Notes on RISC-V, part 2: The Explosive Rise

This is part 2 of a 2-part series. Read the first part "The Origin Story" here.

In July 2022, Calista Redmond, CEO of RISC-V International, said, “We estimate that there are 10 billion RISC-V cores in the market already. Linux did this for software, and we are doing this for hardware.”

It’s an interesting quote. Linux was released as an open-source OS in 1991 and grew quickly as many software developers contributed. It became so popular that it challenged Microsoft, the undisputed emperor of (proprietary) operating systems. In fact, in 2004, Microsoft released a marketing campaign called “Get the Facts” to combat the rise of Linux. The campaign was taken down in 2007, but The Wayback Machine allows us to check out that page’s 2003 and 2007 versions.

Today, Linux is everywhere, and its impact on the world is impossible to understate. From the looks of it, RISC-V is emulating Linux’s rapid rise. Look at the recent news: 

• July 2022: Speaking at the Embedded World Exhibition, the CEO of RISC-V International says, “Estimated 10 billion cores” shipped, primarily in embedded devices. Interestingly, of these, 650 million were shipped by Qualcomm. 
• September 2022: SiFive, the leader in RISC-V computing, announced that it would provide the core CPU for NASA’s next-generation High-Performance Spaceflight Computing (HPSC).
• January 2023: Google announces that it wants RISC-V to be a “tier-1” Android architecture - meaning that we may have devices that run Andoird running RISC-V cores. Devices first and then phones. 

What explains the sudden rise of RISC-V? Let’s look at it from three different lenses - (a) Geopolitics, (b) Arm and (c) open culture.

Lens 1: RISC-V and Geopolitics

From the outside, it might feel like the semiconductor industry is a utopia where companies worldwide come together to make the magic stuff (microchips) that powers the modern era. Who am I kidding? Everyone knows that the knives are out.

It is true that the microchip supply chain is indeed extremely globalised - spread across the US, UK, Europe, Japan, South Korea, Taiwan, China, SEA, and now India. At the same time, it is also true that there are specific chokepoints. The two most notable ones are ASML and TSMC.

ASML

The Dutch company ASML is currently the only company in the world that makes the EUV machine needed by fabs to produce cutting-edge chips, like the Apple M1 and M2, at 5nm technology mode. Described as “the most complicated machine ever built,” it costs upwards of $200 million a piece. 

Can’t anyone else build this machine? ASML has spent decades building this technology and is a decade ahead of everyone else in the space. For its monopoly, the market has rewarded it suitably: As of writing this article, in 203, ASML is valued at over $250 billion. In comparison, Intel is worth $130 billion.

ASML’s machines are so valuable that the US government has lobbied hard to get the Netherlands to stop ASML from selling to companies from China. The company has been one of the focal points of the US-China chip war. 

TSMC

Backstory: After spending 25 years in Texas Instruments and leading its worldwide semiconductor business, Morris Chang returned to his home country - Taiwan - in 1987 and set up the world’s first “pure-play fab.” Taiwan Semiconductor Manufacturing Company (TSMC) was set up to manufacture other companies’ designs. 

By staying true to its original goal, TSMC has placed itself at the centre of the global semiconductor business. As of March 2023, TSMC earned 58.5% of the global semiconductor manufacturing revenue. In second place is Samsung, with 15.8%. The gap between them is a staggering 42.7%. Even more impressive is that it holds a 90% share in the sub-10 nm technology mode, with Samsung at 10%. TSMC’s market capitalisation stands at $ 437 billion. 

Clearly, TMSC is too essential to the global economy. But can anyone change this dynamic? China-based foundry SMIC wants to. But SMIC was put on a trade blacklist by the US in 2020, which means it can’t access certain technologies and machines (including ASML’s EUVs). This means that SMIC will continue to lag behind TSMC (Taiwan-based) and Samsung (S.Korea-based)

That leaves Intel. In Jan 2022, Intel announced its decision to invest $20 billion into a new fab in the US and a similar amount in a facility in Germany. Only time will tell if it succeeds in breaking TSMC’s dominance.

The US-China technology cold war

The executive summary of the US-China “chip war” is such: China intends to gain dominance in the semiconductor space and has achieved technological advances. The US claims this progress has been thanks to technology stolen from US firms and that Chinese companies are putting US technology to military use.

The US has restricted access to US technologies and markets to certain Chinese companies, most notably Huawei and SMIC. These companies can’t access US technology or buy US-made chips or equipment without approval from the US government. Separately, the US has levied additional tariffs on Chinese electronic goods on the market's side. 

