Korea Moves Deeper Into the Global Market for Long-Term Capital

As Japanese yields rise, Korean government bonds enter global benchmarks and U.S. public and AI-related borrowing expands, Korea is being drawn more deeply into the portfolios that set the price of long-term capital.

Japan’s 10-year government bond yield moved above 3 percent at the start of September, taking a market that had spent most of the past quarter-century near the floor of the global interest-rate system into territory not seen since 1996. The immediate consequences are domestic: higher financing costs for a heavily indebted government, a different environment for the Bank of Japan and a meaningful sovereign yield again available to institutions whose liabilities are denominated in yen. Yet those institutions also belong to one of the largest pools of international fixed-income capital in the world. Japanese investors accumulated roughly $2.4 trillion of foreign debt during the long period in which domestic bonds offered little income, and by August 22 they had sold a net ¥3 trillion of overseas debt in 2026, the largest year-to-date reduction since 2022.

The amount is far too small relative to the existing portfolio to describe what is happening as a wholesale return of Japanese money. Insurance companies still need assets that match liabilities stretching decades into the future, pension funds operate within strategic allocation ranges, and foreign bonds continue to provide diversification and potentially superior returns. The more important change occurs when an asset matures, a currency hedge has to be renewed or new contributions have to be invested. A Japanese institution making that decision today faces a domestic bond market that looks materially different from the one under which its overseas portfolio was accumulated.

South Korea has entered that decision at almost exactly the same time. Korean government bonds began their phased inclusion in the FTSE World Government Bond Index in April and will be added through eight monthly tranches ending in November, moving won-denominated sovereign debt into portfolios tied to one of the principal benchmarks for government bonds worldwide. Foreign purchases accelerated around the beginning of the inclusion process, and Japanese investors became prominent among the buyers, with Japanese purchases of Korean bonds exceeding ₩3 trillion in both April and May and roughly ₩12 trillion accumulated over about three months by early July. Japanese institutions are therefore acquiring more Korean debt while higher JGB yields are simultaneously improving the case for holding more fixed income at home.

Those flows make the familiar idea of Japanese capital simply “coming home” too narrow. A pension fund or insurer can increase domestic bonds without abandoning foreign assets, while benchmark rules can create a separate reason to hold Korean government debt and U.S. securities remain attractive if their additional yield compensates for currency and duration risk. The same institution can make all three choices at once across different portfolios and mandates. What has changed is the set of alternatives against which each new investment is priced.

When Japanese yields become competitive again

A Treasury security with a higher headline yield does not automatically provide a Japanese institution with a higher usable return. An investor whose liabilities are primarily in yen must decide whether to leave the dollar exposure unhedged or pay to protect against exchange-rate movements, and the cost of that protection has absorbed a significant share of the additional income available on foreign bonds during periods of wide short-term interest-rate differentials. An unhedged position eliminates the hedge expense but places the value of assets against future yen liabilities at the mercy of the currency. A JGB avoids that decision altogether.

The rise in Japanese yields can therefore change relative value well before Japan offers the same nominal return as the United States. During the long zero-rate period, going abroad was often necessary for an institution seeking positive fixed-income returns without moving far down the credit-quality spectrum. A 10-year JGB above 3 percent makes the domestic option credible again, particularly for insurers and pensions whose liabilities naturally match yen assets. A survey of 82 Japanese corporate pension funds released in early September found the net share planning to increase domestic bond holdings at its highest level since the survey began in 2008, while overseas debt exposure continued to be reduced amid expensive currency hedging.

The portfolios involved are too large and too constrained to turn rapidly. Japan remains the biggest foreign owner of U.S. Treasuries, and decades of accumulated overseas holdings will not be dismantled because a domestic benchmark crossed a round number. Even the ¥3 trillion net reduction in foreign bonds this year represents only a small fraction of Japanese investors’ overseas debt. Its significance lies in what no longer happens automatically: the marginal yen that once had a powerful incentive to search abroad now has a competitive home-market destination.

