Nvidia Acquires ARM For $40 Billion

Nvidia announced on Sunday evening that it has reached an agreement to acquire Arm Limited from SoftBank for a cool $40 billion.

In this age of headlines that use the b-word in place of nine zeros it’s easy to lose track, so you may be wondering, didn’t SoftBank just buy Arm? That was all the way back in July of 2016 to the tune of $32 billion. SoftBank is a holding company, so that deal didn’t ruffle any feathers, but this week’s move by Nvidia might.

Arm Limited is the company behind the ARM architecture, but they don’t actually produce the chips themselves, instead licensing them to other companies who pay a fee to use the core design and build their own chip around it. Nvidia licenses the ARM core for some of their chips, and with this deal they will be in a position to set terms for how their competitors may license the ARM core. The deal still needs regulatory approval so time will tell if this becomes a kink in the acquisition plan.

There’s a good chance that you’re reading this article on a device that contains an ARM processor because of its dominance in the smartphone and tablet market. Although less common in the laptop market, and nearly unheard of in the desktop market, the tide may be changing as Apple announced early in the summer that their Mac line will be moving to ARM.

Chances are you know the Nvidia name for their role as purveyors of fine graphics cards. They got a major boost as the world ramped up Bitcoin and other cryptocurrency mining hardware which early on was mainly based on the heavy lifting of graphics processors. But the company also has their eye on the ongoing wave of hardware targeting AI applications like computer vision. Nvidia’s line of Jetson boards, marketed for “next-generation autonomous machines”, all feature ARM cores.

Assuming the deal goes through without a hitch, what will be the fallout? Your guess is as good ours. There is certainly a conflict of interest in a company who competes in the ARM market owning the Arm. But it’s impossible to say what efforts they will make to firewall those parts of the business. Some might predict a mass exodus from the ARM ecosystem in favor of an open standard like RISC-V, but that is unlikely in the near-term. Momentum is difficult to overcome — look at how long it took ARM to climb that mountain and it was primarily the advent of a new mobile ecosystem lacking an established dominant player that let ARM thrive.

75 thoughts on “Nvidia Acquires ARM For $40 Billion

      1. The huge tech lead Apple has makes that pretty much impossible.

        NVIDIA still have to provide something competitive in the market. If non Apple devices had the performance crown maybe they could get away with bleeding it, but the non Apple mobile market is already bleeding in the first place … it needs saving at the moment.

        NVIDIA might be able to save non Apple mobile ARM… and then they can raise prices from a position of strength, which would be fine.

        1. Apple’s tech lead is from licenses to use Arm technology (ARM don’t design or make the chips for Apple – Apple do that for themselves). NVidia (ARM) will charge a license fee that Apple are “happy” to pay or Apple will use a different technology.

          1. Indeed, and usual license terms seem to include a number of chips per annum type fee structure – something that makes the license owner money for each device shipped. Which makes you wonder why Arm was for sale at all now Apple have publicly said everything they do will be ARM based… Much as I don’t like Apple products they are a license to print money, and sell in vast quantities – so the previous/current owners should be able to make a minor fortune from that which probably rivals the profit of the sale in very little time..

          2. Apple were there at the start of ARM, one of the three founding investors. I’m sure they have always been given a pretty sweet licensing deal. And they were also the very first customer using the chips in their Apple Newton PDA’s. I do not think that anyone else would have got the same deal as Apple.

          3. Apple almost certainly has a perpetual license.

            If so as long as Armv8 and the level of changes they can make to it while obeying the license stays viable they don’t need to pay anything beyond what they agreed to already.

    1. There was already ARM7TDMI architecture before in microcontrollers, it was pretty widespread by the time.

      Cortex-M, IMO hit the spot of ‘good compiler support’, ‘compact code’, ‘saneish architecuture’ and ‘standardised CMSIS’ (basically – interrupt controller and basic peripherials were portable between CPU makers – and that is a very big deal for microcontrollers. + it was cheap, compared to AVR. Cheap enough that it was possible to use it where previously only 8-bit AVRs were used.

  1. My biggest concern is will Nvidia keep ARM as a licensing company or will they try to create their own ARM cores and slowly fade out the licensing section by providing worse and worse support until they are the only ARM manufacturer. I think what helps ARM thrive is how many different people use ARM for all these different jobs, you see them in FPGAs, smart phones, hell you see them in pregnancy testers. I don’t think we as makers that use ARM chips benefit from having 1 guy make ARM and I hope Nvidia considers long term value of the license vs short term gains of killing it.

    1. I don’t think so because there is many options for MCU cores. MIPS core, RISC-V core and others. RISC-V in gaining traction so if NVIDIA play the game of killing ARM M core licencing they could kill ARM core as well.

