Came here to say exactly that. Markets are zombies and dollars are brains. But the title is less thrilling than "Ok everyone knows how to make a tablet now, that won't be a high margin business." In this case margin being the proxy for expertise in putting together a tablet.
But there actually is a story here which VentureBeat came so close too and missed, the cost of compute.
Even in the 80's when I went to work for Intel people 'in the know' thought it was insane that Intel could capture 33% of the value of a PC with just the CPU + chipset. They worked really hard to make that the case, using every gimmick at their disposal, from patents to copyrights to support contracts to outright strong-arm tactics. Even today that survives in the $300 you pay for the CPU on the motherboard. Mostly it was about the software base though, you really couldn't do anyhing about it unless you were willing to bite off on the problem of the software and tools. Apple did that, but nobody else did.
ARM is different, the ARM folks take their tax, but you can buy out their tax if you know you are going to sell enough chips, and the cost to make chips has gone down. Even though the cost to build chip factories has gone up (remarkably so). I'm not nearly as engaged with this problem as I once was but around the turn of the century the 'sunk' costs of doing a chip was on the order of 5 - 7 million dollars. A wafer start (with anywhere from 1 to 12 300mm wafers) is like $48K [1]. Less on the older processes, so may $1 base cost per tested good packaged die. So you if you price to cost + margin there was a huge amount of room below the Intel x86 artificial cost 'floor'. ARM software has gotten better and with Linux and then Android creating volume, additional players entered the market. And some, like All Winning have very modest margin targets on their CPUs, others like BroadCom/TI/QualComm have been playing in the smartphone market where these CPUs don't carry as much margin as general purpose compute devices do. You need look no further than the Raspberry Pi to see what that means in terms of consumer cost.
Anyone can become an ARM cpu vendor, its damn difficult to become an x86 compatible processor vendor. And that is the story we're seeing play out here. From the sidelines I'm watching to see what Intel does next, they are already scrambling (see the Cedar Trail stuff) and they have a lot of smart folks who can do great stuff with manufacturing and architecture. For the first time since the Motorola 68000, they have legitimate competition.
Is it mine in the sense that it spewed out of my keyboard without much forethought? Yes. But the concept that markets are an unkillable mechanism that seek out money is universal.
What about AMD, should they copy ARM business model to compete with Intel in desktop market? Or maybe AMD is too similar to Intel and cannot adapt ARM licence model due to its internal costs structure? (Innovator's Dilemma)
AMD is a good example of how competing with Intel doesn't work well. They kicked Intel's ass with the AMD64 architecture but they didn't get commensurately rewarded.
Look at their lawsuit vis-a-vis Intel and their predetory approach to Dell. They accused Intel of threatening to cut off Dell if Dell put an AMD chip in their servers. At NetApp where I was arguing for Opteron rather than the turd Intel was pushing (Itanic) as a 64 bit solution I got to see all of it, the whole "We might not want to buy your products ..." or "We really only provide NDA roadmaps for our 'comitted' customers ..." etc. Not so with ARM vendors, there isn't anyone to cut your leg off. Don't like the TI part? Hello BroadCom! Can't get docs on their GPU, hello Exynos! It's competitive and its easier to play in.
Tangentially, because I'm finding your comments so fascinating: it looks to me like AMD doesn't really know how to compete with Intel from a technical standpoint in the desktop and server space, but their default approach works much better in the cheap and ARM space.
For example, AMD seems to have bet a lot on their bulldozer architecture, but as far as I can tell it's been almost a total flop in the space where Intel really dominates. Even aside from the OEM space, system builders like my hardware guy look at options from Intel and options from AMD and, currently, go with Intel every time (which is a shame, I personally would like to see AMD giving Intel a rough time).
So what's your take? What happened with Bulldozer? Does AMD have the technical chops to beat Intel?
As an OEM stepping away from Intel is a very risky thing to do. Intel executes really really well, AMD does not. Intel lives and breathes process technology, AMD does not. Intel has discipline to walk to the end of the road, AMD does not.
