back to article Beijing-backed server chip startup formed by ex-Arm China execs

China could prove problematic for Arm once more, amid claims key staff from its local subsidiary have left to form a server chip design biz with government backing, and are eyeing up ex-colleagues to help. Arm China is 49 percent owned by private equity investor Softbank, which still owns most of Arm in the rest of the world …

  1. cornetman Silver badge

    Honestly, I'm not sure what Biden expected to happen but what is playing out now in China is exactly what anyone with a few brain cells to rub together would have guessed.

    It was easier for Chinese companies to import from the west, but with supply cut off, they are going to rapidly grow their own capability and sell it back to us. They were starting to do that already. These actions are just accelerating the process.

    1. Anonymous Coward
      Anonymous Coward

      Here we go again...

      The usual refrain: "Sanctions are hurting the sanctioner more than the sanctioned".

      We've heard that for Cuba, Venezuela, Iran, North Korea, and now... Russia and China.

      I wonder why I'm a bit skeptical... and whether there could be some disingenuous agenda behind such a theory. Really, I'm wondering.

      1. cornetman Silver badge

        Re: Here we go again...

        > "Sanctions are hurting the sanctioner more than the sanctioned"

        Not really sure how you read that sense from my comment. These sanctions are clearly hurting China, but we cannot expect them to just lay down and take it.

        In a few years time, the US will find it self competing directly with China home-grown tech, precisely because the sanctions will make them prioritise their own development of it.

  2. Anonymous Coward
    Anonymous Coward

    It's an, um... arm race

    Predictably, we're going to be served the usual soup of "sanctions-are-the-best-way-to-make-sure-China-will-catch-up-and-overtake-the-West" stories.

    Granted, China will have no other solution than to build their own semi-conductor industry.

    But honestly...

    1. Where they not doing that already? With very little success, one has to admit.

    2. Does anybody believe that free countries will stay idle and wait for China to catch up? And then marvel at the miracle? Surely not! Some of the brightest brains, supported by powerful investments are at work, across all the many disciplines involved in the semiconductor industry. Something to reckon with.

    3. Improving feature size is a dead end already. Optoelectronics is the new game. In this domain, front runners mostly dwell East of the Pacific.

    The pattern is: people take measures to solve their problems. Till they're faced with more problems, and then, they take measures again.

    That's why we still develop drugs and vaccines against pathogens. Granted, pathogens evolve mechanically to escape these drugs as well as our anticipated immune system responses, but, in the meantime, we develop new drugs and new vaccines, tailored to the evolved strains. It's an arm race, it has always been one.

    Sorry for having to state the obvious, but addressing issues is more beneficial than ignoring them.

    1. Anonymous Coward
      Anonymous Coward

      Re: It's an, um... arm race

      > China will have no other solution than to build their own semi-conductor industry...Where they not doing that already? With very little success

      Let's see, shortly following the great famine in 1959 to 1961, the Chinese Cultural Revolution, from 1966 to 1976, aimed to purge not only traditional Chinese society but also Capitalist tendencies and ended up killing or driving into refuge (failing that, onto forced labour farms) scientists, scholars and "the elite". Universities and schools were closed - if not ransacked.

      Now, what state was the semiconductor industry in The West by 1976?

      By 1978 we had the BBC Horizon "Now the chips are down" examining the potential of ICs, and not just as a force for good. 1980 onwards and "soon, every home will have their own computer" (not quite, but we all knew someone who did).

      Now, even taking into account that China acquired tech by dubious means during that period, consider the rate of progress that they have made to fill the gap from 1976 to 2023 - not even the duration of a single working lifetime - and you are saying that they have had "very little success" in building up their semiconductor industry?

      This difference in pace is why the US is reacting the way it is (whether that is sensible or not is a matter for a different comment) - you should not be so casually gung-ho in dismissing it.

      > Sorry for having to state the obvious, but addressing issues is more beneficial than ignoring them.

      Ditto.

      1. Anonymous Coward
        Anonymous Coward

        Progress is not linear

        It's much easier to catch up with the pack when you're far behind than to become a market leader.

        Where is Chinese x86 production (Zhaoxin and Hygon)?

        They might be good enough for low end domestic servers, but in no way a threat to Intel/AMD market dominance.

        1. Anonymous Coward
          Anonymous Coward

          Re: Progress is not linear

          > Where is Chinese x86 production

          Is there any reason the Chinese would care about x86 production?

          It is only of interest to The West because of the dominance of Windows. The interesting stuff is in Arm (and soon enough, RISC-V) in handhelds and for servers. The higher-tech stuff is GPUs, not x86.

          > in no way a threat to Intel/AMD market dominance.

          Again, do the Chinese actually have any reason to care about that? As you point out, they have Zhaoxin and Hygon, who seem quite capable of supplying basic business needs, the section of their home market that is running Windows. If they can get Intel devices off the back of a lorry, they'll flog them, of course (perhaps with a bit of naughty rebranding - we saw you, Powerleader).

          And they are working to get further away from x86 - for example, the Longsoon CPU - which isn't up to the grunt levels of current Intel/AMD but, again, good enough for basic business use.

