MIPS Acquired by AI Hardware Vendor Wave Computingby Ryan Smith on June 15, 2018 7:00 PM EST
- Posted in
- Machine Learning
- Wave Computing
After a year-long on-and-on search for a suitor, MIPS has finally found its partner. Today the company has announced that it’s being acquired by Wave Computing, an AI startup with close ties to the company, and whom will be using MIPS’ technology to further advance their own AI product portfolio.
MIPS sale has been more than a year in the making, going back to 2017 when Imagination Technologies began shopping the division around in the wake of losing Apple as a GPU customer. Imagination opted to focus exclusively on their GPU and related businesses, leading to the company putting MIPS and other non-GPU divisions up for sale. At the time Imagination was unsuccessful in selling off MIPS, and eventually Imagination itself was acquired by Canyon Bridge, which ultimately resulted in MIPS being sold to Tallwood Venture Capital due to geopolitical concerns.
At the time Tallwood did not announce what they intended to do with MIPS, but with a history of rehabilitating and flipping companies, a resale was likely. And indeed, this is exactly what has happened.
Wave Computing, in turn, is a small but not unknown name in Silicon Valley. The company has been around since 2010 but is still venture-capital backed and operating as a startup, with a focus on machine learning and artificial intelligence. Importantly, Wave Computing already has close ties to MIPS both as a customer and in terms of personnel. The company is a recent MIPS client, tapping MIPS’ CPU designs for their planned second generation of products. Meanwhile several staff members, including CEO Derek Meyer, whom was once a MIPS VP. The end result being that MIPS has essentially been acquired by one of their clients for their non-exclusive use.
For their part, Wave Computing has stated that they intend to use MIPS technology to expand their machine learning product portfolio. Specifically, the company is looking to branch out from datacenter ML training to ML inference in edge devices, where MIPS’ existing products and their own work into inference products would be a natural counterpart to Wave Computing’s existing technology. Ultimately this would allow Wave Computing to offer a singular top-to-bottom machine learning product portfolio, with common technology ranging from datacenter training to the actual use of machine learning in edge and other end-user devices, giving them a leg up over vendors who can only supply one or the other.
It goes without saying that machine learning is considered to be a highly lucrative market by investors right now, especially on the back of NVIDIA’s phenomenal growth in the space. The market is still very young, and this marks the latest of many venture capital-related plays in the field to either grab a piece of what some analysts expect to be a very large pie, or to at least bump off the current kings of the market.
As for MIPS itself, it will continue to be operated as a business unit of Wave Computing. And importantly, Wave will continue to license out MIPS technology. Given Wave’s unique focus, the company has unsurprisingly already made it clear that they intend to offer future MIPS designs with their machine learning technology. Otherwise neither Wave nor MIPS have said much more about future plans, and while Wave has little reason not to continue with MIPS current initiatives, this is unlikely to lead to any kind of resurgence of MIPS in the high-end CPU IP market, as some have hoped to see.
Wrapping things up, Tallwood and Wave Computing are both private companies, the terms of the deal are not being disclosed. Though it is interesting to note that due to Wave Computing’s small size, Wave is ultimately acquiring a company almost as big as they are. Wave reportedly has just 100 employees, with MIPS being similarly sized.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Findecanor - Sunday, June 17, 2018 - linkMIPS has a long and proud history. If it had been managed better it might have been competing well with ARM by now.
Good to see that it has a home after Imagination.
ws3 - Sunday, June 17, 2018 - linkIt’s just another of the casualties of the Itanic.
The only reason ARM avoided the same fate is that ARM chips were thought to be “small” and therefore not in the same league, and therefore no threat. Every other RISC architecture apart from Power was cast aside for the fools dream that was Itanium.
mode_13h - Monday, June 18, 2018 - linkDidn't MIPS have a fairly strong foothold in the embedded space - notably telecoms?
fteoath64 - Tuesday, June 19, 2018 - linkYep, until SDN comes in with x86 being the most cost effective. There are some custom telecom serial chips with MIPS cores but not updated as often. Dying domain for MIPS I think.
Let us hope MIPS put some serious ML blocks and get more mileage with their architecture. We can all reminscient those huge SGI systems with large MIPS processors of bygone years ...
mode_13h - Monday, June 18, 2018 - linkBTW, I'm pretty sure x86 had a lot more to do with MIPS demise than Itanium. SGI was already onboard the x86 train, even before Itanium saw the light of day. And x86 certainly had a lot more to do with the demise of SPARC than Itanium ever did.
strtj - Monday, June 18, 2018 - linkSGI was onboard the Windows NT4 train, which was perhaps not the worst of ideas but was misguided. What do you do when suddenly your graphics technology has matured to the point that real people want to use it at a reasonable cost?
But honestly, Itanium was a horrible noose around the neck of HP and SGI for its entire life, which started at about the same time as the awful NT workstations. This was much earlier than people remember, probably because it shipped so late. The architecture was an interesting idea, it turned out it didn't work, and it sank MIPS and HPPA with it.