Five thoughts on SiFive

The leading commercial proponent of the RISC V project, SiFive, announced yesterday that it had raised $175 million in its Series F fundraising. With this week’s news of layoffs at RISC V rival Arm, there is some powerful momentum working in RISC V’s favor. In our coverage of Arm, we have written a fair amount about the prospects of RISC V, which we would characterize as cautiously optimistic. As part of the press circuit for their fundraising, SiFive’s CEO gave an interview to Dylan Martin, now at The Register, which is interesting in its own right and emanates that momentum clearly. Here we want to give our take on SiFive specifically, and on RISC V more generally.

Some background. Arm and RISC V are instruction set architectures (ISAs), the blueprints for some of the crucial math operations needed for every processor chip. (More background on this here.) Arm develops and maintains the Arm ISA. By contrast, RISC V is an open source ISA, developed and maintained by the community. Neither Arm nor RISC V sell chips, they provide designs that other chip makers use to design their own chips. SiFive takes the RISC V ISA and designs its own chips as well as help others design RISC V chips. For those more comfortable with software metaphors – SiFive is the Red Hat, to RISC V’s Linux.

The basis of RISC V’s momentum rests on the troubles of Arm. As we pointed out in those past notes, Arm has mis-priced its products for years, penalizing new entrants. And as we detailed yesterday, they have not done much to really advance the Arm ecosystem in several years. On top of that, Nvidia’s acquisition attempt for Arm was a major catalyst to drive large, established chip companies who feared Arm would end up owned by an arch competitor.

Hence the RISC V wave.

SiFive is looking to capitalize on all this. With the latest funding round, they have now raised over $350 million in venture funding. With the addition of $210 million they will get from selling off their connectivity business, SiFive has momentum and a giant pile of cash. Are they now unstoppable?

In short, we think the answer is ‘not exactly’. We think they are incredibly well positioned, and seem to be executing very well, but we see a few obstacles that may not stop them but will require effort to surmount.

1 – Selling on top of open source software is hard. It took Red Hat 20+ years to reach the point where it could be acquired by IBM (of all people). SiFive does not own the underlying ISA, which means they will be perpetually subject to new entrants, many of whom are also SiFive customers. This does not mean SiFive cannot grow, but it may mean a fair amount of back and forth for their growth profile.

2 – Managing communities is hard. RISC V is open source, which means SiFive will have to ‘influence’ the whole RISC V community to achieve their ends. True, their value rests in the designs they provide that sits on top of the open source components, but community management will require significant resources.

3 – Software. We really enjoyed the interview with SiFive’s CEO. He came across as enthusiastic without being overly booster-ish. That being said, SiFive’s ambitions to move into PCs, servers and mobile (?!?) will depend heavily on having the right software to run on those chips. This is incredibly challenging. It took Intel 20 years to build the software ecosystem around its data center chips, and it remains a formidable competitive moat. For the data center, this will require a few hundred people to help port the code, or convincing software developers to do the porting. Both are expensive propositions. And as for PCs, we see little sign of Microsoft porting Windows any time soon. (We would not be surprised if they already have a team working on this, but it has taken Microsoft more than ten years to port Windows to Arm.)

4 – What is SiFive’s business model? We raise this point mostly to give them credit for having actually sorted out a major issue. For years, SiFive was trying to do many things at once, so it was unclear what they were good at. They now have a new CEO who has been getting strong reviews, and while they still have two sides to their model (their own chips and design support for others), the sale of their connectivity business speaks to him making strategic decisions.

5 – China. We now end every review of semiconductors with this topic. While all the major US chip companies are working on RISC V designs, the project’s ‘Base’ is in China, where there are dozens (if not hundreds) of RISC V designers. There are also several SiFive competitors, notably Andes (from Taiwan). SiFive has business in Greater China, but given the state of the world, their ability to gain traction will face constant headwinds from China’s desire to reduce its dependence on foreign technology providers, especially in a greenfield area like RISC V.

Despite all this, we are fairly optimistic about SiFive’s prospects. But they are raising a lot of money at a decent valuation. In the not too distant future, as they confront the challenges outlined above, their momentum will seem to fade, the glow will come off, the press will start to ask about all these challenges. When that happens it will help to remember today’s momentum.

Leave a Reply