DEEP DIVE: Canada needs to make our own computer chips.
Be careful; that Yankee PC has an Intel Inside™.
What if I told you that America has infiltrated every Canadian home, every place of business, and every government office? What if I told you that backdoors exist into mainstream computer processors, enabling remote access and control over every personal device and every database?
This isn’t a myth. We know from the Edward Snowden leaks that the NSA has worked for over a decade to plant vulnerabilities and backdoors into American technology products.
The Intel Management Engine and AMD Platform Security Processor have been included in all Intel and AMD processors since 2008 and 2013 respectively; despite denials that these are backdoors, they perform exactly the utility that the NSA stated they wish to implement.
What do you think happens to a military when another country can remotely access their computer systems? The invading country launches their invasion, while the defending country’s planes are grounded and missile defences unable to launch.
Not pretty. Not a situation I would ever like Canada to experience. And fear of this kind of circumstance is exactly what has motivated the Chinese government to create an independent supply chain for electronic components.
I believe that we should be taking similar actions to China, actions which South Korea and Taiwan have already completed. I believe Canada should enter the semiconductor industry, and create a crown corporation to design and manufacture computer processors and memory, for both the domestic market and for export customers.
Mind you, the financial undertaking will be…large, to put it mildly. There are three big things that all need to align for a fully independent computer architecture in Canada, uncompromised by the Yankees:
We need to obtain modern lithography equipment to make wafers on advanced process nodes.
We need to design modern CPUs and accelerators which support existing interfaces
We need to provide robust drivers to connect the hardware we build to the software that people run.
Let’s go over each of these in turn:
What is a “Process Node”?
Good question! When manufacturing semiconductors, whether they be processors or microcontrollers or memory chips, the fabrication plant operates on what we call a “Process Node”.
The node is measured, roughly, by the distance in nanometres between two transistor gates on the chip; some 7nm nodes will have higher transistor densities than other 7nm nodes, but they’re a useful categorization…if arbitrary.
What does this mean in practice? With all else held equal, if you fabbed the same chip on a 10nm node vs a 7nm node, the 7nm chip will do the same work with fewer watts, or more work for the same amount of watts.
Obviously, the denser you can push the transistors, the better, with some asterisks. This generally holds regardless of whether you’re trying to build a massive supercomputer for the TOP500 benchmark, or a smartwatch sipping slowly on power over a 36-hour battery life.
The most power-efficient nodes are naturally the most expensive, and many semiconductors are still produced on older nodes based on the industry they serve.
Embedded applications like office printers and washing machines don’t need much power, and ruggedized systems like NASA’s Curiosity rover care more about features like radiation shielding than raw processing power, especially when data can be streamed back to Earth for heavy processing.
To manufacture on 7nm and 5nm nodes, it was initially believed that an EUV lithography machine was necessary; these machines are very expensive, and are only made by Dutch company ASML Holding.
But in an attempt to block China from advancing their semiconductor industry and competing with Taiwan and the West, the United States imposed restrictions that prohibited ASML from sending EUV machines to China, forcing them to rely on the last-generation DUV lithography.
And yet, China did the impossible; they made advanced nodes using stockpiled DUV machines previously obtained from ASML. They did so by innovating new techniques, allowing them to produce 7nm wafers at high enough yields for commercial viability.
As a result, the Huawei Mate 60 features a Kirin 9000s system-on-a-chip (SoC), designed and produced in Mainland China on a 7nm node, using DUV lithography. It is inferior to the Kirin 9000, which used a 5nm node from TSMC, but it is made entirely on the mainland, and it is merely the first generation.
Now, they appear to be on the brink of cracking 5nm with that same DUV equipment. China has even begun producing their own DUV equipment to reduce reliance upon stockpiled western machines; in time, they could even figure out how to break the ASML monopoly on EUV.
The lesson here is that necessity breeds innovation, and that smart people can take good advantage of the right tools, if the government puts the tools into the right hands. The good news is, Canada has a much stronger relationship with The Netherlands than China does.
During World War Two, Canada safely harboured the Dutch royal family in their exile. In 1943, the Crown Princess and future Queen Juliana gave birth in Ottawa to her daughter, Princess Margriet. Canada legally declared the maternity ward in Ottawa temporarily extraterritorial, to ensure the Princess would remain solely Dutch in citizenship.
