Reductions in force at SiFive, aka redundancies, have been on the cards for a while I guess with rumours in Spring this year suggesting about a year’s runway left and revenue not making up the shortfall. Internal discussions about cuts were already going on.
When SiFive pulled sponsorship for the RISC-V Summit the rumour, which I think they encouraged, was that they were annoyed at being pushed off key committees. But the timing didn’t fit: I’m told SiFive pulled out of the Summit before the vote results, saving a little cash.
It’s a tough time. Other IP companies are also making cuts at a similar ~20% level, for example my old company, Imagination Technologies. SiFive has cut 130-200 at least, but expect more to go as projects are completed.
They each have their own strategic issues. But they share some too, especially inhashtag#Chinawhere business is down and political tensions high. SiFive are no longer as close to StarFive as they once were and have been trying to sell direct in China, while StarFive is focused on creating data center designs for Baidu. Anger at Arm’s new business model seems to have softened and post-IPO Arm is looking strong and closing business.
Almost everyone I know in thehashtag#RISCVworld was counting onhashtag#Qualcommto develop the market and there are some positive signs: RISE and a wearable processor project in particular. But the kicker is that Qualcomm is proposing a direction for RISC-V that is contrary to the ISA’s original concept and the product roadmap of key players. Put simply, Qualcomm would like RISC-V to be as much like Arm as possible, enabling them to reuse significant amounts of their existing processor designs. It’s a brilliant strategy, but only for Qualcomm. Is it possible? Sure,hashtag#MIPSalready did something similar, creating RISC-V processors from existing MIPS architecture designs.
I’d love to see what advice Qualcomm has received regarding the patent position. I know Arm has a watching brief on RISC-V ready for a future patent challenge if the opportunity arises.
Now SiFive is pivoting to a custom design focus, similar tohashtag#Codasip, whose custom compute strategy seems to be working. Hopefully it will be enough of a change, but at this point they must be regretting spurning what was a very reasonable offer from Intel to buy them.
Strategy? I’d be seriously looking at M&A to get out of this hole. There are other options, but I’d go that way if they can find someone with a big enough RISC-V shaped need who can reassure customers that SiFive are safe to buy from.
Comentarios