Geopolitics & Conflict News & Analysis Tech & AI Policy 10 min read

The GPU Shortage and the Sovereignty Race: Why Nations Are Nationalizing Silicon Supply Chains

The US government took an equity stake in Intel. China set an 80% chip self-sufficiency target. Europe pledged EUR 31 billion. The race for semiconductor sovereignty has become the defining industrial policy battle of the decade.

Semiconductor sovereignty manufacturing facility with silicon wafers
Reading mode

The global chip shortage never really ended. It transformed. What began as a pandemic supply crunch has evolved into something more permanent: a race for semiconductor sovereigntyA country's ability to design and manufacture its own advanced computer chips without depending on foreign suppliers. that is reshaping how nations think about silicon the way they once thought about oil.

In August 2025, the United States government did something unprecedented in peacetime: it took an equity stake in Intel, converting $5.7 billion in unpaid CHIPS Act grants and $3.2 billion from the Pentagon’s Secure Enclave program into $8.9 billion of common stock[s]. The federal government now owns 433.3 million shares, representing a 9.9% stake in America’s most important chipmaker[s]. This partial nationalizationThe process by which a government takes control of private industry or assets, transferring ownership from private entities to the state. joined a 15% Pentagon stake in rare earth company MP Materials[s].

The reasoning is straightforward: the United States accounts for only about 10% of global chip manufacturing capacity[s], and Intel is the only American company with leading-edge fabs capable of producing chips at the most advanced nodesA chip manufacturing process using the smallest transistor sizes, measured in nanometers; smaller nodes enable faster, more power-efficient chips.[s]. Letting it fail would mean ceding semiconductor sovereignty to foreign competitors entirely.

Why Semiconductor Sovereignty Matters Now

Three forces are colliding to make chip independence a national security priority.

First, artificial intelligence has created unprecedented demand for advanced processors. NVIDIA’s data center revenue hit $57 billion in Q4 2025 alone, up 62.5% from the previous year[s]. The company now captures roughly 44 cents of every dollar hyperscalersLarge technology companies that operate cloud computing infrastructure at massive global scale, such as Meta, Google, and Amazon. spend on AI infrastructure[s].

Second, Taiwan Semiconductor Manufacturing Company (TSMC) produces the vast majority of the world’s advanced chips, and it sits 100 miles from mainland China. This concentration riskThe potential for significant losses due to over-exposure to a particular investment, sector, or asset class. has prompted TSMC to invest $165 billion in Arizona manufacturing facilities[s], with the first fab already in production and additional facilities planned through the decade.

Third, the memory bottleneck has exposed how fragile the supply chain truly is. High Bandwidth Memory (HBM), essential for AI acceleratorsSpecialized computer chips designed to speed up artificial intelligence computations, such as GPUs and TPUs., is made by only three companies: SK Hynix, Samsung, and Micron[s]. Samsung and SK Hynix together control approximately 80% of global HBM production[s]. Both are based in South Korea, which imports 70% of its crude oil through the Strait of Hormuz.

China’s 80% Self-Sufficiency Target

Beijing has responded to US export controlsGovernment regulations restricting the transfer of sensitive technologies, materials, or data to foreign entities for national security reasons. not by retreating but by accelerating. China’s semiconductor industry has set an official target of 80% domestic self-sufficiency by 2030, according to 13 top company leaders[s].

The results so far are mixed but notable. SMIC, China’s largest chipmaker, posted record revenue of $9.3 billion in 2025[s]. US export restrictions have paradoxically added “rocket fuel” to domestic chip demand, according to Paul Triolo of Albright Stonebridge Group[s].

China has captured roughly 30% of the global market for legacy chips[s]. On advanced nodes, Chinese firms now produce 7nm processors comparable to what TSMC was shipping in 2018[s]. The gap is closing, even if slowly.

Europe Joins the Race

The European Union has launched its own push for semiconductor sovereignty through the European Chips Act, which aims to double Europe’s global market share to 20%[s]. The Commission has approved over EUR 31.5 billion in investments for semiconductor facilities across member states[s].

In September 2025, all 27 EU member states signed a declaration calling for a Chips Act 2.0 to address remaining vulnerabilities[s]. The message is clear: semiconductor sovereignty has become a bipartisan priority across democracies.

The Collateral Damage: Gamers and Consumers

The race for AI chips has created real casualties. NVIDIA has reportedly cut RTX 50-series gaming GPU production by 30 to 40% to prioritize AI accelerators[s]. If analyst predictions hold, 2026 will be the first year in three decades without a new generation of NVIDIA consumer GPUs[s].

“Every bit of memory that’s out there, I think, is really getting prioritized to AI computeComputational resources including processing power, memory, and storage used for AI model training and inference.,” said Bernstein analyst Stacy Rasgon. “That dynamic is starving the overall industry of the type of memory that is traditionally used for more consumer type applications.”[s]

What Comes Next

The semiconductor sovereignty race will not end when factories are built. New geopolitical flashpoints keep emerging. Tungsten prices have surged 557% in just over a year after China added the metal to its export control list[s]. Tungsten is essential for chipmaking, and China controls 79% of global production.

Meanwhile, NVIDIA has restarted H200 chip sales to China after a ten-month freeze, under an arrangement where the US government takes a 25% revenue cut[s]. The chip war has entered a new phase where even “restricted” trade continues through managed channels.

The $165 billion question is whether any nation can truly achieve semiconductor sovereignty, or whether interdependence is simply the price of modern technology. The answer will shape the next decade of geopolitics.

The global chip shortage never really ended. It transformed. What began as a pandemic supply crunch has evolved into something more structural: a race for semiconductor sovereigntyA country's ability to design and manufacture its own advanced computer chips without depending on foreign suppliers. that is reshaping national industrial policy around silicon supply chains the way energy policy was once shaped around petroleum.