Canada would then lead the military liberation of The Netherlands from the German occupation. This special relationship is commemorated every year with the Canadian Tulip Festival, where the Dutch send their love to Canada in the form of many, many tulip bulbs.
Even if Canada does not compete in semiconductor design, the government spinning up a fabrication plant would allow us to compete in production. And advanced process nodes are valuable enough to be worth their weight in gold.
I believe that we can obtain EUV, or at least DUV lithography machines from ASML Holding; even if they are forcibly compelled to withhold ALL lithography equipment, I believe we can obtain what we need from the Dutch through back channels; China has managed to evade sanctions to do the same, and the Dutch like us a lot more.
What is an “Instruction Set”?
At the lowest level, any computer program will translate down to machine code called “instructions”. These instructions map out the basic tasks that your software performs, in the simplest logic possible; while you’re scrolling through this article, your phone’s CPU is executing billions of instructions per second.
When you write a piece of software, you place your program in either a compiler or an interpreter; both of these transform your program into instructions that the CPU can run.
To make sure that programmers don’t need to recompile their code every single time a new CPU is released, CPUs will target an “Instruction Set Architecture”, or an ISA for short.
Newer CPUs in an ISA family may add new instructions to that ISA as an “extension”, and these extensions accelerate certain tasks. CPUs can be developed on Field Programmable Gate Arrays, which allow transistor gates to be defined virtually; many CPU designs exist only as soft cores simulated on FPGAs, and have never been fabbed.
The most famous ISA is likely x86, which originated with the Intel 8086 launched in 1978; the Intel 8088 variant is the CPU at the heart of the original IBM PC which ran DOS. If you’ve ever touched a Windows PC, you owe it to the PC…and the fact that IBM failed to stop people from cloning it.
Today, both Intel and AMD manufacture 64-bit CPUs under the x86_64 ISA; while Intel owned the patents over the original 32-bit ISA, AMD invented the 64-bit extensions necessary to use more than four GiB of memory, and after many years of lawsuits they settled by cross-licensing all relevant patents.
That’s the real rub; while you can legally reverse-engineer an ISA and emulate or simulate it, many aspects of x86_64 have not been legally reverse-engineered, especially modern extensions released after 2005.
The major competitor to x86_64 in the current day is the ARM family of ISAs, specifically ARMv8 and ARMv9. ARM licenses not only their ISA, but pre-designed CPU cores as well; many Android phones use an SoC with these off-the-shelf ARM designs.
But the most famous example of a mainstream computer using ARM is probably Apple’s new “Apple Silicon” series of Macs; all Macs containing an M1 to an M4 series use Apple’s custom-designed ARM cores, a privilege for which ARM charges Apple a premium compared to licensees for the off-the-shelf ARM cores.
So, if Intel and AMD will never license the x86_64 ISA, and ARM has placed significant limitations around their own licensing terms, and reverse-engineering for both is limited, then what is the alternative?
It turns out there’s a third competitor in the ring, and it’s growing in popularity for a pretty clear reason. RISC-V is an ISA with many similar technical merits to ARM, but with one clear benefit; the consortium that designed the specification released it for free, without a licensing fee.
The benefit of this is clearly obvious; anyone can design a RISC-V CPU core, whether they wish to manufacture their chip and sell it, or whether they’re a hobbyist sharing their design as a soft core. Engineers can focus on building good designs, and not reverse-engineering unlicensed technologies, a process of trial-and-error.
The European Union has dedicated €270 million Euros to fund the domestic design of a new RISC-V CPU for supercomputers and servers; China is itself pursuing RISC-V, which has enraged Yankee politicians, knowing their corporate sponsors are facing competition they cannot suppress.
Indeed, there is very little that America can do to sanction RISC-V, or any open-source project; you can’t criminalize an idea. The specification is open; anyone can design a CPU compatible with the ISA, and the instruction set is intentionally small to make implementations simple.
Alibaba Group’s research division, DAMO Academy, has released a series of open-source designs for multi-core RISC-V processors. These designs are now spread across the internet, and theoretically anyone could contract a fabrication plant to manufacture them, if they find a plant and can fill an order quantity.
