled everything Polaris and shut down all of Intel's Teraflops Research Chip project (as a perceived Intel-internal threat to his own imaged project), for the sole reason of Polaris being NOT x86-based. Only to then push a DOA x86-based approach called “Larrabee” …
—
What's Larrabee you ask?
Larrabee was the prominent DOA-project of nothing but a bunch of already badly-aged cores off their x86-architecture slapped together, and that was literally it …
Technical base of Larrabee as a HPC/General Purpose-computing architecture;
– Lame-o age-old P54C Pentium-cores from 1994 (fifteen years later in 2009/2010), helplessly gobbled together as a wannabe compute-cluster.
-- Those Pentium-cores were only used for the sole reason, that one could've had more of them on the same die-space, as they're *smaller* …
– Getting their plain superscalar age-old Pentium-cores slapped crippling In-order execution (derived from Atom-cores).
– Getting plastered with nothing but their pumped SSE vector-processing units added (4×128-bit wide; making it 512-bit wide altogether)
Ironically, adding their pumped up SSE-units actually nullified the initially saved floor space, but you cares at that point, right?
As evident, performance left the equation early on, and so did logic, of course.
Since simply put, project Larrabee was actually nothing but the embodied morônic attempt put in hardware, of trying to basically brute-force their way into graphics, with as many x86-cores as possible (since to this day, x86 is all Intel can think of) — It was of course accompanied by a load of marketing-crap, being allegedly superior and faster than anything Nvidia.
Theoretically, you could've ran Windows on Larrabee, since it was basically a bunch of clustered CPU-cores.
So all in all a fundamentally flawed and just plain laughable design-approach to begin with (since a general-purpose CPU-design like x86 cannot possibly ever be as efficient, as a specialized and well-tailored ASIC for the same given purpose of computing!) and it failed for exactly that reason – That's also why its mere rehash Xeon Phi failed as a compute-cruncher, also software-stack.
→ Trying to bring a highly sped-up yet serialized design for a highly-paralleled computing work-load.
A approach which was necessarily doomed to fail from the beginning already on paper, even before put to work. Pushed in favour of Polaris.
—
Now guess who, as Intel's CTO and prominent clown manically dancing around on their IDF for years, went on and knifed Intel's “Teraflops Research Chip” project aka »Polaris« in favour of Larrabee?
Tip: It's the very same key-person who lately publicly declared, that Nvidia "just got lucky with AI" … and it's also the very same person who (still can't seem to let go of his personal baby Larrabee, obviously), declared that Intel shoulda, woulda, coulda a Trillion-dollar company today, if it weren't for Intel firing him back then in 2009 and knifing his Larrabee.
No it won't. Microsoft will announce that 'incompatible hardware' is not supported and one update later it's not running.
W 11 is not running *now* on anything 'incompatible', whatever that is. That is not going change.
Intel is like an old tree: It has been absolutely rotten from inside a long time: It's just so big it takes a long time for it to fall.
But nothing can save it now: Competitors are eating their lunch and they've fired all the competent people. Or they left when none of their ideas went to production.
That's literally what happened to Nokia when an MBA took the CEO position: "Not enough profit, we're not making this". Do that couple of years and the actually competent staff leave.
MBA is inherently incapable to understand what happens and starts to cut costs when sales goes down ... and the Cycle(tm) speeds up.
Agreed. It's a waste of time, money and talent to try to catch up with NVidia and AMD on the GPU front. Maybe they can purchase some innovative start-up in the A.I. field to get them going.
CPU's aren't sexy these days but they're bread-and-butter for Intel. They should focus on remaining competitive there.
BTW I've steered clear of Intel CPU's since they introduced the Intel Management Engine, the spy-CPU within my CPU with no way to turn it off. I'm not saying this has been their downfall but it certainly hasn't helped. Maybe other potential buyers have the same gripe. They should investigate if this has had any impact on their sales.
As things stand, Intel is unlikely to have a competitive GPU or AI accelerator for at least a year, and that's assuming that Intel's GPU team survives the next wave of layoffs.
I picked up an A770 for my work rig and it was dire, unstable, glitchy, a driver update lost my third screen (support were absolutely useless even when provided all the information and pointing out where the driver change broke things) that I eventually pulled it out and got a 5070. The driver updates being bigger and bigger steps backwards in terms of stability was the last straw (bsod so hard I had to re-install windows).
Stuffed it in my NAS, figure I could at least get some value out of it for hardware transcoding of video, and nope, kept locking the system up and preventing me logging in, absolute waste of sand, sits on a shelf now as a reminder to never buy a v1 of anything.
