Chips are one of those things where being the best is simultaneously very important and not at all important. Making a 10% better chip gets you the entire market. But it makes practically no difference if your cellphone or laptop or server or whatever has a 10% worse chip.
> Making a 10% better chip gets you the entire market
Gets you the entire datacenter market maybe. End user (PCs, cellphones etc) stuff is much more concerned about perf/$ (up front cost) than perf/watt (long term cost), and the embedded market (electronics, appliances etc) mostly care about 'good enough' as cheaply as possible - performance isn't a concern at all for many use cases.
And the corporate market mostly cares about (perceived) reliability/liability concerns over everything else - see how hard it's been for AMD cpus to penetrate despite being measurably better in every category compared to Intel at various points in time.
