Everything About AI
NextFin News - Federal Reserve Chairman Kevin Warsh said on Wednesday that he pays "probably more attention to the Korean economy than you might think," casting South Korea's technology exports as a leading indicator for the artificial-intelligence investment boom now driving U.S. growth and inflation. Speaking at a press conference after the central bank's latest two-day policy meeting, Warsh said the country's shipments of semiconductors and related components are "feedstocks" for the technology surge, telling him what is "on planes and ships coming this way" to the United States.
The remark elevated a midsize export economy into the Fed's inner circle of signals — a deliberate choice by a chairman who has made the size and shape of the central bank's balance sheet a signature concern, yet who answered a question about Treasury markets by talking about cargo manifests.
The Exchange: A Balance-Sheet Question, a Supply-Chain Answer
The exchange came from a reporter for a Korean economic daily who asked how the Fed would avoid disrupting foreign official holders of Treasuries and dollar funding markets in South Korea and Japan if it shrinks its balance sheet further. It was the kind of technical question that usually produces a technical answer about reserve scarcity and repo-market plumbing. Warsh's response spent more time on why he watches Korea than on the mechanics of runoff.
"I pay probably more attention to the Korean economy than you might think because when I'm looking at um, uh, the feed stocks effectively for this new surge of technology, a lot of it's coming from Japan and Korea, some from Singapore," Warsh said. "So, I'm particularly interested in in what's happening in the Korean economy, usually because it's telling me what's what's on planes and ships coming this way."
He then pivoted to the balance-sheet review. The question is now "solely within the four walls of the balance sheet task force," he said, and recommendations will reach the Federal Open Market Committee "probably early in the first quarter," though he added, "don't hold me exactly to that timing." Any changes would follow "serious consideration" of the task force's proposals, and he said it is "unlikely to have any material effect on this policy conjuncture."
The two threads of the answer are connected, and the connection is the story. Warsh is reviewing how large the Fed's balance sheet should be — it stood at about $6.75 trillion as of October 22, down from a pandemic peak of roughly $9 trillion — while simultaneously watching the real-economy signals that tell him whether the AI-driven expansion is broadening or narrowing. South Korea, whose exports pulled decisively ahead of Japan's this year, has become one of his gauges. A chairman who watches cargo data is a chairman who does not trust lagging aggregates.
The Transmission Mechanism: Why Korean Exports Reach the Fed's Inflation Forecast
Warsh's logic runs through the supply chain, not the currency market. Semiconductors, memory chips, and advanced packaging are intermediate inputs — goods that enter other goods — so a pickup in Korean chip shipments precedes, rather than follows, the final demand that shows up in U.S. retail sales and capital-expenditure data. U.S. economic statistics arrive with lags and revisions; a Korean customs release lands within days of the month's close. If chip exports from Busan accelerate, U.S. data-center construction, server orders, and business investment are likely to follow within a quarter. If they stall, the AI capital-spending cycle that has supported U.S. growth may be losing momentum before domestic surveys detect it.
That timing advantage is why a central bank cares about a country that accounts for about 1% of global GDP. The Fed's dual mandate — price stability and maximum employment — requires it to forecast, not to record. Korean export data arrive roughly two to three weeks before comparable U.S. investment figures, giving policymakers an early read on the direction of the single most important driver of the current cycle.
The mechanism is not merely about growth. It is also about prices. A surge in demand for high-bandwidth memory and advanced logic pushes up input costs for U.S. technology producers, which can feed into equipment prices and, eventually, the goods component of inflation. Conversely, a supply glut in memory chips — a familiar pattern in the semiconductor cycle — would ease cost pressure. For a Fed chairman trying to separate a durable productivity boom from a cyclical capex spike, the Korean export print is a low-lag, high-signal variable. It is, in effect, a canary for the AI capex cycle, and Warsh is telling the market which cage he is watching.
The Numbers Behind the Gauge
The data justify the attention. South Korea's exports reached $693.6 billion in the first eight months of 2026, surpassing Japan's $517.4 billion by $176.2 billion. The growth rate, 52.9% year-on-year, was nearly three times China's 18.3% over the same period. In 2005, Korea's exports of $284.4 billion were less than half — 48% — of Japan's $595.3 billion; twenty-one years later, the order has inverted and the gap has widened.
Semiconductors are doing almost all of the heavy lifting, and the concentration is extreme. In August, chip exports surged 209% from a year earlier to a record $46.65 billion, accounting for 47.5% of the country's $98.25 billion in goods exports, according to the Ministry of Trade, Industry and Resources. Jeff Ng, head of Asia macro strategy at Sumitomo Mitsui Banking Corporation, estimated that semiconductors accounted for nearly 80% of August's export growth. Computer exports, which ride the same AI wave, jumped 419.5% to $6.24 billion.