          Sure, they would love it if their CPU designs started to make inroads into Intel/AMD markets - that would be a coup and drive the US nuts - but, just as the US is loathe to use Chinese tech, the Chinese would be glad to be shot of US dominated tech in their domestic market.

          1. Anonymous Coward
            Anonymous Coward

            Re: Progress is not linear

            Ever visited a Datacentre?

            So According to you, all these Linux servers run on ARM?

            Claiming that x86 is only needed for Windows is simply laughable.

            1. Anonymous Coward
              Anonymous Coward

              Re: Progress is not linear

              Sigh.

              What is relevant is not what is inside an existing datacentre (the CPUs for those have already been purchased, duh) but what is going to go into next year's centres - *that* is what creates the market now! Even American datacentres are building with Arm for reduced power costs. Let alone those in countries who want to be rid of the US hegemony.

              You mention Linux (btw, just one capital letter - you may want to bone up on your tech terminology) - you do know that that is famously portable to other CPUs?

              You do not *need* to build a datacentre with x86, but you do *need* x86 for Windows (as their moves away have failed - oh, you do know that even Windows has tried to free itself from x86, yes?)

              1. Anonymous Coward
                Anonymous Coward

                Re: Progress is not linear

                Oh yeah, right. LOL.

                The folks who can't even compete in the hyper-commoditised x86/x64 markets are suddenly going to turn into ARM foundry experts. LOL.

                Also, please note, that most Linux data centre servers in the world, including USA, are still x86 based. ARM is only relevant in some newer HCP data centres with custom infra. Ask Red Hat. They must have an idea.

                So, sorry to tell you, but you're mixing things up, the headlines and the on-the-ground reality. When you're wrong in the first post, the best attitude is to admit it, learn, and move on - as opposed to doubling down with some pathetic nonsense only denoting users of unregistered Chinese copies of Windows XP.

                The point is: if you can't develop a profitable mainstream x86 PL, forget about competing in the high end semiconductor arena.

                That said, I can agree with you that the market is rife with low quality Chinese imitations of mainstream chips.

                As discussed here: https://www.youtube.com/watch?v=7epnv43jGV8

                PS: I started spelling "Linux" and other unix's probably before you were even born. So, thanks for the nitpick. That's all you will get right in that thread.

                1. Anonymous Coward
                  Anonymous Coward

                  Re: Progress is not linear

                  > The folks who can't even compete in the hyper-commoditised x86/x64 markets are suddenly going to turn into ARM foundry experts. LOL.

                  Sigh once more. You are fetishising competition and ignoring just getting away from the US suppliers. The West cares about competing and winning, at the moment the Chinese are more concerned with not needing to give a damn about what the US is doing - to get away from any damage that US restrictions are causing.

                  > The point is: if you can't develop a profitable mainstream x86 PL, forget about competing in the high end semiconductor arena.

                  Again, their point is not to compete in any arena than their home market (over which the central government have more control than the equivalent in the West. Their point is to just to be able to tell the US to get knotted. If - note, if - they then manage to get a foothold outside their home market and start "competing in the " then that would just be gravy.

                  > Also, please note, that most Linux data centre servers in the world, including USA, are still x86 based.

                  Once again, what the current DC are composed of is not as important as what the next lot *can* be composed of. Any existing DC is only in the market for replacements to failed units or a little bit of expansion. A piddling piece of the upcoming market, compared to the chance to outfit new DCs from scratch.

                  > I started spelling "Linux" and other unix's probably before you were even born.

                  Ah, you naive youngsters. Unix didn't even exist when I started in this game, let alone that upstart Linux.

                  1. Anonymous Coward
                    Anonymous Coward

                    tl;dr repeating himself.

                    Unix started on a PDP-7 in 1969 in Bell Labs.

                    So if you really started working before 1969, you should be at least in your seventies. LMAO.

                    ...or making things up.

                    1. Anonymous Coward
                      Anonymous Coward

                      Re: tl;dr repeating himself.

                      You do know that people can live into their seventies? And have spare time to waste arguing with people?

                      Oh, and you may not know this one either, but if you are halfway good at what you do, like tech, you can get gainful employment in your early teens? You could even get apprenticeships in most fields (although not tech, as guilds hadn't caught up with it): real life started early back then.

                      By the way, using LOL and LMAO *really* does not help you sound like a professional whose word is to be trusted.

                      Bored now.

                      1. Anonymous Coward
                        Anonymous Coward

                        ROTFL

                        A the time of IBM, Burroughs, Sperry, Univac and Control Data, IT was mainly APL, Fortran or Cobol (please feel free to correct casing, lol).

                        If you were a fresher engineer, you'd still have 4 or 5 years of math such as numerical analysis of finite element analysis specialisation behind you. So you were at least in your early twenties. Not younger.

                        If you were 18y old and just employed in IT in 1966, then you were mounting tapes, feeding printers with stacks of dot matrix paper, or shuffling decks of punched cards around. At best.

                        So, next time, are you going to claim you went to school with Noam Chomsky? LMAO.

  3. Kevin McMurtrie Silver badge

    I see Softbank mentioned and know that a good disaster story follows.

  4. t245t Silver badge
    Linux

    Beijing-backed server chip startup formed by ex-Arm China execs

    Who knew /s

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like