ARM Holdings (ARMH) -- the designer of the eponymous instruction set architecture, as well as a set of micro-architectures that implement that instruction set -- reported a very strong quarter, coming in at $227M in revenues for the quarter against mean analyst estimates of $222M, up 20% y/y. Pre-tax profit of $108.5M was up 22% y/y.
In light of this, many people seem to believe that ARM is the new Intel (INTC). This isn't true and never will be true -- they don't compete at all in the same way. To put things in perspective, in what was widely considered a "bad" quarter for Intel, the company took in $13.5B in revenues and had a net profit of $3B. Why? It all comes down to fundamentally different business models.
How ARM Makes Money
So, here's how ARM makes money. The company allows other companies to have the right to either:
- Design custom processors compatible with the ARM instruction set architecture
- Use ARM's micro-architecture designs to build custom systems-on-chip around a standard core.
ARM receives 1.2% of the sales of each chip built on a "Cortex" core (the high-end mobile products). In essence, ARM allows other semiconductor design houses to build processors using a common instruction set, and in the case of easier time to market, off-the-shelf core designs (which are quite good).
Essentially, ARM is an "arms dealer" to these other companies. It takes a vast amount of time and software sweat equity to design a brand new instruction set, and even the most elegant, well designed one will not be successful if it is not widely adopted. That's where ARM's real advantage lies: it is a well supported, highly popular instruction set architecture that already has a huge software ecosystem.
Comparisons To Intel - Completely Nonsensical
Compare this to Intel, which designs its own instruction set architecture, micro-architectures, and systems-on-chip. After designing the products, it also has the infrastructure to actually manufacture these products and sell them to the device vendors.
So the comparison to ARM is not at all correct, and should not be used as a basis for an investment decision. The more appropriate comparison, however, is with the ARM licensees such as Nvidia (NVDA) and Qualcomm (QCOM). These companies license the ARM instruction set and/or micro-architectures, design products using this IP, and then send the designs to be manufactured at a foundry such as Taiwan Semiconductor (TSM).
While the purpose here is not to determine whether the fully integrated model is superior to the fabless/fab relationship, the point I'm trying to make is the following: ARM does not exist in a vacuum. Somebody needs to design the final solution and someone needs to build it. There are costs to SoC design, and there are costs to chip manufacturing that many seem to disregard when chanting that Intel will suffer serious gross margin compression. In fact, this is downright ludicrous.
Somebody needs to get paid for the actual final product design work (yes, Qualcomm and Nvidia have >50% gross margins), and somebody needs to get paid for manufacturing the products (at the expense of the fabless manufacturers). And in the ARM licensee scenario, ARM needs its cut from the ISA/micro-architecture license. Companies will not compete in the mobile SoC business if it is not lucrative, so the dream of a million ARM-licensees competing for tablet and smartphone market share at small/non-existent margins is a great dream for the consumer, but impractical in the business world, which is why it won't happen.
Intel isn't doomed, ARM will continue to prosper with its business model (especially as ARM does far more than smartphone/tablet chip IP). They're not competitors, and any comparisons between them for investment purposes is ill-advised.