My Enemy’s Enemy Will Still Eat My Lunch

Amidst the collapse of Nvidia’s bid for Arm last week, an interesting announcement came out of Intel. They announced that they were joining the RISC V Foundation and establishing a fund to invest in RISC V projects. Many observers tied the two events together. Knowing what we do about the length of Intel’s press release approval process, we suspect the timing was largely coincidental. Still, it does seem like a smart strategy for Intel. Outflanking their long-standing rival Arm, by seeding the open source alternative.

That being said, there are some major flaws with the strategy.

First and foremost, Intel’s announcement was a bit more nuanced. Their investment is really a bid to attract customers to Intel Foundry Service (IFS), their nascent foundry business which is a bit short on customers right now. And this seems to be less a venture fund and more a commitment to build the tools necessary to allow RISC V customers to actually be able to access Intel’s foundries. Tools are a big problem for IFS.

The second, more important problem, is that RISC V is very unlikely to play out to Intel’s long-term strategy.

Some background. Instruction Set Architectures (ISA) are the underpinnings of modern chip processors. They handle a lot of the low level math (here is our plumbing analogy that explains it fairly well). For decades Intel’s x86 ISA has fought it out with Arm’s ISA. RISC V is an open source ISA alternative to both x86 and Arm. As with all open source projects it has its fair share of friction, and credit to Intel, their investment looks to smooth over some of that friction.

RISC V exists because Arm got too comfortable with its monopoly position. Its licensees, which include some of the biggest chip companies like Apple and Qualcomm, are entirely dependent on Arm. This is how they persuaded the US anti-trust authorities to block the Nvidia deal. And like all monopolists, Arm mis-priced its products to the point that it made no sense for new entrants to the chip market to work with Arm when a free alternative was available.

Put simply, the x86 versus Arm fight ended up with x86 dominant in PCs and Arm dominant in mobile. x86 is still the leader in the data center, but Arm is making real inroads there as well. That means the real battlefield is shifting to two new theaters – Autos and IoT. We will focus here on IoT, it is too soon to tell how Autos will trend, but we have some theories. Every self-respecting chip company CEO will spend a lot of investors’ time explaining the importance of IoT to their business. This is a major opportunity.

But what is IoT? It is really a catch-all term for thousands of different types of devices, many of them needing very little in the way of compute performance. It is also a new, greenfield market. If we scroll up a few lines, the problem becomes apparent. Arm is not doing particularly well in IoT because they do not do well with new customers. Because IoT is so diverse and so new, there is ample room for new chip entrants. And very few of them want to use Arm. To make matters worse, many of these new entrants are in China, where Arm is having a few problems.

We believe, with increasing conviction, that the market for IoT chips is not going to go to Arm. But it is not going to go to x86 or Intel either. It is going to go RISC V.

So while Intel investing in RISC V may seem like a satisfying snub of a long time rival, Intel is also enabling the competitor most likely to take home the prize that Intel’s design business would very much have liked to win.

6 responses to “My Enemy’s Enemy Will Still Eat My Lunch

  1. IoT is not a “new, greenfield market”. It’s been the subject of conversation, an about-to-arrive wonderland, for at least two decades; last time I was asked to work on the technology was while employed at Sun Microsystems. Since then I have been aware of multiple big IoT-infrastructure efforts, all of which have missed their targets catastrophically. For many years now we’ve been promised that all the world’s industrial installations will be full of little devices that are smart enough to be useful but somehow too dumb to do basic protocols like HTTP and run OSes like Linux, so they need special hardware and software. Been promised, that is, by the people who want to sell the hardware and software. I’m not saying there’s no there there, it’s just that the people who are preaching it have extremely damaged credibility.

    • I mostly agree with you. (I wrote this four years ago on the mess that is IoT – https://digitstodollars.com/2018/11/07/why-is-iot-such-a-mess/)

      My only point here is that there are 1,000 chip companies in China – all designing chips for things that aren’t laptops or smartphones. Industrial machines, robots, all the stuff that gets lumped into IoT. And none of those companies are going to do that on Arm.

      And this is happening at a time when the cost of putting some form of more sophisticated “compute” in all those dumb IoT devices is reaching economical viability.

  2. Good article. Can you elaborate on this part “more a commitment to build the tools necessary to allow RISC V customers to actually be able to access Intel’s foundries.” ? What sort of tools and their functionalities?

  3. Good article. Can you elaborate about this part – “more a commitment to build the tools necessary to allow RISC V customers to actually be able to access Intel’s foundries.” What sort of tools and their core functionalities are expected to be built? Any papers or books that I can read about to know further?

    • A big part of running a foundry is having tools that allow others to input their design files into your system. Think of it as an interface between an ERP system and a technical review of the product. TSMC has worked with the big EDA companies for years to smooth out this interface. But Intel has largely relied on internally designed tools to accomplish this task – going from the IDM side to the fab operations team. No one outside Intel is familiar with these tools. Enter RISC V which generally lacks tools to support foundry processes, RISC V is just so new that no one has built a complete system. Intel is committing to design tools to facilitate porting files from the design companies to their fabs. They are in the process of improving their tools for everyone, adding RISC V support is not a big deviation from this.

  4. Pingback: Arm: Credit where credit is due | Digits to Dollars·

Leave a Reply