Talk about apples and moonrocks
ARM confounded analysts during its fourth quarter growing sales by 19 per cent while chip giant Intel saw revenues down 3 per cent compared to a year ago.
Even with a slight drop Intel's profits are still enormous.
And then further down "Intel is throwing Atoms under ARM's wheels". Er, no. While Intel has indeed got Atom's to impressive performance/power levels so that the couple of phones with them compete favourably with ARM based stuff, the chips are still significantly more expensive than ARM chips and depend on Intel's better fabrication techniques.
Of course, ARM faces risks in the future. But so does Intel and, as Intel insists on being a manufacturer those risks are higher - sinking billions into a new fab is considerably riskier than trying out a new chip design. The rewards of success are also higher but those fabs just get more and more expensive to build.
Next to the fabless versus fab is also the difference in business culture. ARM's culture of co-operative competition amongst its customers and occasionally with them (Qualcomm, nVidia, PowerVR) and Intel's attempt to monopolise through instruction set. AMD only exists because IBM insisted that Intel licence a second supplier of x86. ARM's model is more in tune with component suppliers in other industries.
As for servers: there is now very little difference between Intel's CISC and ARM's RISC. Intel went RISCy with the Pentium and everyone has been adding instructions to their chips since MMX. ARM still gets more done with fewer transistors. Servers are likely to profit from exactly the same kind of commodification of components that has benefited Intel over the last 15 years. x86 has remorselessly gained market share from Power, SPARC, MIPS, etc. proving that new chips have a chance. This has meant vastly improved toolchains and compilers in the process from which ARM stands to benefit and the APUs of AMD and nVidia are starting to point the way: yes, a single instruction makes life easier for the compiler but if you can offload number crunching to GPUs then you will just get so much more bang for your buck. If ARM's big.Little architecture can demonstrate this kind of switching in real life then we can expect hybrid servers to really take off. x86 translation can be provided in hardware if necessary as the designs for the necessary silicon (Transmeta) already exist.
But it is also simply naive to put ARM up solely against Intel. Unless Intel goes fabless they are not competitors. Rather Intel is competing increasingly with nVidia, Marvell, Broadcom, Qualcomm, TI, Samsung, et al. and the competition will increase as the PC market fails.