Undoubtedly, there are graduate students at the University of Waterloo who are working on RISC-V designs right now. And they will get snapped up in a flash by Yankee companies across the border like Qualcomm, or by Chinese companies like Alibaba.
I would much rather those bright minds be hired by a government crown corporation, well-compensated for their expertise, actively working on designing Canadian CPUs and GPUs and AI accelerators and more, not just to sell domestically but globally as well.
A Canada that sells these things to the world is a Canada that becomes very wealthy, and very influential; this is one of the biggest plays we could make to diversify the Canadian economy, and we would reap immense trade benefits.
Even if we just sold commodity memory, we would be disrupting a stagnant market with three major players; SK Hynix and Samsung are both South Korean, and Micron is American. There’s a reason these countries still compete with Taiwan on semiconductors, but we’ll get to that later.
What the hell is “Linux”?
There’s just one thing missing before you can use your Canadian-made computer to watch Netflix CBC Gem: a bunch of digital glue.
Most people in the 21st Century run an operating system like Windows or macOS, and software runs within that operating system, while the operating system controls access to the bare metal hardware. For the hardware to be recognized, a driver needs to be written for each operating system.
Because of this, you cannot natively run a program compiled for Windows on your Mac or on Linux. You will either need a compatibility layer, sacrificing performance and accuracy, or you will need to recompile the original code, which you cannot do if the creator does not give it to you, which they may never do.
But what is Linux? An operating system that is free and open-source, which runs on PCs, smartphones, and servers.
The code written for Windows and macOS is closed-source, meaning that nobody knows what they actually do, and if there are hidden “surprises” that corporations or government entities don’t want you to know about.
This is not the case for Linux; all the code is open to the public to audit. Frequently, the public spots security vulnerabilities, fixes them, and then shares those fixes immediately with the rest of the world.
You are likely already using Linux in your daily life, and you may not even realize it. This is because Android, which controls a majority of the market share for smartphones, is actually a version of Linux.
And yet, even Android has alterations from standard Linux, and while Android is ostensibly open-source, Google’s web services that many Android apps rely upon are not open-source.
Things are even more lopsided when it comes to servers, where Linux dominates over Windows Server due to superior performance and reliability. Even Microsoft distributes their own version of Linux for customers to use on their Azure cloud.
Linux is growing in global popularity as a PC operating system, with both the Indian and Chinese governments encouraging uptake; I personally use Fedora, which has different UI options similar to Mac or Windows. Valve’s Steam Deck is a handheld gaming PC, similar in shape to a Nintendo Switch, and it essentially runs normal Linux.
You would be shocked how easy it is to use Linux as your daily PC. There are very few things it does not do that Windows does, and there are very few things they both do that Linux does not do better. There’s a reason that even Microsoft has relented and embraced Linux.
For those who still need to run a specific Windows program, a compatibility layer called Wine allows Windows applications to run under Linux. Another compatibility layer, Box64, allow x86_64 Linux programs to run on other ISAs like ARM or RISC-V.
When you combine Wine and Box64, a RISC-V Linux machine now has the ability to run x86_64 Windows programs, bridging the gap and keeping old programs backwards compatible. It can only run so fast, but it runs much faster than you would expect.
These are not perfect projects, but they are open-source, and as a result dedicated funding can pay for developers to improve their accuracy and speed. Valve, for their part, funds work on the Wine compatibility layer, because it benefits the product they sell.
At the bare minimum, a Canadian chipmaker will need to release Linux drivers for the CPUs and GPUs they release, and keep them updated with fixes and new features. In doing so, Canadians can manufacture and sell fully-functional computers on Canadian soil, compliant to open industry standards.
Are we really going to do it though?
Some would say “not a snowball’s chance,” but I don’t think we have a choice.
It would not be an understatement to say that we could spend several hundred billion dollars on this plan. This would dwarf the cost of Canada’s largest previous public infrastructure projects…but it would also dwarf them in reward.
Taiwan’s semiconductor industry leads the world because TSMC has the most advanced node, and everything needs semiconductors. They charge a price premium for being on the bleeding edge, but Samsung makes healthy profit as well.
