Intel’s ARMageddon

In ten years we’re going to rec­og­nize this week as the moment Intel ceded its lead­er­ship posi­tion in the global chip busi­ness. Microsoft’s deci­sion to embrace ARM sig­naled more than a desire to take on Apple’s iPad — we are wit­ness­ing the end of the ‘Win­Tel’ era and the emer­gence of ‘ARM­Dows’ — a seis­mic shift in the IT indus­try is about to take place.

Intel and Microsoft are like a mar­ried cou­ple who can’t stand the sight of each other and yet whose for­tunes are so inter­twined that they can­not afford to get divorced. Despite what you may read in the press releases there’s no love lost between the man­age­ment teams of the two com­pa­nies. Rather the oppo­site might be closer to the truth. Both have been seek­ing a way out of their co-dependency but the urgency has been greater for Intel.

Chip mak­ing is a com­mod­ity busi­ness. The only thing that mat­ters is per­for­mance and price per watt i.e. how cheaply can they pro­vide the most per­for­mance with the low­est energy con­sump­tion. That real­ity is borne out by the dizzy­ing ros­ter of proces­sors Intel pro­duces. Each is uniquely ‘tuned’ by the mar­ket­ing depart­ment to tar­get a very spe­cific eco­nomic demo­graph­ics of the com­puter buy­ing public.

Intel’s con­stant night­mare is the relent­less path to com­modi­ti­za­tion for x86 archi­tec­ture proces­sors. The pres­sures cre­ated by AMD in this regard lead to behav­ior which the Euro­pean Commission’s com­pe­ti­tion author­ity deemed to have “dimin­ished com­peti­tors’ abil­ity to com­pete on the mer­its of their x86 CPUs.” In des­per­ate cir­cum­stances, com­pa­nies some times resort to des­per­ate measures.

Intel’s dreams came true when it started play­ing around with Linux. Here at last was the per­fect part­ner; no one really con­trolled Linux. It could be molded to Intel’s needs and would finally offer the oppor­tu­nity to escape Microsoft’s cold embrace. To say that Intel’s increas­ing infat­u­a­tion with Linux didn’t go down well in Red­mond would be like say­ing the gulf oil spill was a small leak.

The day Intel announced their engage­ment with Linux was the day SteveB and the boys in Red­mond started work­ing on the divorce paper­work. At CES in Las Vegas this week the writ was served.

The ARM archi­tec­ture can now be found in over 90% of all mobile phones. It pow­ers the iPad the fastest sell­ing com­put­ing device of all time and it is now about to become a very direct alter­na­tive to Intel chips pow­er­ing hun­dred of mil­lions, if not bil­lions of devices which will run Windows.

Microsoft’s abil­ity to craft a com­pelling response to iPad and the emerg­ing tablet mar­ket in gen­eral still remains very much in ques­tion but that was not the most impor­tant issue at play in this week’s announce­ments. Intel has already been writ­ten off in the low mar­gin, low power, high vol­ume, inter­net con­nected device mar­ket. Despite being demon­stra­bly unable to curb the ARM architecture’s grow­ing dom­i­nance of this mar­ket this is not — I sus­pect — what keeps Intel’s senior lead­er­ship team up at night.

The very prof­itable end of Intel’s busi­ness comes from another domain alto­gether; servers. The inex­orable growth in demand for more and more server com­put­ing capac­ity enables Intel to charge a huge and prof­itable pre­mium for their high end — server tar­geted — proces­sors. For the last 10 years, Intel have had almost zero com­pe­ti­tion in this mar­ket killing off com­peti­tors — like Sun’s Sparc and IBM Power series — by lever­ag­ing very com­pelling vol­ume eco­nom­ics and — most impor­tantly — by rid­ing Microsoft’s growth in the enter­prise soft­ware business.

The scale eco­nom­ics of the Win­Tel plat­form — Intel proces­sors and Microsoft’s Win­dows Server stack — has dec­i­mated com­pe­ti­tion and estab­lished the dom­i­nant regime to be found in most data-centers around the world today. IDC’s server num­bers — the most respected for this type of data –make the point. In Decem­ber last year IDC pub­lished their third quar­ter server mar­ket share num­bers — Intel servers run­ning Microsoft’s Win­dows Server oper­at­ing sys­tem accounted for 47.7% of over­all quar­terly shipped fac­tory rev­enue. The near­est com­peti­tor was Linux based Intel servers with 17.5% of over­all revenue.

What seems to have been lost in this week’s Microsoft and ARM announce­ment is the impli­ca­tions for the server busi­ness. As this BBC report points out, ARM have been qui­etly work­ing away at an opti­mized proces­sor design for server appli­ca­tions for the last last eigh­teen months. You can bet that the Win­dows Sever team is now very engaged in that devel­op­ment process. In my view, we’re about to wit­ness a fun­da­men­tal trans­for­ma­tion of the server busi­ness and it looks like Microsoft has just lined itself up to be the prime ben­e­fi­ciary. As ARM boss War­ren East points out in the BBC report “They [Microsoft] are ambi­tious to put that oper­at­ing sys­tem every­where … It works for both of us.”

Over the next ten years we will wit­ness a mas­sive relo­ca­tion of com­pute capac­ity out of pri­vate cor­po­rate data-centers into high den­sity cloud com­pute facil­i­ties. In ten years time it is very unlikely that any CIO will want to carry the cost, risk and com­plex­ity of run­ning their own server farms. Cloud com­put­ing will become the util­ity IT ser­vice provider of choice. As this trans­for­ma­tion moves for­ward old Intel servers will be retired from tra­di­tional data cen­ters while a new gen­er­a­tion of low power, highly effi­cient — ARM based — servers will come online in the cloud and — in Microsoft’s view of the world — the major­ity of them run­ning some ver­sion of a Microsoft oper­at­ing sys­tem — Microsoft’s Azure Cloud stack is after-all just a tuned ver­sion of Win­dows Server with some very clever and inno­v­a­tive man­age­ment soft­ware run­ning on top.

These mar­ket dynam­ics com­bine to become Intel’s worst night­mare; a clas­sic mar­ket squeeze between mul­ti­ple ultra-high vol­ume, low mar­gin providers of ARM based proces­sors at the low end and a com­pelling ARM based value propo­si­tion — based on effi­ciency and low oper­at­ing costs on the very high end. Intel ends up with what is left of the so-so tra­di­tional PC mid­dle mar­ket. Even that gets increas­ingly can­ni­bal­ized by ever more pow­er­ful and capa­ble low end devices. Ulti­mately Intel nei­ther com­mands the vol­umes or the high-end mar­gins required to sus­tain the R&D expen­di­tures needed to stay ahead and the inevitable down­ward spi­ral becomes inescapable.

Of course Intel’s man­age­ment is far from being unaware of these strate­gic chal­lenges and with their for­mi­da­ble resources you can expect them try their hard­est to com­pete and inno­vate their way around this chal­lenge. They need to because the very future of Intel will depend on it.