Last week was ARM’s quarterly earnings call. Simon Segars, the CEO, and Tim Score, the CFO, presented from London.
First, the numbers:
- revenue was up 17% year-on-year in dollar terms to $309.6M, but only 9% in Sterling terms due to exchange rate moves to £187.1
- profit before tax was $94.2M
- licensing was up 42% year-on-year with 41 new processor licenses signed during the quarter. Version 8 licensing was particularly strong. Cumulatively ARM have no licensed v8 products 50 times to 28 different companies, 7 in the last quarter. All of the top-10 semiconductor companies that build chips for smartphones have licensed v8, and 9 of the 10 companies that build application processors for tablets have licensed v8. And 4 out of 5 of the top companies selling chips into the networking business.
- During the quarter 8 Mali multimedia processor licenses were signed bringing the total to 96 licenses with 61 different companies. It is now the most widely licensed GPU in over 75% of digital TVs, over 50% of Android tablets and 25% of Android smartphones (of course this wasn’t mentioned on the call, but although iPhones are ARM-based they don’t use Mali, they use Imagination’s GPUs)
- royalties were up 2% year-on-year. Don’t forget that royalties are one quarter in arrears so this is royalties on products that shipped in Q1 (licensees ship product in Q1, then in Q2 they work out how much they shipped and write a check to ARM). Royalties are higher on v8 licenses but realistically it takes 1-2 years for a license for a processor to show through into volume production. Simon explicitly guided expectations that the second half of the year would have higher royalties
- 2.7B ARM-based chips shipped during the quarter, up 11%. Growth was especially strong in…no, not mobile…enterprise networking and microcontrollers (those are the Cortex-Mx cores)
ARM is obviously dominant in mobile with penetration essentially at 100% of smartphones. Of course there are other processors in smartphones too, out of sight hidden in things like wireless and bluetooth. The really interesting battleground is in the server space. Intel owns this space today, since everyone builds out datacenters using standard PC architectures. But that is not an optimum solution for power, physical size or cost if you are someone like Facebook building datacenters for very specific purposes and where total throughput is more important than the absolutely fastest single thread performance (which is what you want if you are doing, say, weather forecasting). ARM have signed 16 licenses for server applications and first production systems are expected next year. ARM expects that by 2018 they will have 10-15% of the total server market. If it turns out to be true then it is a big change. It will be material for ARM in terms of royalties but it will be even more material for Intel in terms of chips not shipped.
Talking of 2018, here is what ARM presented as their expectations across all their markets. $40B ARM-based chips.
For comparison, here are the actual numbers for last year (2013).
One frequently-asked-question of ARM is what happens as integration goes up. When two chips become one, does the royalty go down? The detailed numbers depend on the deals various licensees negotiate but roughly speaking ARM gets a 2% royalty for the first processor and a 1% royalty for the second. So with those numbers, if a company is shipping a kit of two parts, a $5 part and a $2.50 part, then the royalty is 15 cents per kit. If these are combined into a chip that ships at $6 then they get 18 cents per kit/chip. Basically 2% of $7.50 is less than 3% of $6. Everyone wins. ARM expects this type of chip, with multiple ARM processors to increase.
Slides from the call are here. Transcript of the call at SeekingAlpha is here.