In August 2025, the United States government executed an unprecedented equity conversion: $8.9 billion in CHIPS and Science Act grants were transformed into primary shares of Intel Corporation[s]. The federal government purchased 433.3 million shares at $20.47 per share, acquiring a 9.9% stake[s]. The funding came from $5.7 billion in unpaid CHIPS grants plus $3.2 billion from the Secure Enclave defense program[s].

This partial nationalizationThe process by which a government takes control of private industry or assets, transferring ownership from private entities to the state. accompanied a 15% Pentagon stake in MP Materials, the largest US rare earth producer[s]. The strategic logic is clear: the United States holds only about 10% of global chip manufacturing capacity[s], and Intel remains the only US-headquartered company with leading-edge fabs capable of producing at the 2nm node and below[s].

The Technical Case for Semiconductor Sovereignty

Three converging factors have elevated chip independence to critical national security status.

First, AI computeComputational resources including processing power, memory, and storage used for AI model training and inference. demand has exploded. NVIDIA’s data center segment generated $57 billion in Q4 2025 revenue, a 62.5% year-over-year increase[s]. Hyperscaler capital expenditure reached $130.7 billion in the same quarter, with NVIDIA capturing approximately 44 cents of every capex dollar[s]. This concentration is unprecedented in enterprise computing history.

Second, advanced nodeA chip manufacturing process using the smallest transistor sizes, measured in nanometers; smaller nodes enable faster, more power-efficient chips. manufacturing remains geographically concentrated. TSMC has committed $165 billion to Arizona fab construction[s]. The company recorded NT$1.134 trillion ($35.67 billion) in Q1 2026 revenue, up 35.1% year over year[s]. Yet Taiwan’s proximity to potential conflict zones means diversification is existential risk management.

Third, the memory supply chain has emerged as the true bottleneck. High Bandwidth Memory (HBM) production is controlled by three vendors: SK Hynix, Samsung, and Micron[s]. HBM requires 2 to 3 times more silicon area per gigabyte than standard DRAM, uses advanced 2.5D/3D packaging with through-silicon vias, and suffers lower manufacturing yields[s].

SK Hynix and Samsung together control approximately 80% of global HBM production and 70% of the DRAM market[s]. Both are headquartered in South Korea, which imports roughly 70% of its crude oil through the Strait of Hormuz. SK Hynix’s HBM revenue grew 82% across 2024, with market share rising from 30% to 44%[s].

China’s Capacity Buildout

Beijing’s response to US export controlsGovernment regulations restricting the transfer of sensitive technologies, materials, or data to foreign entities for national security reasons. has been acceleration rather than capitulation. China’s semiconductor industry has established an 80% domestic self-sufficiency target for 2030, endorsed by 13 leading company executives[s].

Progress is measurable. SMIC posted record revenue of $9.3 billion in 2025, a 16% year-over-year increase[s]. Analyst estimates project $11 billion for 2026. US export restrictions have functioned as “rocket fuel” for domestic demand, according to Paul Triolo of Albright Stonebridge Group[s].

China has captured approximately 30% of the global legacy chip market[s]. On advanced nodes, Chinese foundries now produce 7nm-class processors powering Huawei’s latest devices, technology comparable to TSMC’s 2018 output[s]. The technological gap persists but is narrowing systematically.

European Industrial Policy

The European Union has operationalized its semiconductor sovereignty strategy through the European Chips Act, targeting a doubling of global market share to 20%[s]. The Commission has approved EUR 31.5 billion in state aid for first-of-a-kind semiconductor facilities[s].

Notable investments include the ESMC joint venture (TSMC, Bosch, Infineon, NXP) in Dresden for CMOS and FinFETA type of 3D transistor architecture where the current flows along a vertical fin-shaped channel for better electrical control and reduced power consumption. production, STMicroelectronics’ EUR 5 billion SiC device facility in Catania, and Silicon Box’s EUR 3.2 billion advanced packagingSophisticated semiconductor manufacturing processes that integrate multiple components into complex chip configurations. facility in Novara[s].

In September 2025, all 27 member states signed a Semicon Coalition declaration calling for Chips Act 2.0 to address persistent vulnerabilities[s].

Consumer GPU Supply Constraints

The AI infrastructure buildout has created measurable supply constraints for consumer hardware. NVIDIA reportedly reduced RTX 50-series Blackwell consumer production by 30 to 40% in H1 2026 because consumer GDDR7 manufacturing capacity competes with HBM production lines[s].

Analyst predictions suggest 2026 may be the first year in three decades without a new NVIDIA GeForce generation[s]. Bernstein Research analyst Stacy Rasgon summarized the dynamic: “Every bit of memory that’s out there, I think, is really getting prioritized to AI compute. That dynamic is starving the overall industry of the type of memory that is traditionally used for more consumer type applications.”[s]

Emerging Supply Chain Risks

New bottlenecks continue to emerge. Tungsten prices have surged 557% in just over a year following China’s addition of the metal to export control lists[s]. Tungsten’s high melting point and density make it essential for both chips and manufacturing equipment at advanced nodes. China controls 79% of global tungsten mine production.

Trade channels have also evolved. NVIDIA restarted H200 sales to approved Chinese customers after a ten-month freeze, under a December arrangement granting the US government a 25% revenue share[s]. Chinese firms have placed orders for over two million H200 units for 2026 delivery.

The semiconductor sovereignty race has entered a phase where interdependence persists even as nations attempt to reduce it. The question is whether $165 billion in combined US and allied investment can shift the manufacturing center of gravity before the next supply shock arrives.

How was this article?
Share this article

Spot an error? Let us know

Sources