Electronics Weekly Electronics Design & Components Tech News
It’s good to hear tales being spun of a post-Wintel world where chips are bought because of their characteristics rather than financial and legal pressure.
Sometimes tales come true. Twenty five years ago Andy Grove used to come over and tell the tale of how the entrenched x86 software base would make the 80386 the dominant computer processor.
Twenty years ago Sir Robin Saxby said ARM would be bigger than Intel.
Now Hermann Hauser says that, in the age of mobile computing, a licensing model for processors will prevail over the make-and-ship traditional model.
“The reason why ARM is going to kill the microprocessor is not because Intel will not eventually produce an Atom that might be as good as an ARM, but because Intel has the wrong business model,” said Hauser.
When Intel went sole-source on the 386 in 1986, denying its previous x86 licensees any share in the architecture’s success, it lit a slow fire under a pressure cooker.
Over the years, to maintain its quasi-monopoly, Intel had to resort to increasingly desperate measures like paying huge sums to its customers to keep them on-board.
So skewed did the PC processor market become that the US FTC Chairman, earlier this year, said that Intel’s conduct had the effect of: “Stunting innovation, diminishing quality and keeping prices higher than they would otherwise be.”
Would the mobile computing industry prefer a licensing model for its processors rather than a make-and-ship model?
You bet.
“People in the mobile phone architecture do not buy microprocessors,” says Hauser, “so, if you sell microprocessors you have the wrong model. They license them. So it’s not Intel vs. ARM, it is Intel vs. every single semiconductor company in the world.”
Hauser argues that, in every new wave of computing, the old industrial order was replaced by a new order. Now, he says, it’s time for the passing of the old order.
But will the sheer weight of Intel’s Market Development Funds, paid to dissuade mobile computer OEMs from installing non-x86 processors, stop that passing?
Well so far it has. Intel employs hundreds of smart business strategists to figure out how to maintain its business model.
But at least a tale how the post-Wintel age will look is being told.
And that may help it happen.
Mobile phone manufacturers do not want to buy IP. Apple does not buy IP – it relies on Samsung to build a chip around an ARM core. Apple does not care if it is an ARM. Same goes for most phone manufacturers. When Intel does come out with a competitive chip for smart phones (it is clearly where they are heading), then they will take a huge market share – they already have Infineon’s 2 and 3G basebands (and hence customers), so they have a very good start in that market.
Competition is good. ARM will not kill intel, but they will compete, which is good for us consumers (and technology).
I am a HUGE fan of ARM….as was. With Saxby at the helm they innovated at the technology, business model and go to market models and met a clear market opportunity.
What Saxby did as a go to market model was a masterclass in ecosystem development. I just wish he’d write the book because its one I’d read.
BUT – the market has moved on a long way, and fundamentally ARM has not. They are imho a sitting duck strategically. Take the ‘get into servers’ thrust – wow so forward thinking – so ‘of the moment’ – spot any sarcasm here? Or the ‘lets beat Intel at the big processor game coz we can be more green’ aaaargh – this is not BIG company thinking. It’s a me too knee jerk reaction that avoids ARM asking themselves what has changed at the market level that they now need to address – coz their old worn value proposition of low power is starting to show signs of weakness from the consumer perspective (note not the engineers – they are NOT ARMs concern anymore – or should not be!).
Intel has cash and market momentum due to the strength of their marketing machine and I am afraid they’ll grind ARM down eventually – but not any time soon – and therein lies the problem, ARM won’t change now because they don’t seem to perceive the need to change. Once they do – it’ll be game over already.
Let me give you a couple of simple examples:
1. I was at APPWORLD today, and Intel was there (ARM was not), and Intel launched a new sw dev platform called AppUp – they see the need to address this new audience directly – ARM has missed that, abdicating responsibility to their fragmented ecosystem. The market has moved on – so should ARM!
2. Ask your kids – what processor is in their phone – I bet 99% answer Intel if they are able to answer at all – after all what else? When that generation starts to develop software for the first time – guess where they’ll start? especially as AppUp will be sitting there waiting to make it real easy for them to get into Intel processors.
Wake up ARM – please.
Yes indeed, sugarplum, if MS eases ARM into computers, and if Google eases ARM into servers, then a good chunk of that $40bn-odd of annual revenues of Intel would be spread out into the wider semiconductor industry – providing a massive boost for innovation.
