The $1 Trillion Chip Market and the One-Legged Economies It Created

The global semiconductor market will surpass $1 trillion in revenue in 2026 — a milestone the industry was not expected to reach until 2030. The World Semiconductor Trade Statistics consortium now forecasts $1.5 trillion, a 90% leap from the prior year, driven almost entirely by demand for the chips that train and run artificial intelligence models. Memory semiconductors alone are expected to jump roughly 3.5-fold. It is the fastest growth rate since 1995 and, in absolute terms, the largest single-year revenue expansion in the industry’s history.

But this historic milestone disguises a structural vulnerability. In four of the world’s largest economies — South Korea, China, Taiwan, and the United States — strip out semiconductors and AI-related spending, and growth either halves or disappears entirely. The chip supercycle is not lifting all boats. It is inflating a single mast on economies whose hulls are taking water. Economists in Seoul have a term for this: “one-legged growth.” In 2026, it describes not just South Korea, but an entire tier of the global economy.

South Korea: The 53% Headline and the 1.7% Reality

South Korea’s export data for May 2026 looked spectacular: $87.8 billion, a record, up 53.2% year-over-year. The Bank of Korea raised its growth forecast from 2.0% to 2.6%, and Governor Shin Hyun-song publicly credited AI-related chip demand with adding 0.7 percentage points to GDP growth — more than offsetting the 0.4 percentage point drag from higher oil prices.

The disaggregated numbers tell a different story. Semiconductor exports alone surged roughly 170%, blasting past $37 billion in a single month. Samsung Electronics and SK Hynix, which together dominate the global market for high-bandwidth memory (HBM) — the specialised DRAM that powers AI training clusters — are operating at full capacity and booking orders quarters in advance. But strip out semiconductors, and South Korea’s export growth collapses to 1.7%. Automobiles, refining, steel, and petrochemicals — the sectors that employ far more people and sustain far more of the country’s industrial base — are struggling.

The concentration is starker than even the 2021–2022 semiconductor cycle. In that period, chips were a significant contributor to Korean exports but not the overwhelmingly dominant one. In May 2026, semiconductors accounted for more than 40% of total exports by value. The BOK’s own projections imply that without the AI chip boom, South Korea’s growth rate would be closer to 1.9% than 2.6% — respectable, but well below the level the headline data suggests. The question is not whether Korean exports are strong but whether “Korean exports” is any longer a meaningful aggregate, or whether it is really just a proxy for two companies’ order books.

China: AI Exports Doubled. The Domestic Economy Collapsed.

China’s semiconductor export surge is perhaps the most counterintuitive story in global trade. Despite US export controls designed to restrict China’s access to advanced AI chips, Chinese integrated circuit exports reached $43.3 billion in January and February 2026 — up 72.6% year-over-year. By April, monthly IC exports hit approximately $31 billion, a doubling from the prior year. Semiconductor exports climbed 110% year-over-year in the first five months of 2026. Goldman Sachs and Nomura estimated that semiconductors, computers, and other AI-related products accounted for roughly half of China’s total export growth.

This is partly a story of domestic substitution. US restrictions have channelled Chinese buyers toward local suppliers — Huawei’s Ascend AI chip line, whose revenue is projected to surge 60% to $12 billion in 2026, and ChangXin Memory Technologies (CXMT), which saw a 130% revenue jump to over $8 billion. But it is also a story about China’s role as a link in the broader supply chain: many of the “Chinese” chip exports are assembled or tested on the mainland using imported wafers, then re-exported as finished ICs to global data centre builders.

The domestic picture is a mirror image. In the first five months of 2026, fixed-asset investment contracted 4.1% year-over-year — exceeding market expectations of a 2.0% decline and worsening from the 1.6% drop recorded in January–April. Property investment, still the economy’s deepest wound, fell 16.2%, accelerating from 13.7% in the prior period. Manufacturing capacity utilisation sank to 73.9%, near a decade low outside the early-2020 lockdowns. Youth unemployment hit 16.9%. Retail spending remains deflationary. In essence, China is operating two economies simultaneously: an export economy growing at double digits, powered by AI chip demand, and a domestic economy in its worst contraction since the 1990s.