Both TSMC and Samsung have reached 3nm, but TSMC’s first-generation node beat Samsung in transistor density, and then TSMC created several refinements of their node while Samsung’s second-generation 3nm node is still lower in transistor density than TSMC’s first generation.
Meanwhile, Intel only reached 3nm by the time Samsung and TSMC released second generations of their nodes, and they’re having enough difficulty that they’ve outsourced significant production quantities to TSMC across various CPU and GPUs.
When you have to beg your own competitor for production capacity, you’re unfortunately a joke. America would not need to rely upon TSMC if their Yankee darling Intel could operate even a competitive fabrication plant.
Apple places such a premium on the power efficiency of TSMC’s bleeding edge nodes that in 2023 they bought the entirety of 3nm wafers from TSMC. Do you understand what that means?
This meant that in the autumn of 2023, the most advanced chips in the entire world were only available if you had a new MacBook Air, MacBook Pro, iMac, or the iPhone 15 Pro; 3nm wafers could not be obtained for server chips dedicated to enterprise, government, or scientific use.
From a national security perspective, this is absurd. If you question why Samsung bothers to compete against TSMC, it’s because competing in the global market isn’t the focus; the focus is maintaining domestic capacity.
It doesn’t matter that a Samsung Exynos 2400 or a HiSilicon Kirin 9000S might not have the most advanced node; what matters is that South Korea and China have the ability to manufacture their own reasonably powerful semiconductors domestically.
This is why America is working closely with TSMC to bring a fabrication plant to Arizona…and this is why many Taiwanese are afraid what will happen to their sovereignty if neither China nor the United States rely upon them to make chips. Controlling production of an advanced process node is a massive geopolitical asset.
So, if Canadians want to be smart, we’ll realize that this isn’t just about the large potential for profit. This is about being on a war footing in an increasingly hostile world, with an incredibly aggressive southern neighbour who will use our reliance upon them against us.
We need to obtain lithography equipment from ASML, preferably EUV, but in the worst case that we are unable to obtain such, China has demonstrated that we can obtain reasonable yields at 7nm and even 5nm using older DUV machines.
And to be frank, the European Union understands just as well as we do how much of a threat both China and America have become. We have a mutual interest: Canada can manufacture chips and provide natural resources, Europe can sell lithography equipment and provide technical expertise, we can collaborate on design, and we can mutually benefit from a stable supply chain for semiconductors.
This would be a transformative effort for the Canadian way of life; through a crown corporation, hundreds of billions of dollars could be added in revenue to the government’s annual budget if we bring competitive nodes and chips to market.
Canada would also become an attractive place to assemble electronics, if we become a major producer of semiconductors. This is one of the biggest moves that Canada can make to diversify our economy, and increase our global standing.
We cannot trust the Yankees, and we know that American semiconductors and American software developers are explicitly targeted by the NSA to force the inclusion of backdoors into their products, whether they’re manufactured by American companies or by TSMC.
But this does not mean we can trust China, and that they are not doing the same. What we do know is that across the globe, open standards like RISC-V and Linux are being adopted because we don’t need to trust their security; sunlight will be the best disinfectant.
You don’t need to be a big nerd to see the benefit of all this. It’s nice when people work together to make good things for society. It’s a shame the Yankees have never understood cooperation, but thankfully we do not share their limited mindset.
Sadly, I fear that the sticker price of this endeavour will be the immediate end of the discussion. Even if we push past the price anxiety, there would be immense diplomatic pressure from America, from China, from Taiwan, and from South Korea, all politely informing us that it would not be in our interest to compete with them.
But that would be the exact reason we need to do this. Because ever since Donald Trump threatened to annex our country against our will, Canadians have decided we must never again live at the mercy of others. And if the US Government has a secret “control switch” for every AMD and Intel processor…then we are in extreme danger.
We must work closely with our allies in the European Union, but we must also increase our self-sufficiency. I am happy to take money from other countries when we succeed in this plan, but we have a greater goal than profit.
We need Canadian governance, and business, and science, and life, to happen on computers that Canadians control. Because nobody should control the future of Canada, except for the Canadian people ourselves.