Lastly, China claims Taiwan as its own territory and periodically threatens to annex it. Taiwan is home to major semiconductor companies - TSMC, UMC, MediaTek, Foxconn, etc. Some are fabs/foundries; others are packaging units; a few are chip manufacturing factories and testing facilities; and some are fabless design companies. An invasion will halt the global semiconductor business and affect computing, automotive, industrial automation, and many more crucial industries. 

The impact on RISC-V

Thanks to the US-China conflict, Alibaba, Baidu and other Chinese companies have invested in building RISC-V chips to reduce their dependence on Intel’s x86 ISA (US) and the ARM ISA (UK-based) for a while now. 

Countries worldwide are increasingly aware that access to semiconductor IP, raw materials, manufacturing facilities, and finished products might be restricted in future. 

As a result, countries are investing billions of dollars in strengthening their own semiconductor technologies and manufacturing capabilities. In April 2022, the Government of India announced a Digital India RISC-V (DIR-V) program. In December 2022, the EU announced it would release €270 million in funds for companies building chips based on the open RISC-V ISA. 

Note: The EU also has a more extensive €43 billion program, called the  EU Chips Act, to reduce dependency on global supply chains. This is similar to the US’ CHIPS and Science Act which provides $280 billion to boost research and manufacturing of semiconductors in the US. These programs will benefit RISC-C too. 

Lens 2: RISC-V and Arm

Backstory: ARM and arm

In 1983, influenced by early work on RISC by a team at UC Berkeley, Acorn Computers, a company based in the UK, started work on a RISC-based CPU design. They called it Acorn RISC Machine (ARM). In 1990, Acorn spun the team into a new company, with Apple Computers and VLSI Technology as partners. The company was called Advanced RISC Machines Ltd – The “acorn” was replaced with “advanced” to signal that it’s no longer a part of Acorn Computers. 

ARM’s cores became popular, counting Apple, Texas Instruments, Nokia, etc., as clients. The popular Nokia 6110 used a TI chip based on an ARM core. In 1998, the company was listed in London and Nasdaq.  (Read more 1, 2)

It was impossible to predict then, but ARM hit the jackpot with the launch of Apple’s iPhone (which carried a Samsung 32-bit RISC ARM CPU) in 2007. In fact, in 2005, Intel wasn’t keen on selling cores to Apple for the iPhone project because it thought the phones wouldn’t sell much. As the kids say these days, “epic fail.”  

By 2010, ARM had a 95% market share in mobiles. In 2016, the Japanese conglomerate Softbank took ARM private in a £24.3 billion deal. And soon, it renamed itself to “arm.” 

Rocking the Arm boat

It’s well-known that Arm and its parent company Softbank have had a few turbulent years - they have been publicly dragged through hell. During this, Arm has made a few decisions that have spooked its customers.

Everything started with NVidia’s 2020 announcement that it would buy out Arm from Softbank for $40 billion to create the “World’s Premier Computing Company for the Age of AI.” Almost instantly, everyone objected to the deal, including Qualcomm, Microsoft, and Alphabet - Arm had been a neutral partner so far, but this deal would end that. The deal came under intense legal scrutiny and was finally called off in 2022.

Separately, in 2021, Qualcomm purchased Nuvia to bring its brand new best-in-class Arm-based cores to Qualcomm products for both mobiles and servers. Arm responded by suing Qualcomm, demanding that the latter stop work on the upcoming Nuvia range of Arm-based chips. Arm claimed that Qualcomm didn’t have the appropriate Arm license, but Qualcomm rebutted that it acquired the rights when it purchased Nuvia. The fight is bitter, with Qualcomm claiming that Arm’s action is “payback” for its opposition to the Arm-NVidia deal. Whoever wins, both stand to suffer. Further, since both companies are significant players, any impact on either/both will also affect everyone else in the industry.

Lastly, in 2023, Arm announced that it would build its chip. Till now, Arm licensed its IP to chip-makers. By making its chips, Arm would become a competitor to its clients. Arm claims that these chips are meant to demonstrate its technology, much like Google’s Pixel phone was meant to showcase Android at its best. 

A counter-argument will be such: Google’s move did not threaten mobile manufacturers because they were never charged for using the Android OS. Google earns when customers use Google Search (advertisers pay), Google Play (margin on customers’ payments) and so on. Google owner Alphabet’s revenue sources are such that the Google Pixel can not possibly contribute any significant share to it. This is not true of Arm.

And finally, there is the case of Arm’s IPO - which has been imminent for a while and has been finally confirmed (April 2022). One can’t stop thinking about how the IPO will change Arm’s relationship with the rest of the world. On the plus side, public market sentiments (as evidenced by the change in share price) might push Arm to be more strategic about its relationships with the industry at large. However, the goal of a blockbuster IPO seems to be pushing Arm to take aggressive actions in the short term - which might remain in the long-term memory of its partners.