Taiwan: When One Company Is the Economy

Taiwan takes the concentration risk to its logical extreme. TSMC — the Taiwan Semiconductor Manufacturing Company — generated NT$1.13 trillion ($35.6 billion) in revenue in the first quarter of 2026, up 35% year-over-year. Net profit rose 58%, beating analyst estimates. Advanced manufacturing processes (7nm and below) accounted for 74% of total wafer revenue, with 3nm alone representing 25%. The company’s annual revenue is forecast to grow 30% in US dollar terms for the full year. TSMC has announced plans to spend up to $56 billion on new fabrication facilities, including major expansions in the United States and Japan.

The structural question for Taiwan is what fraction of the island’s economic identity is TSMC. The company is the world’s largest contract chipmaker, fabricating the most advanced processors for Apple, Nvidia, AMD, Qualcomm, and an expanding list of AI-focused clients. Its revenue alone is equivalent to roughly 15% of Taiwan’s GDP. When TSMC has a strong quarter, Taiwan’s export data surges. When TSMC guides conservatively, Taiwan’s growth forecasts are revised downward. No other advanced economy has this level of dependence on a single company for its macroeconomic trajectory.

This creates a geopolitical dimension that other one-legged economies do not face. Taiwan sits at the centre of the US–China technology decoupling, and TSMC’s fabrication capacity is the single most strategically significant industrial asset on earth. The company’s planned $56 billion in offshore expansion is partly a hedge against this risk. But in 2026, the immediate concern is simpler: Taiwan’s economic growth is, for practical purposes, TSMC’s revenue growth. And TSMC’s revenue growth is, for practical purposes, AI chip demand.

The United States: AI Capex Carried Q1. Consumer Spending Nearly Stalled.

The BEA’s third estimate of US Q1 2026 GDP, released on June 25, told two stories. The headline was reassuring: 2.1% annualised growth, revised up 0.5 percentage points from the prior estimate. Beneath it, the composition was striking. Business investment in equipment surged 15.8%, and intellectual property products rose 13.8% — categories dominated by AI-related capital expenditure. Consumer spending, meanwhile, was revised sharply downward to 0.5% — well below the 1.4% initially reported. Services demand dropped from 1.8% to 0.5%.

The five hyperscalers — Amazon ($200 billion), Google ($175–185 billion), Microsoft (~$150 billion), Meta ($115–135 billion), and Oracle (~$65 billion) — are collectively spending approximately $725 billion in capital expenditure this year, nearly all of it on AI data centres, GPU clusters, and the infrastructure to power them. Pantheon Macroeconomics calculates that without AI-related investment, US corporate equipment spending would be negative. The Atlanta Fed’s GDPNow tracker estimates Q2 growth at 2.5%, but the composition question persists: how much is broad-based economic activity, and how much is five technology companies building data centres?

The US consumer is no longer the reliable engine that powered post-pandemic recovery. The Conference Board’s Leading Economic Index stands at 99.3, with 6-month and 12-month rates still negative. Consumer spending growth of 0.5% — barely above zero in real terms — is the weakest quarterly print since Q4 2022. If AI capex were to decelerate meaningfully, the US economy does not have a second growth engine running fast enough to compensate.

The Downstream Cost: 600,000 Fewer Cars

The semiconductor supercycle is not a victimless boom. As AI data centres consume an ever-larger share of global chip production, other industries are being squeezed out. AI data centres are projected to consume 70% of all memory chips produced by 2026, and the competition for foundry capacity has become a zero-sum game between hyperscalers and automakers.

The economics are straightforward. A single AI training cluster generates more revenue for a chipmaker than thousands of automotive-grade microcontrollers. Foundries like Samsung, SK Hynix, and Micron have accordingly reallocated manufacturing capacity toward high-margin AI chips, deprioritising automotive orders. UBS estimates that up to 600,000 fewer vehicles may be built in 2026, with disruptions escalating into significant production halts by 2027 and 2028. This is not a repeat of the 2021–2024 pandemic-era chip shortage, which was caused by supply chain disruptions and just-in-time inventory failures. This shortage is structural — a permanent reallocation of capacity toward the industry’s highest-value customers.

