Qualcomm to Acquire Nuvia: A CPU Magnitude Shift

(anandtech.com)

163 points | by pella 114 days ago

15 comments

  • nirv 114 days ago
    Quite sad news, to be honest.

    I hoped that Nuvia, given its great team of ex-Apple CPU architect engineers, would maintain its independence and become a full-fledged competitor in the desktop and server market. Instead, they sold to a company that is often described as "the Oracle of hardware companies" and "a law firm with a few engineers"[1][2][3].

    [1] https://news.ycombinator.com/item?id=18333270

    [2] https://www.bloomberg.com/news/articles/2019-05-22/qualcomm-...

    [3] https://old.reddit.com/r/Android/comments/71rjyx/why_exynos_...

    • robotresearcher 113 days ago
      Viterbi was a founder. He is an historically significant engineer.

      https://en.wikipedia.org/wiki/Andrew_Viterbi

      https://en.wikipedia.org/wiki/Viterbi_algorithm

      • desultir 113 days ago
        Qualcomm once had 7 engineers and zero lawyers. They hire lawyers much faster than they hire engineers
        • dis-sys 113 days ago
          and every lawyer in QCOM is at least treated like a Senior Director.
      • wyldfire 114 days ago
        > maintain its independence and become a full-fledged competitor in the desktop and server market.

        If I had to guess, Nuvia's expertise will be used to optimize Qualcomm's ARM, Adreno and Hexagons for the next generation of Snapdragon cx. Qualcomm and Microsoft dipped their toe into this market and now that Apple is all-in, they need an answer to the M1. They won't get there with reference designs from ARM.

        But once they're producing optimized ARMs for SoC apps processors, who's to say that they couldn't re-enter the server space?

        • singhkays 113 days ago
          > They won't get there with reference designs from ARM.

          This nails it! I blogged about how the expected Geekbench scores for Nuvia were better than M1

          - https://singhkays.com/blog/arm-cpu-faster-than-apple-m1/

          Nuvia was aiming for the server space but Qualcomm probably saw an opportunity to leapfrog many generations of in-house design and catch up to A and M series chips from Apple.

          For reference - Qualcomm's upcoming Snapdragon 888 flagship lacks behind 2019 Apple A13

          - https://images.anandtech.com/graphs/graph16325/116944.png

          • com2kid 113 days ago
            As an aside, on mobile your site has ads on top and bottom of the browser window, leaving a bit of room in the middle to read the content which is just as often ads as your own words.

            Reading that article was a distinctly unpleasant experience.

            • singhkays 113 days ago
              Sorry about that! That does sound like a horrible experience :( Let me look into it
        • oblio 113 days ago
          > Instead, they sold to a company that is often described as "the Oracle of hardware companies" and "a law firm with a few engineers"

          Didn't Qualcomm beat out a ton of competitors?

          • ohazi 113 days ago
            Yes, using patents.

            They got lucky and a handful of their early patents became required parts of early cellular standards. Since then they've lobbied the standards bodies to include features that require their newest patents. When that fails, because the mobile manufacturers are tired of Qualcomm's monopoly, they simply buy up more patents to make up the difference.

            • ksec 110 days ago
              That is just so far from truth. The reality is no one manage to compete with them even when 5G had significant tech not derived from Qualcomm. And yet they still had the best Modem.
            • baybal2 113 days ago
              Yet, very remarkable given that they were nearly stomped out of the market around 2010-2012 by previously "no-name" SoC makers, making cookie cutter SoCs cheap, fast, and selling by tons.

              Yet, they recovered, and clawed back their way to the top, with only MediaTek now threatening them.

              I don't expect MediaTek to be as inept, and Pavlovian if Qualcomm will ever come to them with a deal, as they did with Allwinner 8 years ago (Allwinner gave up on mobile market, and Qualcomm gave them an obscene cut from their low-end Snapdragons in return.)

              And knowing Taiwanese, they simply don't sell companies owners spent their life working on. Very ego driven business culture.

              • robert_foss 113 days ago
                Mediatek has 31% market share of the smartphone SOC market, and Qualcomm 29%.

                https://www.counterpointresearch.com/mediatek-biggest-smartp...

