Where wealth flows: where chip money gathers

In 2026's first half Korea at 496.3 and Taiwan at 416.6 billion dollars passed Japan; the 2025-2032 buildout is projected at 10.3 trillion dollars. Home bills arrive late.

The picture changed in the first half of 2026. South Korea exported 496.3 billion dollars. Taiwan reached 416.6 billion dollars. Japan stayed at 384.4 billion dollars. Both passed Japan for the first time. Chip exports made up 30 percent of Korean and Taiwanese totals. In Japan the rate stayed at 5 percent. The numbers belong to the January-June period. The comparison base matches. The period matches. The unit matches. So the gap is real. The ranking belongs to one half-year. It means no permanent order. It means a photograph. Let us read the photograph right. [1] [2]

How does this table touch our homes. Not directly; it touches with delay. Chip money goes to the factory first. Then it rides the device. Then it reaches the bill. Phones get dearer. Computers get dearer. Power grids strain. Data centers want water and power. Municipalities debate it. Homeowners pay it. Macro data descends to micro life this way. It descends slowly. But it descends. Today's exports become tomorrow's prices. Let us keep this in mind.

The money's address is known too. Korean chip exports were 149 billion dollars. Taiwan printed 133 billion dollars. Japan stayed at 21.2 billion dollars. Advanced chips come from TSMC and Hon Hai in Taiwan. In Korea they come from Samsung and SK Hynix. Value gathers there. Japan is not out of the game. Its role differs. Tokyo Electron, Advantest, and Lasertec sell equipment. Japanese equipment exports were 15 billion dollars. Korea stayed at 5.2, Taiwan at 3.5 billion dollars. This is real business. But it is not the same story as finished-silicon volume. Selling machines and selling chips earn differently. Both earn. The size differs. [2] [3]

The warning sits in the same text. One half-year counts as no permanent order. The lead depends on a few big buyers. It lasts while hyperscaler customers keep spending. If currencies move, the table moves. If chip prices shift, the table shifts. Pressure rises from below too. CXMT closes the LPDDR6 gap with SK Hynix and Samsung. The Korean press read the numbers similarly. So competition never sleeps. Summits grant no comfort. Today's leader may be tomorrow's second. Trade is read this way. With periods. With units. With bases. [4] [4]

Let us read the size from America too. The buildout will total 10.3 trillion dollars from 2025 to 2032. That means 3.63 percent of US national income per year on average. A bigger ratio than the railroad, highway, and telecom booms. Yearly investment will run near 3.6 percent through 2032. Railroads and power grids fall behind. Meta and Microsoft now spend beyond cash flow. The gap closes with leases. It closes with joint ventures. It closes with private credit. It resembles the 2000s internet and fiber booms. Resembling is not equaling. But the lesson comes from the same place. [5] [6]

The researcher calls it no bubble. He does not use the word. He says the job is hard. He writes what regulators should watch. He writes what investors should watch. Debt looks hard from outside. Structures hide debt. Meta's Hyperion project was given as example. Lease debt stands off balance sheet. Outsiders cannot see it. Unseen risk cannot be measured. Unmeasured risk cannot be priced. From my home the question runs thus. On whose bill will this debt be written. The company's, the customer's, the taxpayer's. No answer yet. The question stands. [7] [8]

Let me add the counter-view too. Booms bring everyone work. Factories rise. Engineers are hired. Suppliers grow. Cities revive. In the short term winners abound. In the long term the question changes. Concentration means risk. Few customers, few countries, few companies. When one sneezes, all catch cold. Home economies run so too. A house on one salary breaks fragile. A country on one customer breaks fragile too. Variety is insurance. We know this at home. It holds in the world too. What I will watch is set. What do second-half numbers say. Does Tokyo play for the upper step. Do Seoul and Taipei make the edge lasting. Does debt turn transparent. Whose bill arrives. I watch these. I never hurry. No verdict before the period closes.

