You're looking at the PR, look at the actual technology.
Anyone can build a big enough supercomputer if you throw enough money at the problem. This high level of spending is getting hard to justify, however, because most of the problems you want to solve can't be solved just by throwing another order of magnitude of computing power at the problem. Building a supercomputer becomes a bit of a boondoggle, a public-relations move.
But China can't build the components for those supercomputers. Nobody can build all of them, but China can't fab the processors. Search for "list of semiconductor fabs", find a table, sort by node, and then scroll down until you find a Chinese node that does something other than just churn out flash memory. You'll find SMIC with a couple 40nm or 45nm fabs. That's not just the last generation of tech, that's three generations old (14 [current] => 22 => 32 => 45).
By comparison, the market is flooded with cheap 32nm Xeons being dumped from our data centers.
This. China built that supercomputer as a pure PR/face move. It isn't very interesting otherwise, and the west stopped building those kinds of things a long time ago because it moved on to more detached distributed computing.
You need to look deeper into the latest system. It's not a pr machine. Even the pervious one tianhe-2 isn't. If you think it is, you're looking at the wrong sources.
And no, we have not moved away from supercomputers or building bigger ones.
SMIC is down to 28nm now[1]. I agree this isn't competitive yet, but they are working on 14nm[2] with some reasonably credible partners (including Qualcomm), scheduled to be delivered by 2020.
That would close the gap to Intel/Samsung/TSMC to one generation.
It will be interesting. I'd call that about one and a half generations, if 10nm ships in 2017 like Intel says. Of course, it is anyone's guess whether SMIC will ship 14nm on time, or whether Intel will ship 10nm on time.
Comparing Taiwan, which has been functionally independent of the PRC and under control of the ROC since 1945, to South Ossetia or Transnistria is absolutely ridiculous. Just because Taiwan isn't widely recognized by the international community doesn't mean that it isn't an entirely separate socioeconomic entity. A brief glance at the last two decades of politics in Taiwan will show you as much.
The UN is totally irrelevant here. Taiwan operates as an independent country in every practical way. (So does Hong Kong, kinda, but Taiwan even more so.)
Anyone can build a big enough supercomputer if you throw enough money at the problem. This high level of spending is getting hard to justify, however, because most of the problems you want to solve can't be solved just by throwing another order of magnitude of computing power at the problem. Building a supercomputer becomes a bit of a boondoggle, a public-relations move.
But China can't build the components for those supercomputers. Nobody can build all of them, but China can't fab the processors. Search for "list of semiconductor fabs", find a table, sort by node, and then scroll down until you find a Chinese node that does something other than just churn out flash memory. You'll find SMIC with a couple 40nm or 45nm fabs. That's not just the last generation of tech, that's three generations old (14 [current] => 22 => 32 => 45).
By comparison, the market is flooded with cheap 32nm Xeons being dumped from our data centers.