I have an Intel Arc chip in my laptop, and I think you're being way too generous by assuming this was just a V1 issue.
Intel's big problem (well one of them) is that, like Sony, even though they are a giant bloated sprawling behemoth they can't concentrate on more than one or two things at once. For the last couple years they've been really concentrated on Server and Mobile (wow, two things at once, better than Sony!) so Desktop/Gaming just fell off a cliff and Graphics continued to be third rate.
I have some friends at Intel and have heard many tales about how, under Pat, the Mandate Would Come Down From the Emperor that Here's What's Important Now and everything just grinds to a halt and everyone tries to course correct (or your reviews are at risk) when, at the very least, you want to keep doing well what you're already doing well, but no that isn't important any more. Maybe the New Emperor at least understands that.
I assume it is logically impossible to focus on multiple things as long as all the core loops rotate around control and monitoring.
So it basically boils down on what the C-suite is interested in and all other stuff is more or less abandoned because people on other projects need to beg audience from CEO erc who are fully invested in their pet projects.
So projects are fine as long as metrics look good. Too bad if there are any major blockers that would require authorisation but nobody will do that. Projects just struggle keeping metrics as good as possible until one day it is impossible and everything just starts imploding. And company axes the project. And when company has axed multiple projects the panic is getting real and they'll probably axe also multiple viable/promising projects just in case. Because losing one's face is not an option for higher ups. They'll rather burn the company down completely.
This post has been deleted by its author
Been an exclusive AMD CPU desktop PC guy through thick and thin since the original Athlon largely due to Intel's illegal "marketing practices" in the early days. I did buy a couple of Intel Inside (TM) laptops because until recently good AMD powered laptops were hard to find. My current laptop is powered by AMD and other than not supporting expanding the memory to 32 GiB from the current 16 GiB, I love it. I don't want to see Intel go away, just to keep AMD on it's toes. Healthy (looking at you Intel) competition is a good thing.
My recent two desktop builds (March 2025, July 2025) used AMD, both with 65W TDP and integrated graphics.
The reason to move to AMD from Intel was purely an upgrade path. The AM5 socket will allow future proofing to some extent. With Intel they change sockets too frequently.
The systems they replaced were a 10 year old laptop which was limited to 8GB, and a 11 year old desktop that still worked as needed, but i upgraded it for fun. (both run linux).
I don't see AI as a panacea for the PC industry or peoples needs, yet Microsoft may be the force that compels people to purchase processors with AI NPU's despite people not needing them.
I actually think the opposite is true regarding needing Intel to keep AMD on its toes.
Intel held too much power over the PC industry. It kept AMD in its lane and prevented any newcomers from entering the market through barely legal deals (e.g. OEM bribes) and withholding x86 licenses.
Now that Intel is weak we are seeing a far more healthy environment develop (Qualcomm, Nvidia, Apple etc.). A resurgent Intel would likely drag the PC industry back to the comfy complacency of the x86 duopoly.
Now all we need is for Microsoft to get knocked down several pegs!
My two most recent laptop purchases have an Apple M4 and a Snapdragon X-Elite chip in them respectively.
Apple's second-slowest laptop chip is about the same speed as Intel's second-fastest laptop chip.
The Snapdragon was in a much cheaper laptop, so obviously you have to calibrate your expectations accordingly, but it is still a pretty decent performance.
Both have very impressive battery life.
"Using a software company who did not own the software
Nonsense. Microsoft owned 86-DOS which they licensing to IBM.
"and buying up a job lot of silicon that no one else wanted because better chips were already available ..."
Interesting. Which CPU would have been a better choice?
MS bought DOS-86 from SCP after discussions with IBM in Dec '80 for porting to the IBM PC.
Other processors around included the 68000 which is a very nice instruction set to develop for, I think even the z80000 with MMU would have been a better option.
"MS bought DOS-86 from SCP after discussions with IBM in Dec '80 for porting to the IBM PC."
True. As I wrote, that contradicts OP's 'Using a software company who did not own the software'.
"Other processors around included the 68000 which is a very nice instruction set to develop for,
Yes. But it wasn't available:
"So why aren’t we all using 68K-based computers today?
The answer comes back to being first to market. Intel’s 8088 may have been imperfect but at least it was ready, whereas the Motorola 68K was not. And IBM’s thorough component qualification process required that a manufacturer offer up thousands of “production released” samples of any new part so that IBM could perform life tests and other characterizations. IBM had hundreds of engineers doing quality assurance, but component qualifications take time. In the first half of 1978, Intel already had production-released samples of the 8088. By the end of 1978, Motorola’s 68K was still not quite ready for production release.
