Intel Buying SiFive?

[Sal Ammoniac]

Hmmmm... I wonder what this means for RISC-V?

https://www.tomshardware.com/news/intel-offers-dollar2-billion-for-risc-v-startup-sifive-bloomberg

[SiliconWizard]

Good lord.

[brucehoult]

An offer is possibly quite far from buying. It seems there may be multiple offers. All could be refused.

At minimum it does help set a base valuation for the next funding round, if there is one.

I struggle to see how such a deal would make sense for anyone. Ok, maybe SiFive founders if they want a "never need to work again" pay day now rather than a potentially much bigger pay day later on. But I think at least Krste, Andrew, and Yunsup actually do want to change the world and burying SiFive inside Intel quite likely would not help with that.

Intel doesn't need to buy SiFive to get a license to make RISC-V chips. Anyone is allowed to. Buying SiFive doesn't give control of RISC-V. Intel doesn't need SiFive technology, unless they're planning to get into microcontrollers and IoT, and even then they can (and already do) simply license cores from SiFive. SiFive's most sophisticated shipping cores are around original Pentium or PowerPC 601/603 in complexity (at much higher MHz obviously) Pentium Pro level cores have been announced but are not yet shipping. That's 25 years behind Intel. Intel could do an Apple M1-style RISC-V very easily if they want to, but they don't need to buy SiFive to do it -- it wouldn't even make it easier, particularly.

Does it make sense for SiFive or RISC-V? It's hard to see it. Too much risk of getting buried inside the behemoth. It's not even as if Intel can threaten to compete with them and squash them -- SiFive's battle for the foreseeable future is with ARM, not x86, and not an M1-level RISC-V (if Intel wants to make one)

[PCB.Wiz]

Sometimes, offers are made to get the engineering teams and their client streams.
There are many strategic benefits to this for Intel.

[FlyingDutch]

Hmmmm... I wonder what this means for RISC-V?

Hi,

there is one more possibility: Maybe Intel wants to inhibit development of RISC-V hardware.

But this opinion catch on conspiracy theory 

Best Regards

[Siwastaja]

It's not a conspiracy theory, it's a completely normal business practice to buy out smaller companies (especially startups) just to inhibit their efforts when seen they could pose a competitive threat but do not fit nicely into the larger buyer company's portfolio. The risk of competition doesn't even need to be very real, this can be done "just in case". It's common and completely legal to do so.

[nctnico]

Hmmmm... I wonder what this means for RISC-V?

Hi,

there is one more possibility: Maybe Intel wants to inhibit development of RISC-V hardware.

That doesn't make any sense. Intel is not in the segment where RISC-V may become big (embedded and mobile). ARM is more likely to be threatened by RISC-V

[SiliconWizard]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

[nctnico]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

And then what? RISC-V is open source; any other company can jump in.

[ataradov]

And then what? RISC-V is open source; any other company can jump in.

Sure, but at the moment SiFive is a significant driver behind the architecture. And if nobody will actually want to jump in, then RISC-V will go back to its "paper only" status.

[brucehoult]

Sometimes, offers are made to get the engineering teams and their client streams.

Both have legs, if they don't like the new owner.

Founders and other staff with a lot of stock can be made to commit to staying several years as part of getting their (multi) million dollar pay day. There is no leverage over staff who joined too late to get much.

[brucehoult]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

And then what? RISC-V is open source; any other company can jump in.

That was actually the best point I've seen.

SiFive is getting to the point of having pretty high performance OoO cores -- announced but not shipping yet -- but there are performance numbers from FPGA and simulation. Something in the Pentium Pro, II, III, Pentium M spectrum which Intel after all is still basically just incrementally enhancing now 25 years later. And I think we'll find they have a pretty good vector processing unit too.

Nothing that Intel can't duplicate in their sleep if they decide to make a high performance RISC-V implementation, of course.

Any other company is of course free to start designing RISC-V cores. But it's a significant amount of work, which companies other than Intel / AMD / IBM / ARM could short-circuit a lot by buying SiFive.

Even those companies would get a couple of years head start from acquiring the current designs.

[rsjsouza]

In my mind, such purchase would only make sense if Intel is trying to buy the people's expertise - the IP is open anyways and (hopefully) there is no legal loophole that would lock it.

[ataradov]

SiFive's IP is not open. The spec is open, but by buying SiFive you don't "get" the spec in any way.

The only reason for Intel to buy them is to remove one of the most significant players in the ecosystem.

[phil from seattle]

That doesn't make any sense. Intel is not in the segment where RISC-V may become big (embedded and mobile). ARM is more likely to be threatened by RISC-V

Perhaps you missed a more obvious reason.  Intel is nowhere in those spaces and they see it as an opportunity to be a leader in yet another market segment.  Pretty much how big companies get bigger.

