RISC-V Foundation just moved[1] from the USA to Switzerland in late 2019 due to geopolitical concerns.

Also, the European Union is heavily betting on the RISC-V for its future hyperscale HPC systems[2][3][4].

I would believe that next-gen European CPU budget declaration is part of this dynamic. And I really hope it will work out.

[1] https://www.reuters.com/article/us-usa-china-semiconductors-...

[2] https://www.european-processor-initiative.eu/project/epi/

[3] https://www.mdpi.com/2297-8747/25/3/46/pdf (PDF)

[4] https://www.nextplatform.com/2019/06/10/europes-homegrown-hp...

The RISC-V foundation moved to Switzerland primarily to help the big Chinese processor companies to work on RISC-V.

Regarding the EU heavily betting on RISC-V, what gives you confidence that this will work? The history of EU funded projects to catch up with the US in AI, in cloud computing (like GAIA-X), in quantum computing, in the space race? HPC is not a mass-market, how should this niche be economically viable to amortise the huge cost of processor development? HPC in the US (and other countries) is often connected to military spending (although this is not always openly expressed). Processor development is extremely expensive, to the extent that Intel is now struggling to finance catching up with TSMC's fab process. Where is the EU's micro-architecture design capacity? ARM is about to be swallowed by Nvidia, and a lot of Arm's micro-architecture development is done in the US already. Now compare the micro-architecture design capacity with that in China (the world's biggest processor market).

   I really hope it will work out.

I do too.

Given 40 years of failure to catch up with the US, and now China, I am sceptical! I expect this to end up like previous EU attempts at catching up: the monies available will be much smaller than originally announced (by at least an order of magnitude), spent over much longer a time-frame, but more importantly, mostly flow to existing EU-based companies that are good ad lobbying, e.g. Siemens, and French conglomerates.

Another prediction: China will soon dominate RISC-V, since they are now forced, by US sanctions, to move away from Arm. I cannot see big Chinese processor companies moving to local ISAs such as Loongson. Alibaba announced a RISC-V processor at ICSA this year [2]. The big US players (Intel, AMD, Nvidia) have much less of a need to move to RISC-V.

[1] https://en.wikipedia.org/wiki/Loongson

[2] https://conferences.computer.org/isca/pdfs/ISCA2020-4QlDegUf...

> The RISC-V foundation moved to Switzerland primarily to help the big Chinese processor companies to work on RISC-V.

Of course. It also removes unnecessary anxiety for the EU that the next ruling power in the US (or elsewhere) may eventually attempt to impose sanctions on the export of the IP. Switzerland, as part of Schengen and Union's single market, fits nearly perfectly into the narrative of European consolidation.

> what gives you confidence that this will work? [...] Where is the EU's micro-architecture design capacity?

I can't see the future, hence I have no confidence. But I express cautious optimism for the following reason: the issue of EU's dependence on some crucial technologies from the third parties is obvious, but there's no direct and quick solution to solve it. The EU can't just turn into venture capital Mecca overnight, the same holds for becoming a low tax heaven. However, the problems/challenges are there, and they must be addressed in the best way under the given circumstances.

The EU has a well-established practice of funding R&D in large areas with non-specific goals or metrics, however this approach is not unique to the EU, it's crucial to the bloc to nourish the interbloc integration and establish logistics between the disparate parts of the market and institutions. It doesn't always lead to unexpected discoveries or breakthrough ventures, but in general it manages the integration task quite well. Which, in turn, contributes to the development of a rich ecosystem. As a tech entrepreneur in the EU, I'm not exactly happy with current affairs, but satisfied with the recognition of my industry's difficulties and proposed attempts to solve them. A failed venture with some new sprouts at worst case, a tangible breakthrough at best.

So far, the EU holds leading position in the lithography systems (ASML). The EU is competitive in radio (Ericsson, Nokia) and embedded (NXP, Infineon, STM) semi manufacturing. There're many blank spaces on the European vertical technology stack, but it's far from a blank slate. This whole commitment is about EU pushing homegrown processor sovereignty across the entire stack. AFAIU, the project puts the focus on HPC as the only specific target platform that needs to be manufactured, but the developments around this project should lay the ecosystem for the boost of the industry as a whole.

Companies such as Apple[1], Microsoft[2], Amazon[3], Google[3], and especially fresh start-ups like Nuvia[4] and Ampere[5] — are working on new ARM-based CPU designs, while SiFive[6] and some others on RISC-V. Also, I'd speculate that AMD and Intel will have to consider switching from the X86 in the foreseeable future. Now is the best possible time for venture into the semiconductor industry in the past few decades. I don't underestimate the complexity of the task. But perhaps a fraction of €145bn would be enough to plant a seed of a viable EU-based CPU semi enterprise; or is it definitely a doomed undertaking, eh? Maybe that French SiPearl[7] or some other company will take chances, we shall see.

[1] https://en.wikipedia.org/wiki/Mac_transition_to_Apple_Silico...

[2] https://aws.amazon.com/ec2/graviton/

[3] https://arstechnica.com/gadgets/2020/04/google-wants-to-dump...

[4] https://nuviainc.com/

[5] https://www.anandtech.com/show/15575/amperes-altra-80-core-n...