And unfortunately for Motorola, the Boca Raton group wanted to bring its new IBM PC to market as quickly as possible. So they had only two fully qualified 16-bit microprocessors to choose from. In a competition between two imperfect chips, Intel’s chip was less imperfect than TI’s."
"I think even the z80000 with MMU would have been a better option."
z80000? You serious?
Why don't add Glaze3D and Copland to replace CGA and PC-DOS?
I stand by the statement that when initially dealing with IBM, there was not even a license agreement in place for 86-DOS. The purchase followed later after a single licence had been sold to IBM.
https://en.wikipedia.org/wiki/86-DOS and that Douglas Adams's quote perhaps sums it up ... https://www.azquotes.com/quote/357838
With regards processors, the IEEE article is a good summary. I was USING the TMS9900 at that time and had a development system driven by it. Yes it was quirky, but so was the PC at that time. I had a memory that AMD was producing a third option at that time, but I may be getting that mixed up with the AMD 29000 which appeared later? That today there is not a single x86 based device in mobile devices also speaks for itself? That said I've just pulled a new machine together with an AMD Ryzen processor, with Linux driving it because the price was right :) The windows machine is going nowhere as Microsoft will not update it to W11 'because the hardware is not good enough'.
Where would we be today if IBM had done a better job 45 years ago?
The 8088 (or 8086) wasn't a real 16 bit CPU. It used 64 K segments. Even the 286 (which was 16 bit) was a flawed 16 bit cpu.
The 8088 in the PC used an 8 bit bus and and was easy to port 8080 8bit code by translation of the binary because it was so similar and crippled. The Z8000 was a bit of a disaster and not ready. IBM even had their own 16 bit CPUs then. Also PC-DOS/MS-DOS had no subdirectories and had no advantages over CP/M 80 till about Version 2.11. Initially mostly ported CP/M applications. Also the real clock speed at 4.7 MHz was really 2.35 MHz. The choice of MS-DOS and 8088 and the practically out-of-a-catalogue ISA bus and hardware and lack of built in peripherals (initially) maybe held back SW & HW development for the consumer by 10 years.
"The 8088 (or 8086) wasn't a real 16 bit CPU. It used 64 K segments."
It is an 16-bit CPU. Compromised but 16-bit.
Are you also in the same camp that doesn't consider DOS as an Operating System?
"Also PC-DOS/MS-DOS had no subdirectories and had no advantages over CP/M 80 till about Version 2.11."
(CP/M-86, not -80)
DOS had the advantage of availability from the start. CP/M wasn't available for PC buyers until later.
"Also the real clock speed at 4.7 MHz was really 2.35 MHz."
<citation needed>
It is 4.77MHz hobbled by slow bus. If you are referencing to memory wait states or such - CPU's are memory starved even today.
"The choice of MS-DOS and 8088 and the practically out-of-a-catalogue ISA bus and hardware and lack of built in peripherals (initially) maybe held back SW & HW development for the consumer by 10 years."
DOS was Hobson's Choice - there was no CP/M for PC. IBM could have waited until DRI pulled the finger out but in those days new computers and computing innovations were coming out every month (if not faster) and time-to-market was critical.
Lack of bus-connected peripherals from the start is always a problem when creating new buses. Which bus would have been better?
I was deliberately comparing with Z80 running CP/M 80 (maybe 6 MHz by 1981). CP/M 86 and MS-DOS were on the ACT Sirius 1 which was in the UK before the IBM-PC, but despite better display and HW was also crippled pseudo 16 bit 8088/8086. It wasn't a real 16 bit CPU. See differences Z80, 8088/8086, the poor 80186 (Not used in regular PCs) and the 80286. Later Z80 systems had also more than 64K.
You could run Wordstar & Supercalc better on some Z80 computers on CP/M 80 ( dual 8" 1 Mbyte drives) than on an original PC (5.25" 160 k byte), which didn't even have graphics originally. Remember the original PC had no HDD. Clock HW, serial and parallel ports where options. Graphics was the later abysmal CGA later unless you got 3rd party Hercules. Even later EGA was pretty miserable.
Then to take advantage of the 286 you had to run Xenix or something. DOS used it as an 8086. All the PC nonsense of EMS and Extended memory and the stupidity of not putting graphics / display access at the end of 1M memory.
The fact that DR hadn't yet done CP/M 86 was irrelevant. DOS was a stupid OS, though as CP/M it was a reasonable compromise for an 8080. It should have been a real 16 bit CPU with something like BSD Unix (released 1978).