Using a software example, if you notice the guys who write the gaming engine put out technically better games than the people who license the engine? I attribute that to the ability to internalize all of the benefits and weaknesses of the engine. If you license the engine you 'kinda' know what it does but because you don't have to know your knowledge never gets deep enough to do what the originator can do.
AMD competes against Intel on architecture, and they do Ok at that, but they cannot compete on process. You will notice they gained server share on Intel when Opeteron released because Intel was caught sleeping (they had targeted Itanimum 64 bits, Pentium/Xeon 32 bits, 8051 Etc for SoC/embedded. They were resisting putting 64 bits into the Pentium because it would cut into their already weak Itanium story. As long as they stayed there, AMD gained ground. They capitulated on architecture and then retook ground using process improvements (TDP mostly), meanwhile AMD didn't cement their lead, rather their CTO left and they played dancing CEOs. Boat stalls, momentum is lost.
So no, I don't think AMD can beat Intel organizationally and that will keep them from beating them technically.
AMD put a lot of resources toward the desktop. And from all the benchmarks I've seen they achieved their goal. Bulldozer is a pretty competitive desktop processor. Bulldozer is powerful.
But not energy efficient. There can be no Bulldozer laptop.
And the world switched to laptops. Right now the split is 18% tablets, 60% notebooks/netbooks, 22% desktops.[1] And desktops are losing ground rapidly.
AMD simply has nothing to compete with Intel in the laptop space.
For the same price, Intel scores something nearly double on the benchmarks (1700 for a4-3400m vs 2900 i3-2350m). For everyday computing in the real world, that performance difference is:
AMD: "This is so slow I want to tear my hair out."
Intel: "The CPU is no longer a bottleneck.
I need a new SSD for CPU speed to even matter."
And the Intel chip does all that using SO MUCH LESS power than the AMD (a4-3420m=35TDP and i3-2350m=15TDP).
And as you move up away from the budget laptops on the price/performance curve, things rapidly get worse for AMD.
AMD's failure was strategic, not technical. Even if Bulldozer had been 50% faster, AMD would still be on the verge of irrelevance.
But there actually is a story here which VentureBeat came so close too and missed, the cost of compute.
Even in the 80's when I went to work for Intel people 'in the know' thought it was insane that Intel could capture 33% of the value of a PC with just the CPU + chipset. They worked really hard to make that the case, using every gimmick at their disposal, from patents to copyrights to support contracts to outright strong-arm tactics. Even today that survives in the $300 you pay for the CPU on the motherboard. Mostly it was about the software base though, you really couldn't do anyhing about it unless you were willing to bite off on the problem of the software and tools. Apple did that, but nobody else did.
ARM is different, the ARM folks take their tax, but you can buy out their tax if you know you are going to sell enough chips, and the cost to make chips has gone down. Even though the cost to build chip factories has gone up (remarkably so). I'm not nearly as engaged with this problem as I once was but around the turn of the century the 'sunk' costs of doing a chip was on the order of 5 - 7 million dollars. A wafer start (with anywhere from 1 to 12 300mm wafers) is like $48K [1]. Less on the older processes, so may $1 base cost per tested good packaged die. So you if you price to cost + margin there was a huge amount of room below the Intel x86 artificial cost 'floor'. ARM software has gotten better and with Linux and then Android creating volume, additional players entered the market. And some, like All Winning have very modest margin targets on their CPUs, others like BroadCom/TI/QualComm have been playing in the smartphone market where these CPUs don't carry as much margin as general purpose compute devices do. You need look no further than the Raspberry Pi to see what that means in terms of consumer cost.
Anyone can become an ARM cpu vendor, its damn difficult to become an x86 compatible processor vendor. And that is the story we're seeing play out here. From the sidelines I'm watching to see what Intel does next, they are already scrambling (see the Cedar Trail stuff) and they have a lot of smart folks who can do great stuff with manufacturing and architecture. For the first time since the Motorola 68000, they have legitimate competition.
[1] http://www.xbitlabs.com/news/other/display/20110912192619_TS...