The Great Chip Race: Why the US, China and the EU Are Fighting Over Silicon

[SignalMage92]

A barrel of oil is valuable because the world burns roughly 100 million of them every single day and has no immediate substitute for the energy inside it. A single advanced semiconductor wafer, by contrast, might be worth more than its weight in gold not because of scarcity of raw material, silicon is one of the most abundant elements on Earth, but because of the almost unrepeatable concentration of human knowledge, precision engineering and capital required to turn that raw sand into a working 2 nanometre chip. Oil can be found in dozens of countries and extracted with technology that has existed for a century. A 2 nanometre chip can currently be manufactured at meaningful volume in essentially one place on Earth, using machines built by essentially one company, and that single fact has become one of the defining fault lines of global power in the 2020s.

Why a chip is harder to replace than oil

The comparison to oil is useful precisely because it breaks down so quickly under scrutiny, and the ways it breaks down explain why the chip race matters more, not less, than a simple resource shortage would. Oil is fungible, a barrel from Saudi Arabia and a barrel from Texas serve the same purpose, and if one source is cut off, buyers can generally find another, at a cost. The most advanced semiconductors have no such substitute. Every frontier AI accelerator, from Nvidia's most powerful GPUs to the custom chips Google, Amazon and Microsoft design for their own data centres, depends on manufacturing processes that exist in only a handful of facilities worldwide, built by companies that have spent decades and tens of billions of dollars perfecting techniques nobody else can currently replicate. There is no OPEC equivalent for advanced logic chips because there is effectively no alternate supply to coordinate, the bottleneck sits with specific companies possessing specific, extraordinarily hard to replicate capabilities, not with a commodity spread across many producers.

The chokepoint at the top: TSMC and the silicon shield

Taiwan Semiconductor Manufacturing Company, universally known as TSMC, sits at the centre of this dependency. The company manufactures more than 90 percent of the world's most advanced logic chips, those built at 7 nanometres and below, on behalf of nearly every major chip designer on Earth, Apple, Nvidia, AMD and a long list of others design their chips but rely on TSMC to actually manufacture them at the scale and precision modern computing demands. That concentration has produced a genuinely unusual strategic dynamic that analysts have termed Taiwan's silicon shield, the theory that Taiwan's overwhelming importance to global chip supply itself acts as a deterrent against military conflict, since any disruption to TSMC's fabs would inflict catastrophic economic damage not just on Taiwan but on every economy and military that depends on the chips those fabs produce, including, notably, the very actors who might consider disrupting them. Analysts studying this dynamic are careful to note the shield is not permanent or absolute, it depends on Taiwan's advantage remaining genuinely irreplaceable, and both the United States and China have launched sustained, expensive efforts specifically aimed at reducing that dependency over time, which if successful would gradually weaken the shield's deterrent value even as it strengthens each country's own supply security.

The chokepoint above TSMC: ASML's actual monopoly

If TSMC is the most important manufacturer, the company that makes TSMC's own most advanced work possible is arguably an even more extreme chokepoint. ASML, a Dutch company founded in 1984, is the sole manufacturer in the world of extreme ultraviolet lithography machines, known as EUV, the technology required to etch the impossibly fine circuit patterns onto silicon that modern chips at 5 nanometres, 3 nanometres and below require. No competing EUV machine exists anywhere, from any other company, and none is expected to emerge within the next five to ten years according to industry analysts tracking the handful of alternative technologies in development. These machines are, by some measure, the most complex pieces of manufacturing equipment ever built by humans, shipped in pieces across dozens of separate cargo containers and assembled on site inside a customer's fab. TSMC alone accounts for an estimated 40 percent of ASML's EUV machine deliveries, and the newest generation, called High-NA EUV, costs between 350 and 400 million dollars per machine and is required for the industry's push toward 2 nanometre and eventually 1 nanometre chips. Because ASML's technology underpins the manufacturing capability of TSMC, Samsung and Intel simultaneously, a disruption to ASML's own supply chain, or a political decision restricting where its machines can be sold, ripples through the entire global chip industry regardless of which country's factories are actually running the equipment.

The export control fight: a case study in how hard this is to control

The United States has spent several years attempting to translate that chokepoint into direct geopolitical leverage over China's own semiconductor ambitions, and the actual history of that effort is genuinely more complicated and less linear than the phrase chip war usually suggests. Washington first cut off Huawei's access to American chip technology in stages through 2019 and 2020, then extended controls to cover any company anywhere in the world using American manufacturing equipment to produce chips for Huawei, a rule that forced TSMC itself to stop manufacturing Huawei's most advanced chips. China's largest domestic chip manufacturer, SMIC, was added to the same restricted Entity List in December 2020. For roughly eighteen months afterward, these company specific controls were widely regarded in Washington as working as intended, until a House Foreign Affairs Committee disclosure in October 2021 revealed that US regulators had, during the same period the restrictions were supposedly in force, approved 61.4 billion dollars worth of export licences to Huawei and 41.9 billion dollars worth to SMIC, exposing how much of the restriction existed on paper more cleanly than it did in actual licensing practice.