"ACT Sirius 1"
In a whole another price category - why bring it into discussion?
IBM didn't develop PC to be a high-end product - they already had those - but as an affordable and expandable personal computer.
"You could run Wordstar & Supercalc better on some Z80 computers on CP/M 80 ( dual 8" 1 Mbyte drives) than on"
Which Z80 computer model are you referring to?
"original PC (5.25" 160 k byte), which didn't even have graphics originally."
CGA was offered from the start. The PC could be bought with dual drives. The drive was 180k, actually. DOS 1 just formatted is as 160k.
"Remember the original PC had no HDD."
I remember. Very expensive. The weak PSU in PC didn't have enough power for a HDD.
"Clock HW"
Which computers did have a battery backed clock?
"serial and parallel ports where options."
Yes, a cost saving exercise. If you needed a printer or a modem, you bought one. Mice were not really a thing yet.
"Graphics was the later abysmal CGA"
Yes. If only all the CGA hacks were known back then...8088 MPH demo
"Then to take advantage of the 286 you had to run Xenix or something. DOS used it as an 8086."
Yes.
"It should have been a real 16 bit CPU with something like BSD Unix (released 1978)."
The obvious questions now are: Which CPU are you suggesting?
I'm also re-asking your choice of bus, which you didn't answer.
What would be the hardware requirements for BSD for this CPU be?
How much would the total cost of PC be after adding more memory? (and probably secondary floppy drives or HDD+controller+bigger PSU)
Intel got stuck with the x86 line. I asked someone reasonably senior at Intel why that legacy architecture, wasting transistor budgets, and was told, for every customer criticising there was one spending 10x as much demanding compatibility. They failed to find a way to make the transition away. But in their heyday they were special. I met Gordon Moore once and I couldn't believe how interested he was in my humble views of Intel's products. He didn't care I wasn't a billionaire CEO, he wanted to know what everyone thought. I saw Intel do a lot of things that were more usual in an innovative startup. They had some truly great people and a work hard, play hard culture. So, I find it sad to see them in decline but to reverse it needs a huge investment and a risky one which demands the right leader, one that is both a brilliant strategist and technologist. Doesn't sound like Tan is that man.
"Twenty, 30 years ago, we are really the leader," [Tan addressing the troops.] "The world has changed. We are not in the top 10 semiconductor companies."
"And it's all your fault ! "
Don't laugh. That is the implicit and sometimes explicit accusation from corporate America and its government. Nothing to do with self serving and incompetent leadership in both. /s
Unfortunately, I think Tan is the final stake in the heart. His credentials are not hardware. They would have been better off keeping Pat I think. And really Tan, not in the top 10? I have to disagree. They are probably around 5 or 6. And they have fabs. Tan is just torpedoing what little moral is left at Intel.
LBT may be the Bad Guy[tm] CEO brought in to cull the ranks, cut projects, shutter buildings/sites, and just generally depress or annoy anyone still around into self-vacating their job.
Y'know: "streamline" the company (read: make it smaller and less expensive to operate, though probably no more productive nor efficient).
In a few years (over-under at 3, place your bets) the Board tries a new "visionary" CEO, perhaps another blast from the past like PG was supposed to be, if they can somehow find one. Anybody's guess what happens then. Probably not much different, if I had to guess.
Not saying this is what Intel *should* do; honestly I have no idea how to climb out of the years-in-the-making hole they've dug into; but given the track record, it does seem a likely path.
"Tan, it is reported, believes a smaller Intel will be better prepared to compete."
Yes, it could be argued that chipzilla could, possibly should, be smaller.
Problem is, getting there has historically seemed to do as much or more harm than good. That is, laying off techs, devs, and engineers is unlikely to accomplish much beyond causing more work for the remaining folks, draining experience and tribal knowledge, and further depressing morale.
Thinning out the management ranks, with a focus on mid- and upper-middle layers, might actually help unravel some of the bureaucracy, moribund decision-making, political maneuvering/infighting, and still make a dent in the payroll bottom line. Might even raise morale a bit among the rank and file too -- that hasn't happened in a while.
Unfortunately it appears that more techs than management are being let-go, which is pretty typical of most corporations and not surprising. Managing by the numbers includes layoffs by spreadsheet.
"Despite repeated attempts, Intel has largely failed to find a foothold in the AI accelerator arena"
So the same as they couldn't hack it in mobile modems or GPUs, the roaring success that the Itanium wasn't, or maybe... you know what, I could be here all day at this rate. They just had the x86 market and they took it for granted. How long before they're circling the drain?