Google bought agnilux to make energy efficient ARM servers, Microsoft is buying ARM licenses. If these two giants gradually move more towards ARM, then intel is surely in trouble.
Selling XScale might pinch them a lot
I’m not sure I’d bet on Intel, here. The thing that Hauser says about Intel against the entire semi-conductor industry is spot on. On top of this, you also have to watch the guys that take an ARM architecture license. Although a lot of them are fabless, the fact of the matter is that the foundries are not sitting on their hands – they realise the threat that intel represents, and they are investing just as heavily. The common platform alliance are heavily capitalised and will catch intel on consumer nodes.
The problem Intel has is what happens when the fabless server chip startups – and big companies like Marvell – start selling a $100 server platform part into a server socket that nets Intel $900 – Intel will lower their prices. But how low can they go before they can no longer afford to swell their fabs at the rate that they do?
We might see Intel fall foul of their hubristic tendency to call out innovations that have been out in the world for years as their own (but by another name) in an attempt to gain traction in the mobile market. It will be money for old rope, and if the rest of the industry is flanking them on servers, how much longer can X86 stand its ground?
What you say, Zinu, sounds just like what Intel was intending to do with TSMC. For reasons as yet unknown, that arrangement foundered – quickly. A resurrection is possible, but it seems unlikely. But who knows? I certainly don’t.
Hi David, since ARM cannot do make-and-ship, it is not so surprising for Hauser et al to somehow jump to the observation that a licensing model is not only superior but also somehow inherently mutually exclusive to a make-and-ship model. Well, in light of the recent news on the foundry services that Intel has started to offer to some fabless design shops, it is not too far-fetching to envision the scenarios in which an x86 licensing model could well serve Intel shareholder handsomely by attracting SoC designs to reuse various x86 cores as IP blocks and take advantage of Intel’s manufacture strength through Intel foundry service. There seem to be plenty of receipes and opportunities to achieve win win for all participants.
I should point out, Zinu, that it wasn’t any of ARM’s execs who said this, it was a VC – Hermann Hauser. Hauser was the guy who, in 1990, put together the agreement between Apple, VLSI Technology and Acorn Computers which saw ARM spun off from Acorn as a separate entity. As for Intel licensing its cores – well, as you say, technically it could, but somehow one doubts Intel’s shareholders would be happy to reap 10 cents a core as ARM does, rather than the $100 per discrete chip which Intel gets.
Thirdly, one might add, Torben, that MS obtained an ARM architecture licence earlier this year.
Intel or AMD can technically license one of their x86 cores any time it makes business sense. So ARM has no inherent unique advantage over x86 in this aspect. There is no fundamental technical barrier for x86 processor designs to achieve better power, energy and area efficiency than ARM or other RISC processors. Like the historic debate/’contest’ between CISC vs RISC around late 80s and early 90s, ARM executive’s latest verbal salvo seems to be another histronic prognosis of Intel’s demise.
Fabrice wrote:
“The same can be said from Microsoft. They have build their monopoly together. They will die together.”
Microsoft has something that looks like an exit strategy from Intel. For one, they have invested heavily in Windows Phone 7, which is not tied to x86. Secondly, they have invested a lot in .NET, which is tied to no specific processor architecture. So they can, in theory, combine the two to make a desktop/laptop OS that is Intel-free.
Everything you say is true, Fabrice, but the reason Hermann Hauser gave for Intel’s demise was not the crazy old x86 architecture, but the Intel business plan – because mobile computer-makers like to buy their processors as licensed cores inside customisable chip-sets not as discrete chips.
Oh absolutely, technogeist, ‘Intel Inside’ was a massively successful stratagem for trussing up their customers like a lot of Christmas turkeys. The turkeys cannot now wean themselves off the MDF. And he who pays the piper . . . . . . .
Intel is doomed … just be cause of this:
– the top engineers at Intel are always allowed to the x86 business.
– to develop a brain damaged architecture (8086 was initially software (not binary) compatible with 8080 (an 8-bit), become more and more difficult.