                • phonon 113 days ago
                  I think that must be in units, not dollars. Qualcomm chipsets are nearly universal on the higher end.
                  • robert_foss 113 days ago
                    Not having looked at it at all, I think so too. But the trajectory of them overtaking Qcom is interesting either way.
                • rektide 112 days ago
                  this would help explain why all winner was on a roll making pretty great chips and has shipped absolutely worthless tripe for! he past 5 years. re-spins of rhe same shit they had been making. the chip making game feels so diseased.
              • blinkingled 114 days ago
                But on the other hand they didn't have many other viable options did they? Their expertise is ARM and they need a BigCo to back their R&D and subsequently manufacturing. So that leaves them with Nvidia and Samsung as options.
                • wmf 114 days ago
                  No, I think they could survive independently on VC funding.
                  • blinkingled 113 days ago
                    Having read HN I got the impression that VC funding isn't always a good thing either. How is VC funding at this level better? (It might well be I just don't have any insight.)
                    • wmf 113 days ago
                      I don't know which route is better (clearly the Nuvia team decided to accept the acquisition) but I think independence would have been possible.
                • ChuckMcM 113 days ago
                  I feel the same. It is always hard to make these sorts of acquisitions work because the in house team that you're augmenting/replacing always interprets this move as a criticism of their abilities. I've seen this lead to some pretty dysfunctional (and destructive) behavior.
                  • newusertoday 113 days ago
                    except in this case there is no internal team as qc switched to stock arm cores, so they would actually be replacing arm which is acquired by nvidia and given that soc design's lead time qc might not want to disclose its future soc plans to arm.
                  • devwastaken 113 days ago
                    With how similar processors are can you even have a small company without being threatened by patents at every corner? Part of the capitalism endgame isn't innovation, but to buy up potential competitors by having grey area ownership of the mechanisms they work with. Sometimes letting those companies then use your patents to continue work.
                  • ksec 114 days ago
                    Replying to distract myself from the Intel news.

                    Nuvia has shown [1] they have designs with Pref Per Watts that rivals or exceed Apple's current A14. Normally these sort of pre-design and slides should be taken with a big pinch of salt from Startup. But this is from Gerard Williams, ARM Fellow and Architect of all current Apple CPU design. So i think they are plausible.

                    The improvement in Single Thread Performance is what Qualcomm desperately needs for their SnapDragon SoC. It will also reboot their ARM on Server work now that Apple and Amazon has the ball rolling.

                    Tremendous respect for CEO Steven Mollenkopf, retiring later this summer.

                    [1] https://www.anandtech.com/show/15967/nuvia-phoenix-targets-5...

                    • BugsJustFindMe 113 days ago
                      > Nuvia has shown [1] they have designs with Pref Per Watts that rivals or exceed Apple's current A14.

                      Your linked article 1) doesn't mention A14, and 2) only contains a marketing image for something that didn't yet exist. It doesn't appear like they _showed_ anything. Did you mean to link to a different article?

                      • IanCutress 113 days ago
                        The article was published pre-A14. The A14 results can be found elsewhere on the same website.
                        • BugsJustFindMe 113 days ago
                          And can Nuvia results be found anywhere, or is it just a marketing graphic?
                    • arusahni 114 days ago
                      I initially read this as "NVidia". I'm still recovering from the shock.
                      • jannes 114 days ago
                        Happened to me too. I was already thinking about Qualcomm suddenly owning ARM (through Nvidia) and how anti-competitive that would be.
                        • shroom 114 days ago
                          Think everyone did :D
                        • m463 113 days ago
                          "all new phones will have 12-pin charging connectors" :)
                      • jabl 114 days ago
                        That's kind of interesting considering Qualcomm spent plenty-o-millions not many years ago to develop the ARM64 Falcor core and Centriq server SoC, only to throw it all away just before it was supposed to hit the market.
                        • klelatti 114 days ago
                          That was in response to the Broadcom bid I think? Corporate vandalism.
                          • wyldfire 114 days ago
                            It is a shame but I think they figured they had to go on a diet or risk an LBO.
                            • robocat 113 days ago
                              Is this a common scenario? Bet all capital on a new product, and get bought out at the low point in market valuation: before the product has shown promise and the market price has still not recovered?

                              Edit: I am wondering about older companies more than startups (presumably it is more common for startups that look like failing but are actually on brink of success).

                              • wyldfire 113 days ago
                                I don't think it's common, no. In the 1980s I think it happened more often (because they had a lot of assets that weren't reflected in the valuation maybe?).

                                But in Qualcomm's case they were arguably undervalued because of ongoing regulation/lawsuits regarding their licensing practices. IIRC they had shipped their Amberwing Centriq CPU but the reception was kinda lukewarm.

                                AWS ended up introducing Graviton not long after - presumably AWS wouldn't have sourced Centriq if they had their own designs in the works.

