Edit: People are very confused about how ARM licensing works. Read this explanation first:
ARM licenses its ISA and/or stock CPU designs to companies. Some companies license both. They pay more. Some companies only license the ISA. They pay less. Apple only licenses the ISA, for example.
Apple's ARM-based designs actually compete with ARM's stock designs. This is why your iPhone is always faster than Android phones.
Yes, Apple is both a customer and a competitor to ARM.
Stock ARM chips are used by companies like Qualcomm, Mediatek, etc. But slowly, these companies are moving away from them and they're designing their own ARM-based chips. This means ARM will make less money from them and they too, will become both customers and competitors to ARM.
ARM's biggest customers are an existential threat to ARM's most profitable business.
I'm not buying. While I believe that ARM CPUs will continue to take marketshare away from x86 on laptops, desktops, and servers, I don't believe the company will be that profitable.
Despite licensing its IP and being at the core of the mobile world, ARM's business model makes peanuts compared to Apple, Qualcomm, AMD, Intel, Nvidia. ARM's last reported revenue was $2b/year. Apple makes that in 1.5 days. Microsoft makes 88x more. Intel makes 39x more. Qualcomm makes 17x more.
In essence, ARM can't compete with the big boys and far more profitable companies are poaching their talent. Take for example, Mike Filippo, a lead architect at ARM, was poached by Apple then poached by Microsoft.
ARM simply can't compete with the money that big tech can throw at their top employees. If you're a top student graduate, you're going to Apple/Intel/AMD/Nvidia. You're not going to ARM Austin Texas. Whenever a big tech company wants to build a custom ARM chip, they go raid ARM talent.
This basically means ARM won't be able to match the performance of big tech designs with their stock ARM cores. And it shows because Apple Silicon is far ahead of stock ARM cores, Qualcomm themselves will be building custom ARM cores again, and Ampere is ditching stock ARM cores for their own custom server cores.
Companies that only rely on the ARM ISA pay far less in licensing fees than companies that also license ARM core designs.
Nvidia buying ARM would have injected the capital needed to compete with Apple, Qualcomm, Amazon, Microsoft, Google, AMD, Intel designs and future designs. And it would have been cool to see ARM SoCs with Nvidia IP. I was/am one of the few people on Reddit who actually liked the deal. The vast majority of people were in the camp of “zomg Nvidia closed source monopoly price fixing omg”. Reddit hive mind is real.
Anyways, I'm bullish on custom ARM designs long-term. I'm bearish on ARM the company because their stock cores can't compete with high-end custom ARM cores and high-end x86 cores. If their stock cores can't outperform custom cores, then they will be reduced to small licensing fees forever. This doesn't justify their $40b+ valuation. Every big player in the mobile space is making tens or hundreds of billions. Somehow the tech that is crucial to the mobile world is making peanuts. This isn't going to change and might get worse.
ARM is essential but not a cash cow and has no potential to be a cash cow.
There is no doubt that Softbank saw the same thing and wanted to unload ARM. First, by trying to sell it to a company who valued its technology more than its financials. Now by trying to IPO. Buyers beware.
Leave a Reply