A number of Chinese companies have made use of the MIPS architecture. Would one or more of them be interested in buying MIPS Technologies Inc. and taking royalties from Broadcom, Cavium, Cisco and others? Would they be allowed to do so?
LONDON – The report that processor and related IP licensor MIPS Technologies Inc. is seeking a buyer, if true and taken through to a final conclusion, would be a case of a one-time industry leader looking like a dinosaur after the meteor struck. And the meteor that struck the planet is the rise of mobile computing.
As with ST-Ericsson, which was kind of a self-assembly dinosaur (see Why ST should sell ST-Ericsson to China) the fast changing nature of the market to which a wealth of technology can be applied is causing fall-out for some but allowing some big corporate animals (Apple) and locations (China) to thrive.
In the case of MIPS it seems to me more likely that one or more of the U.S. licensees rather than a Chinese licensee would make a bid. The reason is that culturally the licensing of intellectual property is still a work in progress in China and as such would probably feel like a less secure business than one based on real product delivery.
Indeed it is thought that some now-legitimate Chinese licensees of MIPS architectures started out as pirates. Although they, and the Chinese government through its Loongson-based supercomputing projects, may have an interest in where the ownership of MIPS technology eventually resides they would not yet seem to be ready to turn all the way from poacher to gamekeeper. But if they should it might prompt a bidding war with the U.S. for strategic control.
Broadcom, Cavium and Cisco all make extensive use of MIPS licenses and one or all of them might consider it unsatisfactory to have their processor roadmap in someone else's control.
And it should not be forgetten that this would not be the first time that MIPS has been acquired. The company was founded in 1984 as MIPS Computer Systems Inc. and started out by selling Unix workstations based on its own designs of RISC microprocessor. It became a seller of RISC chips to workstation builders.
The architecture became so important to graphics workstation vendor Silicon Graphics Inc. (Sunnyvale, Calif.) that the company acquired MIPS in 1992 to make sure a particular 64-bit design was not abandoned. It was during a subsequent SGI downsizing that MIPS re-entered the market as a stand-alone licensor of processor designs.
So is a Cisco or a Broadcom going to save MIPS? Possibly. Or maybe a consortium of licensors can put a syndicate together to hold the ownership of the intellectual property.
But 20 years on from SGI the situation is different. There is a viable alternative to MIPS that is already striving to bring up networking and higher performance architectures that goes by the name of ARM.
It seems to me that a year ago Sandeep Vij (CEO) was seen as MIPS savior (wasn't he one of your top ten executives who made a difference in 2010?), and that MIPS would soon be eating ARM's lunch. What happened?
The 1980's DLX instruction set is possibly in the public domain. There are a host of 32-bit RISC uControllers and bus interfaces that don't need anything from MIPS (ARM's AMBA etc). Why buy a company when you can license the ISA? Why should China buy ST or MIPS or any other failing USA or European companies? And to add insult to injury, the share price goes up -- like when the Lexra consortium looked into buying MIPS some time back, but then walked away (most are probably ARM licensees now).
For MIPS it is about aptented instructions (less then word loads and stores). It is mostly legal story: without those instructions you can't call it MIPS, and if you make CPU with them you have to pay for them (and those insturctionsa re not commonly used). But Chinies made their own implementation from scratch so it is not fair to call that coping. Or just say that AMD is coping Intel x86 CPUs.
Are all the comments that say the Chinese just want to copy stuff and not play IP games based on up to date info or are they just prejudice?
Before getting too arrogant it is worth remember that some of the greatest American engineering feats claimed- like the space program - were based on foreign IP.
I suspect that as Chinese engineering, business and research changes we will see a change in the way IP is handled too.
MIPS has been looking for a buyer for the last year or more.
Perhaps Synopsys should finish the acquisition that it started in 2009 with the analog portion of MIPS. Synopsys already has ARC (via Viragelogic), so it shouldn't make matters any better or worse with the ARM relationship.
Market research suggests that ARM had 50 to 60 percent of the embedded processor market in 2010. The next most significant architecture was MIPS with about 10 percent, closely followed by a number of other architectures.
Although the market has probably moved towards ARM in the last year I think the vast size of the embeded market in terms of units will support a number of licensors, particularly as the detail of the hardware becomes less important compared with the overall package of Java Virtual Machine on hardware, Android, etc.
However, the the strength of ecosystem support and having critical mass is also important.
So for the usual market reasons I think at least two main architectures will continue to have enough scale. But I also expect other architectures to come through based on lower cost than ARM, and where roadmap is less important, and to occupy niche positions. These include ARC under the auspices of Synopsys, specialist DSP cores and hardware accelerators, and far-eastern startups such as Andes Technology.