      1. Agreed. RISC-V is already pretty well developed for deployment, last time I looked a few features that would be nice to have were lacking. But it should become the ARM/AMD64 ubiquitous processor almost immediately if the ability to use the existing designs is reduced.

        NVIDIA might play that game for the short term cash grab to keep the shareholders happy… but its a dumb move.

        1. Yeah my concern is not is it smart to do this it’s will Nvidia do it which you seem to agree is a possible/probable thing.

          With that said I won’t share the sentiment that RISC-V is ready for prime time. Just because they have a good system does not mean they have the backend infrastructure to support large industry projects. Without that they just have the hobby market. Currently I buy ARM cores for a few bucks at most, if RISC-V lacks the big bulk buyers I suspect (have not checked would have happy to be proven wrong), the MCUs will run me a lot more for comparible products. Additionally I lose all of the ARM libraries already developed, all the OS integration, basically any open source ARM project. Some can be migrated but who knows how many or how easily. ARM has a large force pushing it forward, Nvidia can crash it into a tree but it will keep going forward, for someone to pickup the scraps of those mistakes will require some serious market pull.

          Who knows maybe Samsung will start puting out RISC-V tablets but I just kind of doubt that. Or maybe all my smart spy fridges will come with RISC-V, not exactly a great market to buy into but whatever gets the ball rolling.

    1. Documentation?
      They can’t provide what they don’t have*. Pretty common for Si companies that do not provide broad market product (ie pretty much the same for qualcom I guess).

      * I worked 5 years at nvidia, and never seen a graphic chip datasheet (I was in networking).
      Same thing for intel, if you want the full datasheet it was often very difficult to have the full one (ie with every bit described) even internally.

    1. The RPI is a paying strategic member of the RISC-V cabal ( https://riscv.org/membership/members/ ). And they are also big enough now that they could get their own chips rolled at any of high end the fabs. But with their strong link to Broadcom (who are not a RISC-V member), they will probably just keep on buying whatever they are selling which currently will not include RISC-V based processors.

  2. As long as NVIDA do not also acquire a Japanese robotics and technology company by the name of Cyberdyne Incorporated and then rebrand itself to be called Cyberdyne Syspems Incorporated – and rise to be the dominant AI corporation. This is all fine.

      1. This makes me smile. I would love to see PowerPC make a comeback. Or whatever else may evolve out of the IBM POWER architecture once these big tyrants finish ruining everything they can get their filthy, glow-in-the-dark hands on.

        I hold hope for RISC-V too, but still a bit skeptical that it won’t suffer a similar fate.

      1. It sure will; arguably the current arrangement already does and has for many years now. As it stands, it’s about to kill a hell of a lot more than innovation. It’s going to kill our future. Big tech is not a power to be taken lightly. If they aren’t broken up, they will become a pseudo-government just like they are doing in many test-ground countries this very moment.

    1. 3Dfx tech went into the GeForce 5000 series. Those were built on the foundation of the last GPU design 3Dfx was working on. Unfortunately it wasn’t quite ready, and the ‘polish’ nVidia put on it wasn’t as glossy as it needed to be. The later refreshes of the 5×00 series combined with driver optimizing did manage to squeeze some better performance out of them.

  3. Libre (SoC) have actually pivoted away from RISC-V for many valid reasons (search for “NLNet Grants approved, Power ISA under consideration” if interested) to use IBM’s OpenPower. And now they are free to integrate some really cool tricks from Seymour Cray’s 1960’s CDC 6600 that drastically simplifies the implementation, reduces gate count, reduces power consumption, … They can do a lot, as long as the Libre implementations are fully-compliant with the OpenPower standards (In theory they could take on ARM, Intel, and AMD. The advanced technology was overlooked for nearly 60 years because the 1967 patent did not describe the key strategic parts).

  4. 40B$? 40000000000 freaking USD?? I am sorry but something is wrong with this world. Some people are jungling with 40B$ and others don’t even have 40$ each month to live with. Yeah.

    Anyway. Let’s keep politics out of here. I am just shocked by those numbers at each of these massive aquisitions (they were a few in semiconductor markets not so long ago, you may remember if you read HaD).

        1. NVDA market cap is insane, it is pure textbook bubble (from 12$ to 500$…)
          I don’t complain because I sell all my nvda stock long ago, not at 500$/each but still I was able to buy my house.

          1. Market Cap is, always has been, and always will be, a completely bogus number.

            For example, let’s say a stock finishes the the day at $2. And tomorrow, near the end of the day, because some guy had a hunch, bids for a 100-share lot at $4, and it ends at $4.

            Market Cap has doubled. Supposedly, this means that the company is in a better financial position.

            In truth, some guy bought 100 shares at a premium from some other guy. The financial health of the company which issued the stock hasn’t changed in the slightest.

    1. it’s not that insane.

      China and India both have an population of about 1.3 billion.