Source passages

  1. Korea, Taiwan Top Japan in H1 Exports as AI Chip Demand Surges | AI Weekly ↗
    TL;DR South Korea's H1 2026 exports reached $496.3 billion and Taiwan's $416.6 billion, both topping Japan's $384.4 billion for the first time. Integrated circuit exports made up about 30% of Korean and Taiwanese total exports in the first half, against roughly 5% for Japan. Japan still leads chip equipment, exporting $15 billion in H1 versus South Korea's $5.2 billion and Taiwan's $3.5 billion. The export scoreboard flipped in the first half of 2026, and it flipped in a
  2. Korea, Taiwan Top Japan in H1 Exports as AI Chip Demand Surges | AI Weekly ↗
    for Taiwan in the first half, roughly 30% of each economy's total exports. Japan's integrated circuit exports were $21.2 billion, about 5% of its total. The advanced chips are coming from TSMC and Hon Hai Precision Industry in Taiwan and from Samsung Electronics and SK Hynix in Korea, and that is where the export value is landing. Japan is not absent from the AI chip build-out, it just plays a different role in it. Its strengths sit in semiconductor manufacturing equipment
  3. Korea, Taiwan Top Japan in H1 Exports as AI Chip Demand Surges | AI Weekly ↗
    and materials, and firms including Tokyo Electron, Advantest and Lasertec have capitalized as fabs expand globally. Japan's exports of semiconductor manufacturing equipment reached $15 billion in the first half, well above South Korea's $5.2 billion and Taiwan's $3.5 billion. That is a real business, but it is not the same volume story as selling the finished silicon. A single-half ranking is not a permanent order of the world. The reporting is a snapshot of one especially
  4. Korea, Taiwan Top Japan in H1 Exports as AI Chip Demand Surges | AI Weekly ↗
    concentrated boom, and it does not tell you how much of the Korean and Taiwanese lead depends on continued AI capex from a small set of hyperscaler customers, or what the picture looks like when currency swings and downstream chip pricing shift again. The Korean lead is already under pressure from below, with CXMT closing the LPDDR6 gap with SK Hynix and Samsung. Korean outlet KED Global covered the same numbers with a similar framing. For anyone tracking industrial policy,
  5. The $10 trillion question: Financing the AI buildout | Brookings ↗
    AI is often described as a software story, but the boom is also a massive physical investment in data centers, power systems, and specialized chips. A new BPEA paper projects that this buildout will total $10.3 trillion from 2025 to 2032, an average of 3.63% of U.S. GDP per year and larger relative to the economy than past booms in railroads, highways, and telecommunications. On this episode of the Brookings Podcast on Economic Activity, Aaron Klein talks to the paper’s
  6. The $10 trillion question: Financing the AI buildout | Brookings ↗
    Nieuwerburgh of Columbia Business School about his new paper, “Financing the AI Build-Out.” EBERLY: The numbers here are staggering. Van Nieuwerburgh estimates the U.S. data center build-out will require something like 3.6% of GDP in annual investment through 2032, bigger relative to the economy than the railroad boom, the interstate highway system, or the build-out of the electrical grid. STEINSSON: Amazingly, companies like Meta and Microsoft are now spending more on AI
  7. The $10 trillion question: Financing the AI buildout | Brookings ↗
    investments than they’re bringing in through operating cash flow. So increasingly, they’re turning to leases, joint ventures, and private credit to make up the difference. This should actually sound familiar to those that followed previous investment booms like the internet or fiber optic booms in the early 2000s. EBERLY: Van Nieuwerburgh walks through one deal in particular, Meta’s Hyperion Data Center project, to show an example of how these structures actually work and
  8. The $10 trillion question: Financing the AI buildout | Brookings ↗
    why they can make leverage that is debt harder to see from the outside. He doesn’t label this a bubble, which is always tricky, but he does flag exactly what regulators and investors should be watching. STEINSSON: Let’s turn it over to Aaron and Stijn. [2:46] KLEIN: Thank you, Jon and Jan. It’s a pleasure to be working for you again. And hello, Stijn. Thanks for joining. VAN NIEUWERBURGH: Hi, Aaron. KLEIN: Where are you based out of?  VAN NIEUWERBURGH: I’m based out of