ARM Holdings eager for PC and server expansion
Record 2010, looking for
Intel killer 2020
RISC chip designer ARM Holding has closed out a record Q4 and 2010, and is laying the foundations to expand into desktop PCs and servers through the aggressive and enlightened self-interest of its growing licensee base.
In the fourth quarter ended in December, ARM (the company) had £113.9m in total revenues, up 34 per cent. Operating margins were up over seven points to 30.1 per cent, and profits after taxes rose by an impressive 71.3 per cent to £29.7m. For the full year, ARM had £406.6m in revenues and brought just under £86m to the bottom line.
This is a record high Q4 and financial year for ARM, which is celebrating its 20th anniversary this year and looks poised to expand itself from the smartphone market it dominates into new markets, such as PCs, servers, digital TVs, set-top boxes, and a flurry of consumer and embedded devices.
"There's a blurring between computers and smartphones, and ARM's success in smartphones is helping us get into computing," explained Warren East, chief executive officer at ARM Holdings, in a presentation to City analysts in London this morning.
ARM doesn't sell chips, but rather establishes base chip designs that others license and then modify to suit their own purposes. ARM gets money for the licenses and then royalties on the devices that others make and sell. According to East, more than 6.1 billion ARM chips were sold in 2010, a 55 per cent unit increase from 2009, and the company projecting that unit shipments will continue to grow at a very aggressive pace in the coming decade.
In the fourth quarter, ARM's Processor Division accounted for £53.8m in licensing revenues, up 51 per cent, with royalties bringing in another £81.9m, up 29 per cent. The Physical IP Division, which sells chip packaging technology, had £11.6m in licensing (up 26 per cent) and £12m in royalties (up 8 per cent). The Development Systems and Services Division had £11.6m in revenues and £8.7m in follow-on services sales, down 9 per cent and up 12 per cent respectively. For the year, the Processor Division had £458.4m in revenues – up 36 per cent – with the Physical IP Division getting £85.1m (up 18 per cent) and Development Systems and Services Division bringing in $87.8m (8.3 per cent).
By ARM's reckoning, the company had greater than 95 per cent unit shipment share in the smartphone market in 2010, and is very excited that the market for such devices is expected to grow from 280 million units in 2010 to around 1.1 billion units by 2015. In mobile computers, where ARM-based devices had a 10 per cent toe-hold in 2010, over 230 million CPUs were sold in 2010 and the company is anticipating that 750 million units will ship in 2015. (Mobile computers include tablets, netbooks, and laptops in the ARM lingo.)
East said that in digital TVs and set-top boxes, ARM-based products had a 35 per cent market share as 2010 came to a close thanks to 60 per cent growth by ARM licensees, more than double the 30 per cent growth of this market segment overall. And while ARM has zero market share in servers and desktop PCs, it has high hopes, as this chart shows:
The reason why ARM thinks it has a shot in servers and desktop PCs aligns with why Intel's CEO, Paul Otellini, has said that Intel will not see server sales flag in the coming years.
"More use of the internet drives more server usage," said East. "ARM is very attractive in terms of power savings and potentially cost savings going forward. Cortex A processors support multiprocessing and that delivers the high level of performance required by server applications. In this space, we have zero market share and we have an ecosystem that we need to develop."
As for desktop PCs, East said the situation was much the same, but could change now that Microsoft has committed to supporting Windows on ARM-based devices. That said, East conceded that this is a "very established and capable incumbent architecture" in the desktop PC space. But the volumes in the server and PC rackets – and the relatively high prices for chips (and therefore the royalty stream to ARM) is attractive enough to make a run at it.
ARM reckons that 70 million chips were sold in servers and 150 million chips were sold in PCs in 2010, and that the server sector will consume 110 million units and the PC sector around 140 million units in 2015. Add those together and they are in the same ballpark as mobile computers and smartphones individually in 2010. And in a mobile computer, a chip might bring in $20, but in a desktop PC it will be more and in a server even more. Perhaps something like this bell curve that East showed in his presentation:
ARM's plan is to range from low-end sensors to high-end servers, from single-core chips that run as slow as 50 MHz and cost 50 cents to multi-core chips that ramp up to 2 GHz and beyond that cost more than $200 a pop. That high-end server SKU has a revenue stream that is 10 times higher than what ARM is getting from mobile computers. And that is why ARM is pumped about servers.