      If you earn 20 bucks from each person, you get 40 billion.
      Now take into an account for the rest of the world and how many arm devices are out there.
      Modems, routers, smart watch, cell phones, automotive, aerospace, communication, industrial and hobbyist parts.

      I can easily be worth that much.

  5. Just want to point out the big difference between offering just an ISA license vs. offering multiple levels of implementations of the ISA. Arm does both, as well as offering GPU implementations as well as other support packages for implementing a huge range of Arm processors into customer products. There’s a reason the price tag is very big.

  6. Frankly i don’t care for what architecture my C compiler compiles for, it could be a souped up 65816 for all i care. If nVidia decides to close up ARM or jack up prices to infinity we as makers will simply move to the next platform the folks in China adopt for our cheap micro-controllers. I doubt anyone cares what CPU core is inside the STM32 on a Bluepill as long as said Bluepill is cheap as sand and can be programmed with the GCC.

  7. A long time ago Commodore’s Jack Tramiel purchased MOS Technology so he could get the 6502s for his computers cheap. But, as ruthless as he was, Commodore still no had no issues selling 6502s to Apple, Atari, or anyone else, or licensing it to other companies like WDC for derivatives. On the other hand, it may be he was forced to be generous by the law or other agreements at the time. I have no idea whether a similar situation will take place with NVIDIA and ARM.

    1. The company Jack Tramiel wanted revenge on was Texas Instruments who almost forced him out of the calculator market. He succeeded too. What other computer companies used Texás TMS9900 processor?

  8. People need to look beyond nvidia as a gaming company. My 2 cents is that this is all about CUDA and CUDA-X.

    This probably won’t mean anything for mobile chips unless they can get CUDA working on Mali GPUs. Maybe Vulcan support will get better as aside effect.

    Instead, I’d expect nvidia to be pushing software into the ARM ecosystem that would enable them to survive getting locked out of the data center by Intel and AMD.

      1. Consider AMD owns ATI and with Ryzen doing well, perhaps Dr. Su will set the focus on useful APU or at least get ATI running well.
        If Intel can manage to not retreat back to just x86 in bad years like they have in the past, they may manage to make an ok GPU and might even find a way to improve things with their Altera acquisition from a few years back.

        So the foundation is there and both companies should be seen as hostile towards nvidia.

        Considering that you aren’t likely to find a Tegra in the data center, nvidia needs a way to proliferate CUDA. If they effectively offer CUDA support IP for free to ARM customers, it can create an install base bigger than their own chips. How they could do that is another question but we’ll just have to wait and see (while being ready to goad governments into enforcing anti-trust and anti-competitive behavior legislation).

    1. A super CUDA realtime, very high quality, HEVC encoding hardware chunk that incorporates all the features of HEVC that nVidia’s nVenc nVenc hardware currently does not?

      nVenc can produce some really nice video but since it lacks some features the files end up larger than the same video done with better optimization using (much slower) software compression.

      1. The NVenc block is generally a dedicated block in the HW. The whole point of the block(s) is to leave the CUDA cores available for the running application. It would make more sense for Nvidia to just update the NVenc units. They might use some tensor cores to try and use AI for better encoding but I have no idea if there are enough units left over after using them to make ray tracing look acceptable , performing a DLSS pass on each frame, running RTX voice, and running AI green screen. This assumes game devs don’t decide to try to implement some of the other interesting AI algorithms like the one for generating animation data on the fly, or use it for enemy AI in games.

        Concerning CUDA, I’d recommend watching the 2020 GTC keynote and watch for the the things built on it and CUDA-X. One example is Magnum IO which you don’t have to try too hard to read between the lines in order to see how Mellanox factors into that. It also sounds like a subset of that has been leveraged for RTX IO announced with the RTX 3000 series.

        So while some of their data center, AI, and scientific computing tech is showing up in their game hardware, the company is trying really hard to be a LOT more than just a gaming graphics company.

  9. Will be interesting to see the outcome.
    ARM do not produce silicon. Apple engineers take ARM designs and add their flair to produce silicon (outsourced).
    If NVIDIA can combine to good parts of ARM engineers with their GPU etc knowhow, they could produce a fantastic AI alternative for the data centre. This could just produce a viable alternative to the intel/amd straglehold.

    1. This acquisition has broken the relation of Arm with its customers. They are no longer IP only and not make chips that compete with Arm customers.

      They can still mess around with the direction of new product development and resources. i.e. transferring good Arm designers into their other divisions or other projects.

    2. Not anti-trust, anti0competitive. That’s two different face of the coin.
      Apple buying ARM will raise anti-trust issues
      Nvidia buying ARM will raise anti-competitive issues
      QCOM buying ARM will raise both anti-trust and anti-competitive issues

Leave a Reply to Gregg EshelmanCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.