                                [1] https://en.wikipedia.org/wiki/Qualcomm_Centriq

                            • imtringued 113 days ago
                              That's how every ARM server project ends. This is not going to be different.
                            • Iolaum 114 days ago
                              Do I remember wrong that Nuvia higher ups have stated that they wanted to bring products to market rather than get acquired? P.S. I know such statements shouldn't be taken at face value anyway. Just curious.
                              • avrionov 113 days ago
                                > NUVIA was originally founded in February 2019

                                Do they hold the record for the fastest and most profitable exit?

                                • zyang 113 days ago
                                  This is deja vu of the Uber - Otto aquisition.
                                  • wmf 113 days ago
                                    Netscape did a $2.9B IPO after 17 months.
                                    • avrionov 113 days ago
                                      Adjusted for inflation Netscape price is much higher, but at least they had a working product and big team.
                                  • liminal 113 days ago
                                    Microsoft must be kicking themselves for not acquiring them
                                    • modeless 114 days ago
                                      I hope this means they are at least going to try to catch Apple. The complete lack of effort by them on the CPU side has been sad. Given hardware timelines, though, I'm sure any improvement is many years away.
                                      • tambourine_man 113 days ago
                                        What are the chances of lawsuits in the near future?

                                        Not my area of expertise by a long shot, but Qualcomm and Apple are not strangers in court and acquiring a slew of ex-Apple CPU engineers is probably like trowing gas on that fire.

                                      • rektide 114 days ago
                                        $1.4 billion. Wow. That's a lot of dough.

                                        I feel like 202x might be another decade where no new chip making companies emerge. I had great hopes for Nuvia becoming one of them. My first most primitive reaction is that this is like another PA Semi[1]: great talent & new prospects, fresh air, vanishing into one of the super-massive existing players.

                                        Qualcomm does some open source work, but most of the support comes from outside engineers reverse engineering their products or otherwise doing the work[2]. The Freedreno graphics driver for example allows their gpu to be used (OpenGL, Vulkan drivers), & has been a huge effort. That work mostly comes from Google engineers like Rob Clark (who started Freedreno iirc), Eric Anholt, Igalia engineers like Danil, & countless others[3].

                                        In the wifi-router world, where Qualcomm also has a huge presence, we rely on hunting around for open-source code drops of what usually seems like pretty old but better than we've got code, such as this recent find[4], which is hopefully going to help us catch a couple years up in support level to what Qualcomm is doing today.

                                        Nuvia explicitly stated they wanted to chase hyperscaler market. And I think a lot of this was positioning. That hyper-scalers wanted good tech, would use good tech. Where-as a lot of the world, they want what they have but better, a lot of legacy concerns, hand holding. Nuvia wanted to build a new platform, a good, well supported platform, with next-gen technologies. We were all expecting open-source perhaps Rust-based firmwares & management systems, pursuant to latest OpenCompute Project specs &c. They were clear about targetting hyper-scalers, & big chips, but somehow the hope always was that they'd help reform & reshape computing & what was available to everyone. They certainly talked the good game, sounded real good on Twitter. Ah well.

                                        [1] https://venturebeat.com/2012/09/18/more-details-shake-loose-...

                                        [2] https://www.phoronix.com/scan.php?page=search&q=Qualcomm

                                        [3] https://github.com/mesa3d/mesa/commits/master/src/freedreno

                                        [4] https://forum.openwrt.org/t/adding-openwrt-support-for-xiaom...

                                        • gigatexal 113 days ago
                                          Qualcomm CPUs for both mobile and non-mobile should be even more compelling in the near future with this acquisition. Apple has a legit competitor here.
                                          • solarkraft 114 days ago
                                            Is it true that this will uniquely position them to compete with Apple's processors while everyone keeps having to use the bog standard (and apparently not that amazing) Cortex cores?

                                            Are there other companies able to produce promising cores/any that could develop a serious M1 competitor?

                                            Was this acquisition mostly fueled by M1 hype (i.e. "Cortext chips are just fine")?

                                            • WatchDog 113 days ago
                                              1.4B, not bad for 2 years of work.
                                              • person_of_color 113 days ago
                                                What sort of exit would a staff engineer get here?
                                                • bitcoinmoney 113 days ago
                                                  Seems like it is a an acquihire. I know somebody who said he gets a modest amount. Seems like they sold out too quickly? Series C or D would be better. But they just had A and sold out without a product?
                                                  • _chris_ 113 days ago
                                                    But if you sell out after Series A you should have a ton of equity still left with little dilution?

                                                    In this case, it looks like they did a $50M Series A and a $240M Series B. IIRC, 5x is respectable, but I can imagine the VCs might be on multipliers here and eat away quite a bit of the common stock.

                                                    • person_of_color 113 days ago
                                                      Did they even tape out?

                                                      They probably got equivalent to staff engineer rsu refresher.