$1.4 trillion dollars. That’s the investment price tag the Chinese government has committed to paying to develop its domestic semiconductor industry in the next five years. This attempt to achieve self-reliance in semiconductors, which power much of the global economy, is not just a strategic objective but an existential imperative. Multiple rounds of U.S. sanctions are threatening to crippled China’s tech sector, forcing national champions like Huawei to charter planes to bring back as many chips as possible from its main supplier, the Taiwan Semiconductor Manufacturing Company (TSMC), before access officially closes off.
But what can money actually buy in the world of semiconductors? Not much, if spent on the wrong people, doing the wrong thing, with the wrong timeline.
The full lifecycle of a chip is complex, but it can be roughly broken down into four stages: acquire raw material, design chips, marry the design to manufacturing capabilities and produce at scale. Some of these stages are capital intensive; others require brain power, judicious R&D, operational efficiency, and long-term patience — things that money can’t buy.
The most capital intensive parts are the beginning and end — raw material acquisition and building large chip-making facilities (known as either chip foundries, fabrication facilities, or just “fabs”).
Given that China already produces more than half of the world’s supply of silicon — the main raw material for semiconductors — this part is mostly under China’s control and any challenges that arise can likely be solved with direct investment. Building large fabs also requires lots of money for land, energy, construction, equipment and people. Indeed, China’s designated national champion of fabs, Semiconductor Manufacturing International Corp (SMIC), has already attracted more than $10 billion worth of investment from state-affiliated funds and a new public listing on Shanghai’s Nasdaq-esque STAR market. We can safely assume that a good chunk of this $1.4 trillion will go to SMIC.
Fortunately or unfortunately, depending on who you are, the power of money basically stops there. The middle two stages of the lifecycle are the most IP-intensive and most vulnerable to U.S. sanctions.
Chip design requires access to a robust and mature Instruction Set Architecture (ISA). An ISA is a list of instructions that tells a chip how to do math — from elementary school math, like addition and subtraction, to graduate school math, like matrix multiplication (heavily used in Deep Learning AI). The two dominant ISAs come from Intel and Arm, a UK company that may soon become American if its proposed $40 billion acquisition by Nvidia gets completed. Thus, both can become weaponized in a new round of U.S. sanctions.
Building a domestic ISA alternative takes serious technical talent and long term R&D. Progress on this front for China, like the Loongson ISA, is nascent at best. There is an open-source alternative, RISC-V, where any party can run, change, copy, and distribute the technology. Chinese tech giants like Huawei and Alibaba have been investing in RISC-V as a key arrow in their quiver to achieve technological self-reliance. But RISC-V is still young and mostly used as a complement, not a competitor, to Intel or Arm. And I haven’t even mentioned the software ecosystem that must be fostered so developers are willing to build operating systems and applications on top of a specific ISA for end consumers. Arm’s vast ecosystem that earned it a 95 percent market share in the mobile technology sector was a multi-decade effort. A single five-year plan won’t be enough.
Even if we assume that China can build a robust ISA alternative to Intel and Arm, the next step of marrying design to manufacturing is a whole different ball game. The key element is electronic design automation (EDA) — high-end precision manufacturing software, the tools behind a system that helps lay down the transistors on tiny pieces of silicons. This is the technology that transforms designs into functioning chips at scale. Without state-of-the-art EDAs, fabs are basically useless. To be clear, you can just buy this technology — a money problem — but all the advanced EDA vendors are either American, like Synopsys and Cadence, or European, like Siemens. And we have already seen evidence of “unofficial sanctions” working to prevent Chinese chip manufacturers from acquiring these EDAs. Earlier this year, the Trump administration intervened to stop an order of ASML equipment from going to China, where the buyer was rumored to be SMIC. If SMIC is officially blacklisted, a distinct possibility, similar restrictive access to EDAs will only intensify.
On this front, again, building a domestic alternative requires more technical talent, R&D, and patience than money.
On this front, again, building a domestic alternative requires more technical talent, R&D, and patience than money. Thus far, China’s domestic EDA effort is not much to speak of; it’s still years away from being capable of making chips that can power the newest iPhone or Huawei Mate. Efforts to foster open source EDA options are even more rudimentary than RISC-V.
Even though this $1.4 trillion figure has yet to be finalized, government subsidies for domestic players are already flowing — and being wasted. One recent, and egregious, example is Wuhan Hongxin Semiconductor Manufacturing (HSMC). Since its founding in 2017, HSMC has attracted roughly $18 billion worth of investments and subsidies, hired a legendary figure as CEO (Chiang Shang-Yi, TSMC’s first CTO), and has been aggressively poaching engineers from Taiwan with 2 to 2.5 times their current salary. Yet, its fab in Wuhan has not been worked on for eight months due to payment delays to the contractors (not Covid-19) while its prized hires work in a makeshift office that looks more like a shoe factory.
It took TSMC more than 15 years to catch up to IBM, and another 15 years to become the world’s leading chip foundry that it is today, beating both Samsung and Intel. While luring its engineers with money could accelerate some learning to catch up, a similar management quality and operational efficiency takes decades of practice to hone. Even with zero waste and corruption, it’s impossible to replicate in short order.
China’s trillion-plus investment should still be taken seriously, but perhaps not so literally, if only because its literal value is less than meets the eye. And when the entire Middle Kingdom is hellbent on getting a piece of this silicon pie, it may just be mo’ money, mo’ problem.
Kevin Xu is the author and founder of Interconnected, a bilingual newsletter exploring the intersections of tech, business, money, geopolitics, and U.S.-China relations.