The Kirin 980 SoC uses the Cortex-A76 which was designed in Austin, TX, United States of America.

The statement "The U.S. still makes — or at least, designs — the best computer chips in the world" is accurate in the context of a manufacturer attempting to build a high-performance microprocessor-based device.

Exactly, thus far China doesn't have the technology to do sub-14nm foundry work at any notable scale, and as we saw with Dhyana, they apparently don't have the internal talent to improve AMD's Zen cores. The foundry issue isn't apt to change anytime soon, as newer nodes are extremely expensive and take years to bring up (see Global Foundries & Intel's troubles with newer processes).

Minor nitpick: As highlighted in the article, power consumption is on par with A12, not efficiency. A12 still (somehow!) leads by ~50-100% there.

So, you think all phones and wireless base stations and telecom equipment ever need is just some mobile phone SoC?

This narative is so so similar to Chinese netizens'，I couldn't help but cringe when I saw so uninformed and forced arguments, this myth has to stop.

The US still monopolizes most telecom related chips and components, Huawei does make a few other chips along with some Chinese companies, but they are generally low-end in performance, couldn't compete at high-end market.

And let's be realistic, free of those marketing frills, Huawei's chips are impressive but still modified ARM products, and they buy a lot other licensed IPs for their chips.

Ditto for Allwinner and other Chinese ARM licensees. They buy the majority of IP blocks from ARM and perhaps a HDMI PHY & WiFi chip from some small company (also not Chinese) cause its cheaper, send the mostly vanilla design to TSMC, and when it doesn't work immediately run back to ARM asking them to fix it (which they do).

Apple & Qualcomm comparatively tweak or redesign most of the IP they license from ARM, which is what gives them such a significant performance advantage over the 300 other ARM licensees.

Another example of this is AMD's recent JV in China, essentially passing a holding company the IP to build an original Zen core. While this is sketchy AF, I think we can expect brutal embargoes if China attempts to sell these chips outside their domestic market, and despite having a year and a half to improve Zen, it appears Dhyana is nearly identical to Zen.

Meanwhile Zen 2 has already started sampling in useful quantities on a newer process node that Chinese foundries likely won't replicate anytime soon, seeing as Global Foundries gave up, Intel is running years behind and massively over budget, with TSMC having the only working sub-14nm process. Even if AMD licenses this architecture, it will take years (even with process theft from TSMC) to make those chips on 7nm in China.

Excellent post but once China subsumes Taiwan - and that day is coming - they'll have TSMC's technology and that much more electronics production monopolistic leverage.

China won't be subsuming Taiwan anytime soon. If China did subsume Taiwan, TSMC would lose some of their process engineers, and this is apt to have disastrous consequences for their sub-14nm nodes as they are so finicky to get & keep working reliably.

The population replacement in Hongkong that's at its another height, will inevitably occur in Taiwan if that starts to be a remote reality, albeit at an even larger scale.

China doesn't have exclusive control of the borders with Taiwan, for population replacement (or just flooding Taiwan with loyal Chinese citizens to alter the demographics) it would require the Taiwanese government to look the other way, or encourage it. Neither is likely to happen, as their government isn't asleep at the helm.

I feel the need to suggest the USA is not making chips. It are corporations doing it. The technology is, as if high priests guarding the sacret knowledge, kept secret. From the perspective of human progress it's a sad afair.

This is a pretty a uncompellling example.

Without citing Huawei specifically, consider some government backed company with tons of cash no practically no IP.

I bet we could go from you and me in a garage, to a pretty high performance ARM compatible SOC in not that many years.

How? First think of how far you can get just by buying everything ARM has to offer including licensing all IP options and technical consulting service options. Instant prettty decent reference implementation.

In parallel we’ve hired 1,000 engineers and scientists to start working on improvements and novel ideas, let by a team on the the best architects on the planet we can woo away by offering each 7 or 8 figure contracts, unlimited first class air travel for family and friends, whatever is necessary based on strategic value.

Competently managed that would provide high probability of a pretty fast ARM chip.

The problem? You can’t win and sustain competitive advantage with just that. It’s performance balanced with so many other factors including power, effificency, overall effecrivenesss of your SOC for given applications etc.

Besides the enormous challenge of all things adding up at once to win, there’s also no guarantee we’d end up with any especially huge amount of novell IP, not just stuff licensed from someone else, that would allow this win advantage to be comfortably sustained for any susbstantial period of time.

They have a lot of smart people doing some interesting stuff, but that particular example I wouldn’t consider very threatening, even if other examples do exist the better make the point.

Plus they said the same about Japan and Korean electronics and cars, and look how that turned out...