The policy has continued to swing since. The Trump administration revoked Nvidia's licence to sell its H20 chip to China entirely in April 2025, a decision that cost Nvidia a 5.5 billion dollar charge, only for a more advanced chip, the H200, to be approved for export to China by December 2025 following direct lobbying from Nvidia's leadership. A new framework announced in January 2026 layered a 25 percent tariff on advanced AI chips meeting certain performance thresholds under Section 232 trade authority, alongside a case by case licensing system replacing the previous blanket presumption of denial, with a volume cap estimated to allow roughly one million H200 chips into China, notably only about half of what Chinese buyers reportedly already had on order. Analysts at the Council on Foreign Relations have described the resulting policy bluntly as strategically incoherent, simultaneously acknowledging the national security risk of exporting advanced AI chips to China while creating an explicit legal pathway to do exactly that anyway. By early 2026, the East Asia Forum reported the White House was actively downplaying chip export controls as a public priority ahead of a planned presidential visit to Beijing focused on broader trade stability, provoking pushback from congressional lawmakers who want export licensing authority shifted more firmly under legislative rather than executive control. Taiwan itself has separately been weighing significantly tighter controls of its own on AI chip exports to China, specifically to close routes where advanced hardware gets diverted or smuggled through Taiwan on its way to Chinese buyers, a move that would align Taipei more closely with Washington's approach but risks a sharp response from Beijing given how directly it touches the cross strait relationship.

Despite years of restriction, estimates suggest China's domestic semiconductor industry can currently produce advanced chips at only roughly 1 to 4 percent of US production capacity, a gap analysts expect to narrow only slowly even as Chinese firms like SMIC continue pushing toward more advanced nodes, SMIC has repeatedly signalled progress toward 5 nanometre production, though independent analysts note the company has delayed that specific milestone multiple times, leaving Chinese firms like Huawei dependent on comparatively dated chip technology in the meantime.

Why every major power is now racing to reduce its own dependency

The strategic response from every major power involved has converged on a similar instinct even as their specific tactics diverge, reduce dependency on chokepoints controlled by rivals, whatever the cost. The United States passed the CHIPS Act specifically to subsidize domestic semiconductor manufacturing capacity on American soil, and TSMC itself has responded to the broader geopolitical pressure by building new fabs in the United States, Japan and Germany specifically to diversify its own manufacturing footprint away from concentration in Taiwan alone. The European Union has pursued its own Chips Act aimed at boosting European semiconductor production and reducing reliance on both American and Asian supply chains, framed explicitly around the language of digital sovereignty. China's own self sufficiency drive continues pouring state investment into domestic alternatives across the entire chip supply chain, from design software to manufacturing equipment to the fabs themselves, precisely because dependency on foreign made equipment, and foreign government permission to buy it, has proven to be a genuine vulnerability rather than a theoretical one.

What this means for global power

The underlying lesson sitting beneath all of this technical and policy detail is genuinely simple, whoever controls the ability to manufacture the most advanced chips controls a meaningfully large share of the ability to build the next generation of military systems, AI infrastructure and economic capacity more broadly, and that control currently sits disproportionately with a small number of companies clustered in Taiwan, the Netherlands, South Korea and the United States rather than being distributed the way oil production is spread across dozens of nations. That concentration is precisely why a 2 nanometre chip has become, in a genuinely literal strategic sense, more valuable than oil, oil shortages cause real economic pain but rarely reshape which countries can build the most advanced weapons or AI systems, while a durable chip advantage genuinely does. Every major power's current behaviour, subsidizing domestic fabs, restricting exports, courting allied suppliers, racing toward the next lithography generation, reflects a shared recognition that this specific chokepoint, once established, does not stay static, and whichever nations manage to reduce their own dependency fastest while preserving or expanding their own chokepoints over others will hold a meaningfully stronger hand in whatever the next decade of global competition actually looks like.

[Current]

Genuinely comprehensive and I appreciate that this stayed factual on the export control history rather than framing it as either side clearly winning, because the actual record, the Huawei and SMIC licensing gap revealed in 2021, the H20 to H200 reversal, the Council on Foreign Relations calling the current framework strategically incoherent, really does support a more messy and contradictory picture than most chip war coverage lets on. Where I think the essay could have pushed harder is on just how much the silicon shield theory itself is contested rather than settled analysis, plenty of security researchers argue the shield logic assumes rational actors weighing economic costs in a way that doesn't necessarily hold if a conflict were ever actually triggered by non-economic motivations, and I think presenting it purely as Taiwan's deterrent without at least flagging that debate slightly understates how fragile that specific argument actually is among people who study cross strait security seriously rather than just semiconductor supply chains.