All the way back in August 2023, in the midst of a series of posts on the RISC-V architecture, I ran a reader poll on the future of the architecture, with the following introduction:
As we’re in the middle of a short RISC-V series, I thought it might be fun to test how Chip Letter readers feel about RISC-V.
There’s lots of ambition behind the RISC-V project. In “Instruction Sets Should Be Free: The Case For RISC-V” Krste Asanović and David A. Patterson wrote:
“.. our goal is grander: just as Linux has become the standard OS for most computing devices, we envision RISC-V becoming the standard ISA for all computing devices.”
Please note the use of ‘the standard’ (i.e. singular) rather than ‘a standard’.
Rather than make this a binary question as to whether you agree with this or not, I thought it would be worth broadening the options out a bit and asking it in the context of a specific timeframe.
Please vote in the poll and share your thoughts in the comments thread.
The results were very interesting, with a majority of readers indicating that they they believed that RISC-V would be the standard or the most popular architecture in 2043.
A lot has happened since August 2023. ChatGPT had been launched at the end of 2022 but the full impact of the dramatic build out of AI facilities wasn’t yet clear.
In the world of RISC-V we’ve seen lots of progress on higher performance designs from firms like SiFive and Tenstorrent as well as the ratification of the important RVA23 Profile Standard.
So when I asked the same question in August 2026, I wasn’t sure what to expect.
That’s quite different!
Now, much less than half of readers who voted think that RISC-V will be the standard or most popular ISA in 20 years, a drop of 22% from the 2023 poll.
Why this change?
First of all, thanks to everyone who voted!
Without looking into your minds and motivations, it’s impossible to come to a definitive answer.
I have two theories, though.
First, the scale of focus and investment in AI has diverted both resources and attention that might otherwise have been devoted to RISC-V. The ISA has been an important role in many accelerators but has played a subsidiary role to the accelerator itself in these designs.
Second, the development of RISC-V is to a large extent dependent on the strength of competing architectures. The last three years have seen Intel’s resurgence and further progress in the development of the Arm ecosystem with a server SoC from Arm itself and more high performance cores from Nvidia and others.
Combining the two together means that RISC-V has lost ‘mindshare’.
What do you think? Comments are open so do please share your thoughts.
Cover Image: Development board for BluetrumAB32VG1 UG7200 2118 RISC-V microcontroller SoC
Credit: By Popolon - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=142507606
The Chinese, who graduate more engineers in a year than we do in 10, really don't like paying license fees. Espressive has been making RV chips for a while now. They are not super fast but they are super cheap. I use both the ESP32-C3 and C6 chips that are limited to 140 MHz. Their new ESP32-P4 is a multi-core 400MHz part. Unfortunately the first silicon was buggy enough to require a retape and it will probably be November before Digikey has devkits. There are a multitude of the first rev chips on boards now, but that is just companies clearing out defective inventory. They sorta run but the video compressor is not what is most used and there are instruction bugs too. They were limited by the manufacturer to 360MHz, but I have seen boards advertising 400MHz only to find the part is a rev 1.3 which is EOL by the manufacturer. Other Chinese companies have produced server grade multicore parts and boards. The Reddit RV forum has lots of information. If I remember correctly, Bruce Holt is one of the mods, I think he is located in Australia; he buys, tests and reports on these boards. I have a couple that will run Linux that could be used for daily use, but not real quick. The newer server grade boards are a bit out of my casual spending range.
I reckon that the cost of an Arm license - architectural or chip IP - is a significant factor for low-end chips, but gets lost in the noise for more capable devices. So then the question becomes - does risc v offer at constant technology a substantial silicon area cost advantage at constant performance? Or a substantial performance advantage at constant area? Or substantial architecture-driven power-savings opportunities? And the answer is pretty clear: it’s yet another RISC, so nope - nothing here.
So I’d guess at best it’ll proliferate in the hobby market.
But now that we can leverage AI to build toolchains and silicon quickly and cheaply, I’d expect that custom architectures will make a comeback and risc-v will wither on the vine.