– to justify its monopoly (build on illegal actions, see Sherman act), Intel pursue an x86 everywhere strategy (saying that x86 arch is just the best out there), up to the point then when they try to make software rendering based graphic chips (Larabee), they says that they envisioned other (new) architectures, but that it finally seemed to them that x86 was the best one for the task; then you have the total Larabee failure. Ok that just what happen when marketing dictate engineer choices. And that the current situation at Intel.
So Intel can only innovate on a technological dead end (x86 architecture). They cannot prepare something really new. This is just not sustainable.
Every mips win on this crazy architecture (from instruction set point of view), will become more and more expensive.
Every empire finally die. That just a fact of life. Intel has locked itself in saying to extend a dirty, crazy, and almost 40 years old architecture, is the best way to obtain performance. And this will consume all the resource at Intel, just forbidding them to invent their future.
The same can be said from Microsoft. They have build their monopoly together. They will die together. For Microsoft, the market (capitalization) is somewhat starting to anticipate this (a low capitalization relative to its earning means the market anticipate a slow decline).
You should not really compare ARM and Intel on turnover or trade value. ARM does not sell chips, so they get only a small fraction of the income of all ARM chips that are sold, where Intel gets the full income of all Intel-made x86 chips. Nor does ARM have a large fraction of its value bound in fabs.
So you should rather compare the complete market of chips containing ARM processors to the complete market of chips containing x86 processors — including those from AMD, VIA and whoever else clones x86.
I agree with Hauser that the mobile market wants licenses rather than ready-made chips. This gives more flexibility of design. Even if few mobile companies design their own SoCs, the fact that there are so many licensees makes the choice so much wider than if all came from a single supplier.
The server market is different: You don’t need a huge number of different SoCs. So, here, Intel’s model is not a showstopper. However, the server market is not so strongly tied to Windows as the PC market, so the main advantage of x86 (it runs Windows) is no decisive as it is in the PC market.
Reading some of the comments here leaves me with no doubt that Intel’s marketing has most of their user base by gonads.
ARM is agile, because it has no expensive fabrication plants sucking money out of their bottom line. ARM licensees are able to share the costs of retooling, while ARM invests more of its cash in R&D. Intel is spends inordinately huge amounts on keeping the gravy train from derailing.
Brits like to moan, because we can.
@Rakesh – GPUs are a great tool for accelerating tasks but have a look at Intel’s upcoming Knights corner and one can see that NVIDIA won’t have this market to themselves for much longer.
Thanks Rakesh, but I think Samsung still builds the odd advanced fab without any help, and Toshiba also builds fabs on its own account – as well as in conjunction with SanDisk – and Hynix, Sharp, Micron and Elpida have all engaged in fab-building on their own account recently. But Hauser’s point about the demise of Intel was that it’s business model isn’t appropriate to the mobile computing world which has got used to sourcing its processors in the form of licensed cores. Is he right? I don’t know. But it’s a debating point.
What’s going to keep Intel on top is their fabs. Intel’s currently the only semiconductor company that can finance a new state-of-the-art fab on its own.
As long as there’s a need for high-performance processors, Intel will remain dominant in processors.
ARM’s been bigger than x86 by at least two orders of magnitude in terms of unit sales for years, and it hasn’t done much to hurt Intel, and the smart phone and pad haven’t yet done any harm to the PC market — they’re still to complementary to truly compete.
GPU’s are more likely to unseat x86 than ARM, IMO.
If you look at our pharmaceutical companies, oil companies and banks, Chris, you could say we have produced businesses of American proportions – but it still won’t ever stop Brits from having a moan when they want to. I think it’s something to do with the weather.
hmm – Intel vs every other semiconductor company in the world ?
I’d probably bet on Intel 🙂 Samsung can run them closest on technology but their business model is very specific.
As for Herman’s other points, I agree people don’t buy microprocessors – they buy a platform that runs the software they want to use. So until there’s a Microsoft Office (not some underperforming and questionably legal clone) on ARM we will all stay with Intel/AMD for professional applications.
Intel has the wrong business model… what, compared to ARM… a company which trades at a P/E of 80 and a market-cap of £5bn compared to Intel’s $120bn (on a slightly more realistic P/E of 11)…
ARM is certainly a technological success but it is a pretty poor business. Perhaps if people stopped bleating about how unfair Intel is and produced businesses of American proportions our economy might be in a better shape.