[6] https://www.reuters.com/article/us-sifive-tech-idUSKCN2571UL

[7] https://sipearl.com

The foundation just coordinates the standards process and sets rules for using the trademark. It has a tiny budget and few staff. All substantive R&D is done by its members, which are companies of various size with headquarters in various countries: https://riscv.org/members/

Changing where the foundation is incorporated doesn’t change where this R&D happens.

Also Switzerland isn’t in the EU.

On the other hand, without the printing press your reach as a novel autor was severely limited - to the point there were no novel authors basically.

Nice! Great point and it might just apply here as well!

ARM IP, even full architectural license, is relatively small percentage of the cost of a semiconductor. Know-how for fables microarchitecture designs is widely distributed. Most fabless work goes to specialized asics.

It's the inability for SMIC to obtain EUV lithography equipment for their sub-7nm process technologies from ASML that prevents China from getting into high-end semiconductors, not the domestic microarchitecture know-how.

I agree with you.

Here is an interesting question: how long will it take China to catch up? I predict 10 years! The Chinese government sees a leading position in semi manufacture as being of extreme strategic importance. There are several reasons why I think China will succeed other than government strategy: the size of the Chinese market (already the biggest processor market in the world), the ease with which Taiwanese semi expertise can be brought to China (legally, salaries are just higher in Shanghai than in Taipei).

Does that change with the China eu trade deal?

I don't think the US will allow that. They probably have enough leverage.

I think they squandered that recently. We will see with Biden but the Trump administration basically encouraged Brexit and repeatedly adopted anti- EU positions in key geopolitical issues (Syria, NATO, Iran, Israel, and Ukraine). These things come with a cost and the EU will try to rectify this. Enhancing Chinas CPU tech by a few generations might be worth it if it can advance EU interests and ensure future good-faith from a supposed ally.

It's not their decision. The ASML scanners rely on many critical US-origin technologies and the US can block shipment unilaterally.

Enhancing China’s CPU tech even a single generation is in the EU’s interest in exactly 0 alternate universes, along with this one.

As long as the US has the world’s only reserve currency and as long as the EU economy - especially the productive bit in the North - is built on exporting to the all-powerful US consumer, the EU has very little leverage in its affairs with the US.

> the EU has very little leverage in its affairs with the US.

The final word has not been spoken. Construction of Nord Stream 2 has recently been restarted despite the US threatening with sanctions.

This thinking is how you end an empire

And in every case the US was right and the EU was wrong.

Only a fool would tie themselves to China as things stand right now.

>I think much of the relevant chip design knowledge we had (ARM) left the EU this morning.

The larger ARM Cortex cores are still designed to a large degree in Southern France.

Yeah but that's ARM IP though. There's a window of opportunity with RISC V, but something tells me that the fact it's open source rubs against some people.

>Yeah but that's ARM IP though.

How does this matter? Do you think ARM won't license their IP to European chip makers?

Are ARM still Japanese? I think there’s a pretty solid trade agreement between the EU and Japan ...

Has the UK even got a trade deal with the US yet?

No it hasn't, and the type of deal it is going to get looks likely to be different under Biden than under Trump. I think your point is relevant regarding ARM-Nvidia, as semiconductor technology has tended to be the subject of trade barriers, and is seen as a strategic national asset.

Yeah that’s what I was kind of getting at ... great British success story, bought by SoftBank Japan coincidentally straight after Brexit. Would have been a good asset to have in trade negotiations. The sale to US based Nvidia 3 months ago is an interesting twist that would have gone against the EU had Trump won again.

ARM will be owned by NVIDIA soon anyway so it's all swings and roundabouts.

> I think much of the relevant general-purpose chip design knowledge we had (ARM) left the EU this morning

Worth noting the Sophia Antipolis arm design centre in France continues to exist. It works on some of the big A-class cores (The A75 for instance: https://en.wikipedia.org/wiki/ARM_Cortex-A75).

Though whether arm would get involved in this endeavour is another question. They have a significant physical IP portfolio: https://www.arm.com/products/silicon-ip-physical and associated skillsets (I think at least some of this is UK based, not entirely sure how much) so would be a good partner to have and I suspect arm would like to see arm cores coming out as test designs rather than RISC-V cores. Of course in the wake of brexit potentially not a partnership the EU wants politically! UK government would spin it as UK company plays key part in major EU initiative with significant funding awarded. EU would spin it as EU brings in major global semiconductor company to new semiconductor initiative.

Both AMD and Apple Are fabless, so maybe The EU can partner with them like how Apple has partnered with TSMC.

The company has offices and design centres in Copenhagen in Denmark; Oulu in Finland; Sophia Antipolis in France; Grasbrunn in Germany; Budapest in Hungary; Galway in Ireland; Trondheim in Norway; Katowice in Poland; Sentjernej in Slovenia; Lund in Sweden.

> I think much of the relevant general-purpose chip design knowledge we had (ARM) left the EU this morning.

Does that even marginally matter?

Even after Brexit, both EU, and UK will remain open economies, unless you believe a conspiracy theory that Boris is a closet communist.

Your analogy works if ASML makes the only printer that can print the letter 'S'. Without it, you can't print and sell the books, regardless of who the author is.