Intel and ARM used to live in peaceful coexistence. ARM designed small chips for a litany of inexpensive devices--mobile phones, disk drives, game systems, anti-lock brakes, washing machines--while Intel's forte was power-hungry CPUs with desktop brawn. Then came the smartphone and the netbook, and the dynamics began to change.
As the lines between smartphone and PC blur, "ARM is coming up from the portable space, and Intel is coming down from the PC space," says Joseph Byrne, a senior analyst at the Linley Group who specializes in semiconductors and processor IP. "Looking forward, these guys are going to collide."
The conflict will play out in the competition between each company's low-power, high-performance chips, which promise to be smaller and more efficient than full-sized CPUs, but more powerful than traditional smartphone ones. ARM's 32nm Cortex design, which will start to appear in devices late this year, will be used in everything from netbooks to mobile Internet devices, achieving multi-gigahertz speeds with less power consumption than bigger-geometry chips. (Intel currently uses a bigger, less efficient chip, the 45nm Atom.) The Cortex chip may win significant marketshare from Intel, which currently dominates the netbook market with 80% market share.
"There is an inherent cost savings in the 32nm geometry because the chips are smaller," says Simon Segars, ARM's general manager of physical intellectual property. "But everyone is under such tight cost control that the exact moment when people will switch over is hard to tell," he says. "It makes for very interesting times in the semiconductor industry."
Intel plans to compete with new plans for the Atom, which is found in everything from Asus' Eee PC to the Dell Mini 9. The company will be porting the processor's cores to work on the technology platform of Taiwan Semiconductor Manufacturing Company, or TSMC, allowing customers to license Atom's blueprints and customize the processor to their needs. "This is a big deal," Byrne says. "There's an existential risk if Intel holds onto x86 too tightly." Intel will keep developing its own Atom roadmap, but may be hoping to achieve ARM-like proliferation with its licensees. Intel also has a next-gen version of its 45nm chip on the way code-named Moorestown, which the company says will be more efficient than the Atom and more scalable than ARM's designs.
Each company approaches the low-power, high-speed netbook market with a drastically different business model. UK-based ARM Holdings [ARMH] doesn't actually manufacture anything; they design the building blocks for microprocessors and license that intellectual property to chipmakers for a fee plus royalties. "With the exception of Intel, we're working with all the big boys," says Segars. "Almost every semiconductor company is using ARM in one way or another, and many are paying royalties."
This doesn't mean that each company gets the same chips; they merely get similar building blocks that they can implement in their own way. "At the block-diagram level, they look similar. But the way the system is put together is very different," says Segars.
That's an advantage, says Byrne, because it allows ARM to diversify the risk of failure of each chip design. "ARM doesn't have to care if any one company does well," he says. "If you're selling the arms to both sides, it doesn't matter which side wins." The downside is that ARM relies on thinner margins--only about 5.5 cents of profit per chip, says Bryne, compared with several dollars per chip for Intel. "Although the ARM partnership is a disaggregated model," says ARM's Segars, "it can still deliver the same technology as someone like Intel, who's doing everything itself and investing massively," he says.
Investing massively it is. Intel [INTC] designs and manufactures its chips all in-house with its 84,000 employees and an R&D budget of about $5.7 billion per annum. (ARM, by comparison, employs less than 1,800 people and reports an R&D budget of just $76 million). In 2009, many of Intel's billions will be aimed at ARM's home turf: Mobile. In February, Intel released its most recent quarterly report and made clear the new primacy of phone chips: "Shipments of our mobile microprocessors exceeded our desktop microprocessors for the first time," in Q2 of 2008.
But that same annual report showed an Intel that had been deeply hurt by broader economic woes. Demand for processors had dropped precipitously in the fourth quarter, falling short of analyst expectations (though recent predictions have been more sanguine.) While Intel was aching, ARM seemed insulated from the tanking economy, posting good numbers last quarter. It used the opportunity to earn a little good press about its flexible business model, amidst predictions that mobile phones sales would continue to grow despite slumps elsewhere.
However, ARM's bad news may simply be delayed. "ARM has two big revenue streams: royalties and licensing," says Byrne. "Royalties are recognized in arrears and trail by a quarter. For the chip industry, Q4 was the big drop, so ARM hasn't released its bad numbers yet."
Intel, Byrne says, hasn't let up on research despite last year's rough patch. "They still have the pedal on the gas in capital investment," he says, and its profitability will bounce back more quickly than ARM's. Still, ARM can leverage their silicon-on-insulator technology, which is currently limited to high-end products like the Sony PlayStation, to give MIDs and netbooks even higher performance with even more cost savings. Whichever company gains the lead, consumers come out ahead.