That concentration is the flip side of the signal's clarity. A gauge that depends on one product category can flash brightly while the rest of the economy dims. Automobile exports fell 29.8% in August, and the Bank of Korea has been raising rates to contain inflation — the base rate reached 3% in August after two consecutive increases. Non-semiconductor exports did climb 20% in August, but the asymmetry is unmistakable: one sector is carrying the economy.
Dave Chia, an economist at Moody's Analytics, framed the risk plainly: "A gradual slowdown would be manageable. An abrupt stall is a different matter, because the economy already runs at two speeds, and the sectors that would need to take up the slack are the ones under pressure today." The two-speed description matters for Warsh's purpose. A feedstock gauge works best when the underlying economy is diversified enough that the signal reflects broad external demand rather than a single industry's inventory cycle.
The Balance-Sheet Question He Deferred
The reporter's actual question was about the Fed's balance sheet and the risk of disrupting foreign buyers of Treasuries — a genuine concern for Korea and Japan. Japan is the largest foreign holder of U.S. government debt, with about $1.12 trillion in Treasuries as of June 2026, according to Treasury International Capital data; South Korea is also a major foreign official holder. A disorderly runoff that lifts their dollar-funding costs would ripple back to U.S. rates, which is precisely why the question was asked.
Warsh's answer deferred the substance to the task force, and the deferral is itself informative. The balance-sheet task force is one of five panels Warsh appointed after taking office on May 22 to reconsider how the Fed operates, alongside reviews of inflation frameworks, data, productivity and jobs, and communications. The panels are co-led by external advisers — economists, business leaders, and former central bankers — and report to the FOMC. Analysts at the Brookings Institution called the review "the most consequential review of U.S. monetary policy at least since the Federal Open Market Committee adopted a formal inflation target in 2012."
Warsh's framing — "unlikely to have any material effect on this policy conjuncture" — is a deliberate downplaying of near-term risk. With the balance sheet at roughly $6.75 trillion, or about 21% of nominal GDP as of March, the question is not whether the portfolio is large but how fast it should shrink without straining the plumbing of the dollar funding system. The task force studies the regime; the FOMC sets the path. That division lets Warsh pursue a smaller balance sheet over time without committing to a pace that would spook foreign holders today. Recommendations are due to the FOMC early in the first quarter, which means the window for market guidance opens in the first months of next year.
There is also a personal thread in the exchange. Warsh noted that South Korea's new central bank governor is "someone I've known for 20 years," and that they met "amid the global financial crisis as it would come to be known." He was referring to Shin Hyun-song, the former head of the Bank for International Settlements' Monetary and Economic Department, whom President Lee Jae-myung nominated on March 22 to succeed Rhee Chang-yong when his term ends on April 20. A Fed chairman with a two-decade relationship to the incoming governor of a major Treasury-holding central bank is a channel, not just a courtesy.
Cyclical or Structural: The Call That Determines How Much Weight the Fed Places on Korea
The deeper question beneath Warsh's Korea comment is whether the AI-driven export boom is cyclical or structural — and the answer determines how much weight the Fed should place on it. Getting this wrong flips the conclusion. If the boom is cyclical, the Korean signal will mean-revert and the Fed would be wrong to treat it as a durable growth impulse. If it is structural, the signal deserves a permanent place in the reaction function.
The cyclical case is strong on the surface. Semiconductor demand has always been a boom-bust industry, and the 209% year-on-year jump in Korean chip exports is the kind of spike that reverses as capacity catches up. Memory prices have collapsed and rebounded in cycles measured in quarters: the 2016-2017 upcycle that preceded the 2019 downturn, the 2020-2021 pandemic shortage that gave way to the 2023 inventory correction. Korea's export concentration — nearly half of all goods shipments in a single category — is a classic cyclical vulnerability. Base effects alone guarantee that the 200%-plus growth rate cannot persist into 2027.
The structural case, however, is stronger. This is not a generic memory upcycle. The demand is coming from a step-change in computing architecture: data centers built for generative AI require an order of magnitude more memory and bandwidth than the cloud build-out of the 2010s. Korea's advantage is not just capacity but position — it sits at a chokepoint in the supply chain for high-bandwidth memory and advanced packaging that cannot be replicated quickly. Japan and Singapore anchor adjacent links, which is why Warsh grouped the three together as feedstocks rather than treating Korea as a standalone export story. The AI compute build-out is a multi-year transition, not a single-year event, so the floor beneath Korean tech exports has moved up structurally even as the amplitude of the 2026 spike is cyclical.