"So the same as they couldn't hack it in mobile modems"
So bad that Apple paid $1B for the business? I've used (and supported) plenty of laptops with Intel WWAN cards. Mostly harmless.
"or GPUs, the roaring success that the Itanium wasn't, or maybe... you know what, I could be here all day at this rate."
Hindsight 20/20?
Intel has had its share of technological dead-end products besides Itanium - but they're hardly alone.
Intel GPU's are very popular in laptops, BTW...
As we said at the time, Intel missed the AI boat, at least in the datacenter, and it seems that Tan would agree. "On training, I think it's too late for us," he's quoted as saying, characterizing Nvidia's grip on the market as simply "too strong." ... Tan is apparently jazzed about the prospect of AI agents that can automate entire tasks traditionally performed by people.
This seems to be a continuation of the same losing strategy - give up on Intel's core strength of design and focus on some far downstream application with vague specs, big hype, and unproven demand. e.g., IoT -> AIAgents. (Sure something like AI Agents is a future thing but it can only be built organically on more efficient hardware and algorithms which will lower costs dramatically, but its not something you can plan/design a chip product for by 2026Q4).
Meanwhile, NVidia is actually planning to cut production of consumer AI-capable RTX5090 this year to keep prices artificially high, a sure sign that the market is screaming for competition. That's a market that is ripe for exploitation, as racked datacenter AI services have problems with predictable billing and keeping client data safe, small business' (factories included) need price reliable alternatives.
As for the rackspace market, Google has the TPU and Amazon has the Tranium - somehow those have both carved out niches and are both competing with NVIDIA.
Intel desperately needs to trim down its workforce. What are those 100.000 employees doing when most of the company's products are manufactured by external fabs like TSMC? All sales, marketing, managers, interim managers and management support? Your idea is as good as mine.
The company is extremely bloated for its manufactured output. I would postulate they can slash two-thirds of their workforce and still remain as productive. It would hugely improve their profits which could be invested in new fabs and process improvements. Intel is desperately lagging there and is continuously playing catch-up with TSMC and Samsung, relegating it to an also-ran.
"I would postulate they can slash two-thirds of their workforce and still remain as productive"
Most probably yes. But, it never is the middle management which is fired: It's always those people who actually produce something. Like new products.
In the end there's only managers basking in their own excellence and company goes bankrupt as they have nothing to sell nor anyone to sell it, if they had something to sell.
Yes seen it, been in the middle of it. On the techie side and junior management side. The worst for the company is when they cut management. The ones that survive tend to be the sycophants that comply and obey, never speaking the truth when they believe their boss wrong. Because their boss is someone who has an ego problem and needs to have his arse licked. I firmly believe in a chain of command and that the boss's final decision must be followed but when people can't feed criticism and honest opinion upwards the outcome is obvious; decline.
You can't be an effective leader without hearing a range of thoughts and understanding them, but unfortunately you can still climb the greasy pole.
In the 70s I worked for Intel's (briefly) largest customer (8008, 2102 and other sundries).
Even then they were a myopic, difficult supplier, and seemingly always having delivery/process issues.
Our buyers claimed it was the 1702A and cross-licenses that kept the company afloat.
I still think Gelsinger was the riyman for the job and when Intel gets better, it will be because of the decisions he took.
À monkey could fire tens of thousands of employees, sell assets, stop R&D marketing and whatnot to jack up apparent but fake profitability.
What Intel needs to relearn is fortitude : stop launching a product or service to systematically abandon it 3 years later. Why should your customers, partners and developers trust you and invest in so risky techs?
In the case of Intel, firing tens of thousands is an absolute prerequisite. Yes, anyone can do it, but can you get the support of the board to do it?
Intel needs bumper profits not to buy-back shares, but to invest in new products and manufacturing prowess. This will take years and years. Any CEO will also have to resist pressure from the shareholders, which ain't easy.
You hit something there; "to jack up apparent but fake profitability."
There is a problem in the shareholder system in the West. It is the dominance of global institutional investors. They don't care if a company goes bust or declines because they also invest in the competition who pick up the lost business. So, they just demand maximun dividends and share performance. This becomes aligned with board bonuses. There is no financial space to retrench and change direction if that means reduced or zero payouts for a while. So they shed costs and sell the assets, which includes people and technology. They have to do this constantly until someone much bigger buys what remains and "might" be willing to inject money. It become nearly impossible to change path.
"Twenty, 30 years ago, we are really the leader," he's reported as saying. "The world has changed. We are not in the top 10 semiconductor companies."
I wouldn't advocate hopeless optimism but that statement isn't going to do anything for productivity and innovation. The only thing that will do is make it easier to trim the workforce.
