With power/performance getting close to par between x86 and ARM, the only way ARM-based mobile SoC vendors can keep up with Intel is on process and manufacturing.
Intel's ace has always been process and manufacturing.
Fabless chip vendors (Qualcomm, nVidia, TI, Xilinx, Altera, etc.) have gotten around this by going to TSMC and other foundries who have managed to keep costs low due to their location and by spreading R&D costs over a large number of customers. Further, the foundries were not forced to compete with Intel on process nodes till now since Intel used its advanced factories for its own captive manufacturing. And the ARM-based Intel competitors have vastly benefited by it.
This strategy will not work any more. Intel, with its new foundry services and huge capital expenditure, is going to force the issue with foundries. Having decided to offer foundry services, I think Intel will go after the premium fabless companies that don't compete with Intel (Xilinx, Altera, and perhaps even TI and several others) - and they, in order to retain their leading marketing positions, would want Intel to manufacture their chips for them.
This will have the effect of cutting off the air supply to TSMC and other foundries. If their premium customers who are capable of paying for capital investment, delivery priority, etc. move over to Intel, the foundries will have to fork up the capital investment themselves...and/or rely on the second-tier fabless companies to do so. Neither is a great proposition for the foundries.
With Intel Foundry taking away the premium fabless vendors, the foundries will be unable to keep up with the investment needed to help ARM-based mobile SoCs compete with Intel. And once they fall behind, it will be difficult for them to catch up with Intel - so Intel will have a long-term advantage!
So Intel wins in all 3 areas:
Power/Performance Check Check
Leading Process Nodes Check No
Manufacturing volumes Check No
Absolutely see this as the end-game. There is no other reason for Intel to expand its manufacturing capacity and capital outlay so significantly.
Anyone who suggests that it is an "end game" for either (ARM or Intel) is exaggerating. The market does not want a single architecture nor does it want a single supplier. Every customer wants to have the lowest price for its component parts to maximize profits and will therefore cultivate competitors for its vendors. These competitors will then be primarily be used as a lever or club to get better pricing and if needed used as an alternate supplier. If Intel did not exist, a competitor would be created.
Think about Facebook creating the OpenCompute group to get better pricing on equipment they want.
The semiconductor industry is a capital intensive industry. There is a high capital barrier to entry (takes billions and billions for entery) and continuation. Cash and investment is the oxygen that allows it to breathe. There are 3 large, heathy sources of semiconductors today: Intel, TSMC and Samsung.
The TSMC BOD and management is faced with funding decisions. They have a number of risks. How to fund. Who funds. How to handle the prospect of just the possibility of a large customer or two vaporizing. Intel/Samsung could capture them through a customer deal or foundry deal. Or not.
Lackeygarret apparently doesn't agree with Wally......and he looks more credible to me!
by lackeygarrett . Jan 25, 2013 9:15 AM . Permalink
Yes, ARM's ecosystem is the key differentiator that gives them an insurmountable moat that no other company can compete with.
Approximately 30% of the semi space has adopted the ARM architecture and the stated goal by ARM's management is to engage the whole semi space for their TAM.
With processors costs decreasing and functionality increasing, more and more processors are being used for 2-way communications. This requires a 32-bit+ processor.
Electrical devices can only run as fast as the network they run on, so as the mobile internet proliferates and the telcom networks keep increasing speed with LTE, more and more devices will be used.
The amount of processors being deployed yearly is growing larger year by year. This expanding market share pie will continue for the foreseeable future.
When LTE-A becomes mainstream the platform used will be all IP (using SoC's) and based on Software Defined Radio, which will be the catalyst for the Internet of Things.
This leads to a Smart Planet, which will use a Smat Grid, which is enabled by a smart meter, where the metering industry for 2-way communicating meters is consolidating onto the ARM architecture.
The smart grid and the internet will be based on heterogeneous computing/networking which merges the wired and unwired world together allowing interoperability between all electrical devices, regardless of the device's platform, protocol or operating system.
That is the requirement for future-proofness.
That is why ARM is a long term investment.
"Anyone who suggests that it is an "end game" for either (ARM or Intel) is exaggerating. The market does not want a single architecture nor does it want a single supplier. Every customer wants to have the lowest price for its component parts to maximize profits and will therefore cultivate competitors for its vendors. These competitors will then be primarily be used as a lever or club to get better pricing and if needed used as an alternate supplier. If Intel did not exist, a competitor would be created. "
[I think it would be erroneous to conclude that we are currently in the "end game". However, I think it's quite possible that it really is "game over" and we are just waiting for it to play out although it will take years for it to be fully realized - perhaps 5 years to be substantially reflected in the marketplace.
Bottom line is that FinFET is a disruptive technology (i.e. a new technology that has a serious impact on the status quo and changes the way people have been dealing with something, perhaps for decades). Of course, ARM would have us believe that this is not the case. But at this point we have no reason to believe ARM can do FinFET. Ortellini said ARM wouldn't have FinFET EVER. Now he's a guy who knows quite a bit about FinFET. I can't see him making that comment without thinking that FinFET is a mighty hurdle to get over. But here's the rub - It's not whether ARM can do FinFET but how long it will take. If it takes ARM four years to reach volume production on FinFET, then that's just the same as never. Because Intel will own mobility if ARM gets stuck at 20nm planar for four years. Intel will be doing 7nm by that time and have a huge advantage in both performance and power efficiency.
But wait, this will never happen because Intel doesn't have the capacity to take over all of ARM's foundry business. Oh, wait. My bad. Yes, they do have the capacity. Not only do they have the capacity but they built that capacity with no serious strain on their cash flow. Intel is ready to flood the market with state-of-the-art FinFET production. And plan on building more.
And this doesn't even include other disruptive factors substantially in Intel's favor. Like 450mm. Intel will be first to 450mm production on 10nm. ARM will still be stuck on 20nm planar when this happens. ARM has nothing to compete with either. And Intel will probably be first to EUV as well.
Intel is therefore on the verge of opening up technological, economic and manufacturing advantages that are easily classified as disruptive. And just because the market wants competition doesn't mean the market is going to get it. If ARM can't do FinFET, then no one else is going to come along with a decade of time and $25 billion to develop it. We know that ARM will be at least 2 years behind on FinFET and 4 years is very possible. So, the question isn't whether it will be disruptive. It already is disruptive. Anyone with half a brain knows how remarkable an achievement FinFET is. The question is how disruptive. Looking pretty disruptive.
All the evidence points to ARM being the next Lotus 1-2-3... ]
Sentiment: Strong Buy
Alex: The definition of "end game": the final part of a game, battle, or political process when the result is decided
I did not indicate that the result would be a complete elimination of ARM and the foundries. Nor the victory of a single architecture.
However, with its integrated design and manufacturing prowess, Intel should and will work towards expanding x86 in the mobile and embedded computing spaces which will have the effect of reducing the ARM/foundries competition to a minor role with presence in some segments.
IMO, Intel will do this with a combination of:
a) better SoCs
b) more advanced process nodes than the foundries
c) higher manufacturing capacity
(b) and (c) both have huge costs - Intel will spread the costs with an expanding Intel Foundry footprint targeted towards premium fabless customers. This will, of course, result in negative revenue impact on the foundries impacting their ability to invest in expensive future process nodes as well as new factory capacity.
This is how I see the next couple of years playing out.