It is a pity that ARM is not moving the quad-core, 40-bit Cortex-A15 to market faster and talking about either 64-bit kickers or how it will manage virtualized 32-bit applications on multi-core ARM servers such that no one cares that they are 32-bit applications. It is very expensive to virtualize six-core Xeon servers with VMware hypervisors, and the server industry is spoiling for a clever alternative.
A slew of startup companies are spoiling to get into the ARM server racket, too, because of the cost and power issues with x64-based chips and their hypervisors. And there is a tidal wave of companies making ARM-based smartphones and tablets. As 2010 came to an end, ARM had 743 licensees, up 35 during the fourth quarter.
ARM estimates that there is a 4 billion unit royalty opportunity for ARM-based chips using its Cortex-A CPU and related Mali graphics chips in application processing (mobile computing, smartphones, servers, and PCs) in 2015. The real-time embedded chip space represents a 12 billion unit opportunity in 2015 and will be addressed by its Cortex-R family, with another 18 billion units addressable with its Cortex-M designs for microcontrollers. ®
If I might offer a word of advise to ARM based on my own previous experience developing sofrware in a partnership with MS:
- do not allow your relationship with MS to distract your senior management, nor allocate more resources to working with MS than you absolutely have to;
- do allow for the possibility (likelihood) that it may all come to naught, at which point you really dont want to have wasted too much time, attention, or other company resource;
- do not assume there is a wodge of money to be got if you can just handle things right (and that therefore it is worth ignoring the first two points).
ARMs are embedded everywhere
These volumes of hefty, multiple hundreds of MHz ARM cores are paltry compared to the 7TDMI you find in pretty much every WiFi or Bluetooth radio. Even lowly SD cards sport an ARM where it's a throw-away item which runs the BIST once and forever after just manages the SD interface. The ARM instruction set has become what Intel's 8051 was, ubiquitous.
"The complet user experience"
"Linux does not offer the same complete user experience (drivers, software)
None of the existing Windows drivers and software will work on Windows for ARM so your point is moot. It will take years for all the 3rd party devs and hardware vendors to get around to porting all their sundry crap over to ARM and even then they won't bother with legacy stuff. And they probably won't even bother starting in the first place unless they are pretty certain that Windows on ARM is going to be a winner and they can see a big market for it.
The irony is that this leaves WinARM in much the same position as Linux. Devs and vendors don't bother developing for it because it doesn't have enough market share and it doesn't gain market share because the "software and drivers" are pretty much non existent.
It happened before when Windows support MIPs and Power et al
MS has more than enough Intertia.
The fact that plenty of people are willing and eager to spout anti-Linux FUD to end users means that Microsoft is in the driver's seat here. They can take their sweet time in deploying ARM ports.
The biggest problem for Microsoft is all of that much vaunted "superior" app and device support will EVAPORATE. All of that stuff has to be recompiled and tested by all of the various proprietary vendors. This isn't just about whether or not Microsoft decides to embrace ARM but the entire "ecosystem" that surrounds it.
Linux (fragmented or not) is at much less of a disadvantage in this regard.
I can recompile my favorite apps myself.
Despite all their claims about porting Windows to ARM, you need to do a lot more than that to make the platform viable. They learnt that when NT came in builds that supported Alpha, MIPS, PowerPC etc. It's all well and good to have your OS support a different architecture, but if the applications aren't also ported across then it is completely pointless.
Microsoft will have a hard enough time porting their own multitude of products over let alone all the third party devs. How many years do you think it will be before Exchange, Sharepoint and SQLServer are fully ported and offer a seamless admin experience? Personally, I don't think that it will ever happen. To the extent that they port Windows to ARM I think they will limit themselves to porting their consumer OS only along with a not-quite-fully-compatible port of office in the hope of capturing some of the nascent tablet market. I doubt very much that they will port their server products at all.
Which brings me to wondering what the point of porting even consumer Windows to ARM is?
Oh, I understand it from the PoV of MS, they want to be able to sell tablets and not miss out on all the fun but from the consumer perspective what is the point? The Windows UI is not well designed for a touch screen and one of the main reasons for using Windows cited by all of the fanbois is having access to the "millions and millions of windows applications". Apparently, if you can't run these millions and millions of applications on your tablet then you may as well not bother. This is one of the main reasons that Linux is scoffed at by the Windows crowd.
Well, guess what fanboys, with Windows on ARM you will be *unable* to run those millions and millions of applications as well.
So, if you can't run all your precious Windows apps and the UI kinda sucks for touch then what, I ask you, is the attraction of a Windows tablet?