The US-China Technological Rivalry Is Leaving the World with Binary Options
- Felix Patrick
- 11 minutes ago
- 5 min read
Contemporary discussion of the technology rivalry between the United States and China often centers on chips and the critical minerals used to make them. The phrase "chip war" became prominent in 2022, when Chris Miller published a book titled “Chip War”, which explained how central semiconductors had become to nearly every industry. That importance has only grown since OpenAI released ChatGPT weeks later. The subsequent AI race between the two countries has pushed semiconductors to the center of debate among policymakers and observers in Washington and Beijing.
Nevertheless, this framing often overlooks an important second layer: this contest will not be decided by whichever country makes the best chips or controls the most minerals. A country with neither can still hold real agency if it controls the software that makes chips useful.
Huawei shows why. When America sanctioned the company, the hardest blow was not a restriction on chips. It was the loss of Google Mobile Services, the licensed bundle of apps and tools that most Android phones need to run properly. Without it, Huawei's phones lost the Google Play Store, the app store most people outside China use. The hardware kept working. The phone became far less useful.
This case explains the shape of the wider rivalry. Trace the value in any modern device, and America's advantage lies less in hardware than in the software layered on top of it. A British firm, ARM, designs the basic blueprint used in nearly every smartphone chip, including Apple's and Samsung's, and charges other firms to use it. Above that sits the operating system, where only two real choices exist: Google's Android, which runs on almost every phone that is not an iPhone, and Apple's own iOS. On desktop computers, Intel and AMD make nearly every processor, and Microsoft's Windows runs most of the machines that use them. Then came artificial intelligence, and a new layer of power emerged: CUDA (Compute Unified Device Architecture), the toolkit Nvidia began building in 2006 that tells its chips how to run AI workloads. A rival that lacks any one of these layers is not merely behind on one product. It is locked out of a whole system built and controlled by someone else.
Huawei's experience taught China a version of the same lesson and pushed the country toward self-sufficiency not only in hardware, but in the software that makes hardware useful. The hardware answer came in 2023, when Huawei's Mate 60 Pro shipped with the Kirin 9000S, a chip that SMIC (Semiconductor Manufacturing International Corporation) designed and manufactured domestically, despite years of American efforts to deny China the equipment needed to make it.
The software answer followed almost as fast. Nvidia spent nearly two decades turning CUDA into the language nearly every AI developer must use to run its chips, an advantage that only began to pay off once the AI boom made that language essential. Huawei's answer, CANN (Compute Architecture for Neural Networks), is a few years old. It already mirrors CUDA's own programming model closely enough to offer near-drop-in replacements for CUDA's tools, cutting the cost of switching over.
The same speed shows up in phones. Google and Apple spent close to two decades each building Android and iOS into the only mobile operating systems that matter. Huawei's answer, HarmonyOS, began as a hedge against exactly this kind of dependence: it launched within months of the first US sanctions in 2019, and by 2024 had shed its last remaining Android code entirely. It has already overtaken Apple's iOS to become China's second-largest operating system, though it remains a minor player elsewhere.
Even lithography, arguably the hardest process in chipmaking, is moving faster than expected. ASML, a Dutch firm, spent roughly three decades and about billions of dollars developing the machines used to print the world's most advanced chips. In 2019, under American pressure, the Netherlands has barred China from buying its most advanced machines. China still has no working replacement for those top-tier machines. But in mid-2026, a state-backed Shanghai firm began mass-producing an immersion deep ultraviolet lithography tool of its own — the first time a Chinese company has manufactured this class of equipment at scale. It falls well short of ASML's most advanced tools, but it cuts China's dependence on foreign equipment for the mid-tier chips that still run most of the world's electronics. Extreme ultraviolet lithography, the technology behind the most advanced chips, remains further off: researchers in Shenzhen built a working prototype in 2025, though most estimates put commercial output years away, probably around 2030.
The pattern repeats across the whole chain, from chip to cloud. On one side sit American chips, American software, and American cloud platforms. On the other side sit Huawei's Ascend chips, its CANN software, and Huawei Cloud. Huawei’s Ascend chips are now mature enough that ByteDance plans to spend $5.6 billion in 2026, with Alibaba and Tencent placing large orders of their own.
One distinction matters more than any other. America built this stack, in large part, to sell abroad. Chips, cloud subscriptions, and software licenses are American exports, and America's own economy depends on the rest of the world continuing to buy them. China is not building its version of the stack for that purpose. It is building it so that China no longer needs the United States not so the world needs China. This partly explains why China and the United States have taken different approaches to open-weight and closed-weight AI models: Washington has kept its most capable models closed, while Beijing has pushed to release its own, DeepSeek and Alibaba's Qwen chief among them as open-weight software that any developer, anywhere, can download and adapt.
That contrast, layered on top of America's growing reliance on sanctions and export bans, is starting to alienate the very firms and countries Washington would prefer to keep on side. China is positioning itself as the more open, more accessible option for any company or country wary of being treated as a security risk by a supplier it still depends on.
The international political economy theorist Susan Strange described power in the world economy as resting on four structures: security, finance, production, and knowledge. America's edge in this rivalry has always rested on the latter. The control over the technology and know-how the rest of the world needs to function. But structural power exercised mainly through denial, rather than through the promise of continued access, tends to erode itself: it gives everyone else a reason to build an alternative rather than stay dependent on a supplier that might cut them off next. China's answer, at every layer from chips to cloud, is a sustained effort to do exactly that, it’s not designed not to overtake the American stack, but to no longer be dependent on it. Washington's habit of reaching for sanctions each time competition intensifies may be reinforcing the very dependency it was meant to maintain.
China's software technology stack is also unlike Germany's chemical industry or Japan's car industry, both of which are export-led and depend on international trade. Sanctions and tariffs Washington's most-used, and arguably overused, policy tools work by threatening a country's access to markets it depends on. They work less well against a country that was never trying to sell into those markets in the first place.
That leaves most other countries with a narrowing set of choices. Few have a chip industry or a software giant of their own, and increasingly, they cannot avoid this rivalry at all. They will have to choose whose phones, whose cloud platforms, and whose chips run their economies, because in the long term, these two systems will drift too far apart to allow countries to use both at once.
This article written by Felix Patrick, Policy and Research Director of Foreign Policy Talks.
Comments