Lens 3: RISC-V and Open Culture 

Let’s go back to that comparison with Linux. What geopolitical factors drove the rise of Linux? None. Which existing company’s public behaviour pushed companies to look for alternatives? The best candidate is Microsoft - which waged a long war on the entire open source movement (calling it “cancer” and “communism”), only to embrace it later. Microsoft vs open source has its own Wikipedia page.

The most significant causal factor in the growth of Linux was that it was free, and customers liked that, especially big buyers like governments of sovereign nations. The other major causal factor was that developers embraced it. The cost of getting started was low. There was no “lock-in.” A vibrant community developed where folks would take from and contribute back to the ecosystem. 

In addition to Linux, these popular tools are also open-source - Apache (webserver), MySQL (DBMS), PHP (scripting language), WordPress (CMS), and so on. Together these tools have enabled millions of developers to build their careers and hundreds of thousands of companies to set up their businesses.

Similarly, the causal factors of RISC-V adoption are (a) end-product cost, (b) low barrier to entry, and (c) no lock-in for engineers and companies. Intel’s x86 and Arm’s ecosystem of tools, products and libraries are (currently) richer, but the ecosystem is only open to those willing to (a) pay and (b) commit to that ecosystem.

And just like Linux, a rich open culture community is getting built around RISC-V. 

What does the future hold for RISC-V

Network effects

In economics, the term “network effects” is used to describe the phenomenon wherein the value of something increases when more people use it. For example, services like Yelp (user review/rating), WhatsApp (personal messaging), and LinkedIn (professional networking) are only useful if many people are using them. These are examples of “direct network effects.”

In “indirect network effects,” the worth of the service to one set of value-seekers increases when another set of value-seekers grows in number and vice versa. For example, Airbnb will attract more travellers only if they have sufficient listings, and property owners will list their rooms only if there is a significant volume of travellers. More examples: Amazon (sellers/buyers), Uber (drivers/customers), TikTok/Instagram (creators/consumers) and so on.

‍

RISC-V is currently experiencing tremendous network effects. Consider three sets of “value-seekers” - Chip makers/sellers, Chip buyers and engineers.

Products that need network effects to grow suffer from the cold start problem - a chicken-and-egg “who comes first” problem. It can take years for the network effects flywheel to start moving.

But here’s the fact: RISC-V has overcome this problem already since billions of chips are already shipped!

Under the hood

Microchips are used in three broad applications - embedded devices, personal computing (including mobiles) and servers. In embedded systems, RISC-V is already well-proven and widely used. 

Many of these applications are in devices where the chip is meant to work silently without drawing attention to itself. For example, a Mindgrove Secure IoT may be used in a smart electricity meter or a biometric device, or a Mindgrove Vision SoC might be used in a security camera cluster installation.

In other applications, a RISC-V chip may be used in addition to another chip. The Google Pixel 7 uses the ARM-based Tensor SoC and a RISC-V-based Titan M2 security chip. 

Mainstream

For RISC-V chips to achieve mainstream popularity, they must be the primary chips in mobiles, computers and server chips. This might take a while. 

As of May 2023: For servers, SiFive and Ventana have announced RISC-V-based cores. For computers, StarFive has a high-performance Single Board Computer (SBC) with an onboard GPU (Imagination IP) called Vision Five 2. This SBC may not work for a full-fledged personal computer but can replace a Raspberry Pi. For mobiles, the first RISC-V mobiles will likely come from Chinese companies like Huawei or Alibaba.

The room for innovation

At first glance, it might feel like the semiconductor industry is obsessed with the latest, newest chip at the latest node. Just recently, attention across the world was focused on TSMC’s new 3nm capability and how Apple has pre-purchased almost all of it.

But not all chips need to be M1/M2/M3/…

For most applications, the latest ARM or x86 chips are simply overdesigned. In real-world applications, only 10-20% of the chip may be used. This is similar to how we drive 200 km/hr rated cars at 20km/hr in city traffic. Just think about it: how much computing does a smart TV, a fridge, or even an industrial robot need?

Not just that, requirements change across applications. Automobile chips for India must work across a temperate range of Ladakh to Vellore and a humidity range of Mumbai to Jaipur. Some chips must process AI on edge—or video (like in ADAS), while others must consume very low power. 

Summary

Depending on who you ask, RISC-V’s growth can be attributed to (a) geopolitics, (b) ARM’s adventures, or (c) open culture. Whatever the causal factors, the truth remains that RISC-V is experiencing explosive growth. This growth is being accelerated by government action across the world.

RISC-V-based chips have already seen wide acceptance in embedded systems and will soon be seen in computers and mobile devices.

One thing is for sure: this is an exciting time for Mindgrove Technologies, with three RISC-V SoCs (Shakti core) due to release soon!