I wouldn't go that far. Samsung and IBM are both quite talented. IBM designs and fabricates much higher clock speed processors than Intel, for example -- as high as 5.5 GHz currently.
Steve Jobs was annoyed that IBM and Motorola/Freescale couldn't always satisfy his processor desires. We thought he was correct in adopting a sole source Intel strategy (announced in 2005 and executed in 2006), but in retrospect maybe he erred. Maybe he should have retained Mac OS X's dual architecture support past 10.5 to allow processor supplier flexibility, with (as examples) a MacPro equipped with powerful IBM Power8 processors and a more powerful Freescale-based Apple TV with real game console capabilities. Nobody delivers perfectly in this business, and it wasn't all that long ago that companies consuming semiconductors insisted on at least two independent suppliers.
Hopefully Apple is still compiling OS X on Power CPUs internally to keep their options open. Apple cannot afford to be so thoroughly dependent on Intel.
That says nothing at all - and is in fact very misleading.
The Power8 is a server processor. That means is has very different design requirements. It is allowed to burn lots of power and cost the earth; Intel's chips have the exact opposite requirements.
The Power8 has 48/96MB of L3 cache depending on configuration. That is enormous! Haskell only has 2-8MB.
The Power8 has a die size of 362mm2 - 650mm2 (!!!). Haskell has between 177mm2 and 277mm2. Not only that, but it's estimated that about 65% of Haswell's die area is taken up by the GPU, which the Power8 obviously doesn't have. That means the largest (4-core) Haswell CPU takes up about 96mm2. You could almost fit 4 (i.e. 16 cores worth) of them on the die of the 6-core Power8.
Oh, and those IBM servers probably have some very sophisticated cooling system which exceeds the kind of cooling you'd typically want under your desk (or on your lap). For the record, you can overlock Haswell up to 4.8Ghz with air cooling that's still probably not as powerful as those servers have rocking.
These chips have completely different design requirements. It's like comparing a Bugatti Veyron to a VW Golf. The Bugatti will go faster, but while doing so it will also consume 12,000 gallons (45k litres) of air per minute and do 2.5mpg while burning through its $40,000 tyres in about half an hour. The VW is more modest, but for its size and relative economy delivers remarkable performance.
Actually, it's probably quite an apt comparison; the Power8 has 1700 integrated voltage regulators. Each Haswell has one (split in to 20 parts internally; still that's 85x less than the Power8). That's similar to the Veyron having 10 radiators.
But I didn'tt mention architecture; I said Intel has the best fabrication technology. The Power8 was announced around the same time as Haswell, but IBM itself have only been able to ship them since a month ago, on a 22nm process. Intel's been shipping cargo-ships full of 22nm Haswells for well over a year. Broadwell is being developed on a 14nm process which nobody else on Earth has been able to do at that scale so far. If we see chips at the end of the year, that's still better scale than anybody else has been able to achieve.
Building chips is really, really hard. Electronics-grade silicon is expensive, and all the processing that is done to it is also expensive. Materials behave completely differently at the nanoscale, and nano-scopic stress fractures and other utterly bizarre quantum forces can totally ruin your work. You can't conceivably test every function of every chip on the production line, so you need a very reliable and well-tested process to be able to get reliable, economical yields out. It's simple enough to demonstrate one 14nm chip once in a lab, but scaling that up to a production line is a whole other story.
Intel is better than anyone at doing that. As far as fabrication technology is concerned, they are the trailblazers of the silicon industry.
Intel's focus has shifted enormously from the Pentium 4 days: they're all about performance-per-watt now, not just performance. I've spoken to Intel engineers who have had really good ideas for improving performance but after rigorous energy analysis it was deemed that the power cost was too great. You don't need to worry about that when you're making enormous server chips.