The irony is acute. The countries most dependent on the semiconductor supercycle — South Korea, Japan, Germany — are also major automobile producers. South Korea’s auto exports, once a pillar of its trade surplus, have stagnated in 2026 precisely because the same semiconductor boom that inflates its headline export data is redirecting chips away from vehicle production. For Germany, already reeling from an industrial recession, the chip reallocation adds another headwind to a manufacturing sector that has shed 157,000 jobs in the past year.

The $1 Trillion Milestone: What It Means and What It Hides

The semiconductor industry crossing $1 trillion in annual revenue is, by any measure, a landmark. It places semiconductors alongside oil and gas, automotive, and financial services as one of the world’s trillion-dollar industries. The milestone arrives four years ahead of the consensus 2030 forecast, driven by a demand profile that no one predicted at this scale even two years ago. SEMI, the industry body, has described it as “the most significant inflection point since the invention of the integrated circuit.”

But the composition of that trillion dollars reveals extreme concentration. High-value AI chips — the GPUs, HBM modules, and advanced logic processors that power training and inference — drive roughly half of total revenue. Yet they represent less than 0.2% of total chip volume. The vast majority of semiconductors shipped in 2026 are the same commodity microcontrollers, sensors, and analog chips that go into cars, appliances, and industrial equipment. Those segments are growing at low single digits or not at all. The trillion-dollar market is, in essence, a $500 billion AI chip market stapled to a $500 billion legacy market that is barely expanding.

This matters because the revenue concentration mirrors the geographic concentration documented above. The countries and companies riding the AI chip wave are experiencing a supercycle. The countries and sectors that consume the other kind of chip are experiencing something between stagnation and shortage. The headline number — $1 trillion — makes the industry look healthier and more balanced than it is.

What Happens If AI Demand Decelerates?

The question that hangs over every one-legged economy is the same: what if the leg buckles? In the case of the semiconductor supercycle, this translates to a specific scenario — what happens if hyperscaler capital expenditure slows?

There is precedent. In June 2026, AI chip stocks experienced a $1.4 trillion crash before recovering, a volatility event that briefly wiped out more market capitalisation than the entire GDP of Australia. The trigger was not a change in fundamentals but a momentary question about whether AI revenue would justify the scale of investment. If that question is answered negatively — if the AI revenue models underpinning $725 billion in annual hyperscaler capex prove slower to materialise than expected — the consequences would ripple through the real economy of every country examined in this analysis.

South Korea’s export growth would collapse from 53% to low single digits overnight. China’s export engine, already the only thing preventing a deeper domestic contraction, would stall. Taiwan’s GDP growth would be revised downward in proportion to TSMC’s guidance cut. And in the United States, with consumer spending at 0.5% and without the AI investment engine, Q1 GDP growth would have been closer to zero than to 2.1%.

The historical parallel is instructive. The railroad boom of the 1870s, the fibre-optic buildout of the late 1990s, and the shale oil investment surge of 2014–2015 all featured massive capital expenditure cycles driven by transformative technologies. In each case, the investment eventually outran near-term revenue, producing a correction that hurt not just investors but the economies that had become dependent on the spending. The AI chip cycle differs in one important respect: the investment is more concentrated. Five companies account for approximately 60% of US nonresidential fixed investment growth. Two companies dominate HBM. One company fabricates roughly 90% of the world’s most advanced logic chips. Concentration at this level means that a deceleration would not be gradual; it would be sudden and synchronised across multiple economies.

None of this means the AI chip supercycle will end in 2026 — the demand signals remain strong, and the technology is genuinely transformative. But the macroeconomic vulnerability is real. When four of the world’s five largest economies derive their headline growth from the same sector, the global economy has a concentration risk that no amount of headline GDP data can obscure. The country rankings look healthy. The dependency structure behind them does not.

Data sources: WSTS June 2026 forecast; SEMI; TechInsights; Korea Customs Service (May 2026); Bank of Korea (May 28 press conference); China General Administration of Customs (Jan–May 2026); Caixin Global; Goldman Sachs; Nomura; TSMC Q1 2026 earnings (April 16); BEA GDP third estimate Q1 2026 (June 25); Pantheon Macroeconomics; Morgan Stanley; UBS (auto production forecast); S&P Global Automotive; Deloitte 2026 Semiconductor Industry Outlook. Country economic data from Statistics of the World, sourced from IMF WEO April 2026 and World Bank.