The verdict: the cycle is superimposed on a structural shift. For the Fed, that means Korean data should be read as a direction signal with cyclical noise stripped out, not as a level to extrapolate. A chairman who understands this will not ease on a single soft Korean print, and will not treat a hot print as proof of durable acceleration. He will watch the slope, not the level.
The Counter-Thesis: A Feedstock Gauge Measures Throughput, Not Final Demand
The strongest argument against Warsh's approach is that a feedstock gauge captures supply-chain throughput, not final demand — and final demand is what drives inflation. Korea can ship record chips into a U.S. inventory build that never reaches consumers. The 2026 export surge may reflect customers front-running anticipated shortages rather than end-user purchases, in which case the "planes and ships" signal leads to a false positive: the Fed sees strength that is really restocking.
There is also the concentration problem. If nearly 80% of Korea's export growth comes from one sector, then watching Korea closely is, in effect, watching one industry's order book and calling it the economy. A central bank with a dual mandate needs breadth — wage growth, services inflation, consumption — that a chip export print cannot provide. A gauge that worked during a single-sector boom may mislead when the cycle rotates.
Warsh's answer to this is implicit in his wording. He said Korea tells him what is "coming this way" — a flow indicator, one input among several — not what the U.S. economy is doing today. The falsifying signal is specific and observable: if Korean semiconductor exports contract for two consecutive quarters while U.S. business investment in information-processing equipment continues to rise, then the feedstock gauge has decoupled from U.S. demand and should be downgraded. The opposite also holds: if Korean exports hold up while U.S. capex stalls, the signal is telling the Fed that the slowdown is inventory-driven, not demand-driven. A gauge that cannot be proven wrong is not a gauge; it is a narrative.
Second-Order Implications: A Korea-Watching Fed Changes the Reaction Function
The first-order reading of Warsh's comment is that Korea matters to the Fed. The second-order implication is that a Fed chairman who watches Korean customs data is a more data-dependent chairman, and that has consequences for how the market should read every Fed communication from here.
If Korean chip exports cool, do not expect Warsh to treat it as a reason to ease unless U.S. investment data confirm the turn. If they accelerate, expect heightened vigilance on input-cost inflation even if the headline consumer numbers are quiet. The market's conventional wisdom is that the Fed reacts to U.S. data; Warsh is signaling that the Fed is now watching the global supply chain that precedes those data. That widens the set of variables that can move rates, and it raises the cost of being wrong about the AI cycle.
There is also a cross-market implication for the dollar funding system that prompted the question in the first place. Japan and South Korea are major foreign official holders of Treasuries, and both are navigating a transition in central-bank leadership — Japan under a governor managing yen volatility, Korea under Shin Hyun-song, a BIS veteran with a financial-stability mandate. The balance-sheet task force's recommendations, due early in the first quarter, will be read as closely in Seoul and Tokyo as in Washington. Warsh's reassurance that changes are "unlikely to have any material effect on this policy conjuncture" should be read as reassurance on the near-term rate path, not as abandonment of the smaller-balance-sheet goal.
What to Watch: Scenarios for the Next Six Months
Three scenarios frame the path ahead, each tied to an observable trigger:
- Base case: Korean chip exports moderate from the 200%-plus pace but remain positive as AI infrastructure spending continues. The Fed holds rates steady and lets the task force complete its work, with guidance arriving early in the first quarter.
- Upside case: exports stay elevated through the first quarter, confirming a durable capex cycle. The Fed stays alert to input-cost inflation but keeps policy unchanged absent broader price pressure. The balance-sheet review proceeds without market disruption.
- Downside case: an abrupt stall in chip demand coincides with weakness in Korea's non-semiconductor sectors. The feedstock signal flips negative, and the Fed faces a growth scare layered on sticky services inflation — the two-speed economy that Moody's warned about becomes a policy problem.
The single most important number to watch is the monthly Korean customs release on semiconductors — specifically whether the year-on-year growth rate stabilizes above zero after base effects roll through. That number, more than any headline, will tell Warsh whether the AI boom is broadening or peaking. The second is the task force's timeline: any slippage beyond "early in the first quarter" would signal disagreement inside the FOMC about the balance-sheet path.
Warsh's message is that in an economy powered by chips, a country with a concentrated export print can tell a central bank what the world's largest economy is about to do — if you know how to read the manifest. The market should read it the same way.
Explore more exclusive insights at nextfin.ai.