[tszaboo]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

And then what? RISC-V is open source; any other company can jump in.

It could be something trivial, we wouldn't think of in a million years. Like they want to use the cores in the management engine for ethernet ports in servers. Or as a core for their 5G modem portfolio.
It doesn't have to be steamrolling the market with chips.

[NorthGuy]

They just have too much money which has to be spent before it's eaten by the inflation.

[rsjsouza]

SiFive's IP is not open. The spec is open, but by buying SiFive you don't "get" the spec in any way.

The only reason for Intel to buy them is to remove one of the most significant players in the ecosystem.

I read some more. You are correct and I confused the open ISA with the closed implementations from different companies. In this case, I suspect that Intel is buying SiFive for the same reasons it purchased StrongARM: to quickly enter a market they have zero know how. Once they get bored with it, they will sell it just like they did with their SA/XScale to Marvell.

[brucehoult]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

And then what? RISC-V is open source; any other company can jump in.

It could be something trivial, we wouldn't think of in a million years. Like they want to use the cores in the management engine for ethernet ports in servers. Or as a core for their 5G modem portfolio.
It doesn't have to be steamrolling the market with chips.

Samsung and Qualcomm already use SiFive RISC-V cores for those purposes. For both the 5G modem and controlling the camera in the high end 2020 Samsung phones (they said so), and I assume but don't know in 2021 also. Same with Qualcomm for 5G.

That's a $1 million or $2 million license. No need to spend 1000 times more buying the company.

[chickenHeadKnob]

They just have too much money which has to be spent before it's eaten by the inflation.

Ding! Ding! Ding! we have a probable winner.  The engineering mindset tends to over-emphasize the engineering aspects of the deal. I think back to the tech stock boom of 1998- 2000 when NorTel was flying high on insane valuations and I tried warning my father to sell them. Under John Roth NorTel was buying up companies left and right while at the same time inflating their order book by subsidizing their customers. Some of the companies purchased made a modicum of sense from the engineering perspective but over-all he was driving NorTel into the ditch.

[HwAoRrDk]

I seem to recall some news earlier this year that Intel were planning on opening up their foundries to external customers. What if an acquisition of SiFive is to ensure they have a guinea pig 'customer' to help refine their capability to produce other companies' chips?

[brucehoult]

I seem to recall some news earlier this year that Intel were planning on opening up their foundries to external customers. What if an acquisition of SiFive is to ensure they have a guinea pig 'customer' to help refine their capability to produce other companies' chips?

That co-operation was announced months ago in March.

https://www.sifive.com/blog/sifive-collaborates-with-new-intel-foundry-business

Again, you can do that without buying them.

[rsjsouza]

They just have too much money which has to be spent before it's eaten by the inflation.

Ding! Ding! Ding! we have a probable winner.  The engineering mindset tends to over-emphasize the engineering aspects of the deal. I think back to the tech stock boom of 1998- 2000 when NorTel was flying high on insane valuations and I tried warning my father to sell them. Under John Roth NorTel was buying up companies left and right while at the same time inflating their order book by subsidizing their customers. Some of the companies purchased made a modicum of sense from the engineering perspective but over-all he was driving NorTel into the ditch.

Back then, Nortel and so many others were riding on the waves of Arthur Andersen's consulting tips for a successful business, where Enron was the straw that broke the camel's back. Would something similar be taking place twenty years later? Definitely plausible, but Intel does not seem to be in the hot water as these other companies were.

[tszaboo]

Even if Intel has no direct interest in RISC-V at the moment, they may be considering buying SiFive just to prevent some other company from doing so.

And then what? RISC-V is open source; any other company can jump in.

It could be something trivial, we wouldn't think of in a million years. Like they want to use the cores in the management engine for ethernet ports in servers. Or as a core for their 5G modem portfolio.
It doesn't have to be steamrolling the market with chips.

Samsung and Qualcomm already use SiFive RISC-V cores for those purposes. For both the 5G modem and controlling the camera in the high end 2020 Samsung phones (they said so), and I assume but don't know in 2021 also. Same with Qualcomm for 5G.

That's a $1 million or $2 million license. No need to spend 1000 times more buying the company.

Yeah, but synergy and vertical integration and other weasel talk.

[NorthGuy]

That's a $1 million or $2 million license. No need to spend 1000 times more buying the company.

In the times like this?

https://www.washingtonpost.com/business/on-small-business/biden-will-need-more-than-52-billion-for-chip-effort/2021/06/09/0f4ebffc-c91b-11eb-8708-64991f2acf28_story.html