That can matter without requiring a large sale of existing Treasuries. When a Japanese institution allows an overseas bond to mature and reinvests the proceeds in a JGB, no Treasury is dumped into the secondary market, but another investor must still be found for the next American security being issued. The effect grows if the same change in relative returns influences a broad group of insurance companies, corporate pensions and asset managers, even when each alters its portfolio only gradually. A global borrower that became accustomed to incremental Japanese demand during the low-yield era cannot assume that the same buyer will appear at the same price indefinitely.

Korea shows how that reduced foreign appetite can coexist with new overseas allocations. WGBI inclusion creates demand that is partly mechanical: portfolios replicating the benchmark have to acquire Korean sovereign exposure as the country’s index weight rises. Funds estimated at roughly $2.5 trillion to $3 trillion track the index, which is why analysts before the start of inclusion projected tens of trillions of won in eventual passive purchases. Japanese pension funds and insurers, already important users of global government-bond benchmarks, are among the institutions for which Korean debt has become more difficult to ignore.

The early experience also shows why index inclusion cannot be translated directly into lower Korean yields. Foreign investors can buy large amounts of Korean government bonds while domestic yields rise if inflation expectations, Korean monetary-policy expectations, U.S. Treasury yields or other risk factors move in the opposite direction. Passive managers can be buying because they have to reproduce an index while active managers, banks and arbitrage funds adjust exposure according to the price of the bond, the won and the derivatives needed to finance or hedge the position. Korea gains a broader investor base without gaining control over the return that investor base will demand.

Korean bonds enter a larger portfolio

The distinction matters because Korea is entering global fixed-income portfolios at a time when its long-term interest rates have become increasingly connected to the United States. A Bank of Korea working paper published in January examined high-frequency movements in Korean and U.S. two- and ten-year yields from 2002 through mid-2025 and found that synchronization at the long end had become more important after the global financial crisis, rising particularly during periods of international financial stress. Domestic inflation, growth, fiscal expectations and Korean monetary policy continue to matter, but U.S. movements have become large enough to affect the financial environment through which Korean monetary policy reaches the broader economy.

The researchers also found that conventional monetary-policy transmission became weaker when U.S.–Korean yield synchronization was high. Tightening in Korea produced more familiar declines in inflation and real activity during low-synchronization periods, while the estimated responses became weaker and in some cases less conventional when global financial conditions were exerting greater influence on Korean yields. The study does not imply that the Bank of Korea has lost monetary autonomy, nor does a change in the U.S. 10-year yield mechanically generate the same change in Korea. It does show why a policy rate determined in Seoul and the longer financing costs faced by Korean borrowers need not move as a single price.

The Bank of Korea confronted that environment directly when it raised the base rate from 2.75 percent to 3 percent on August 27. Its policy statement said Korean government-bond yields had fluctuated considerably in response to stronger domestic growth as well as changes in U.S. Treasury yields and international oil prices. The decision to tighten was based on Korean conditions—strong exports, recovering domestic demand, inflation expected to stay above target and persistent financial-stability concerns—but the longer end of the market remained connected to prices formed outside the country.

WGBI inclusion does not remove that exposure. It reduces some of the frictions that previously kept global institutions from owning Korean government debt and gives the market a more diversified structural investor base, but those investors still compare Korean securities with what they can earn elsewhere. A Japanese institution can weigh a KTB against a higher-yielding JGB at home. A dollar-based fund can compare a Korean government bond with a Treasury after allowing for exchange-rate and hedging costs. An active investor can sell even while index funds continue buying.

Korea can consequently attract more global savings without making long-term money uniformly cheaper. The foreign investor base can deepen at the same time that the return demanded for holding 10 or 30 years of interest-rate risk rises, because the two developments originate in different decisions. Benchmark rules determine part of where capital must go; markets continue to determine the price at which it stays there.

That distinction becomes more important when the market setting one of Korea’s key external reference prices is itself asking investors to absorb large financing requirements.

More American borrowing meets the market

The U.S. Treasury estimated in August that it would borrow $739 billion in privately held net marketable debt during the July-to-September quarter and another $628 billion during the final three months of the year. Much of that financing can be raised through bills, so the totals should not be treated as equivalent additions to long-term bond supply, but the Treasury remains a large and continuous issuer further along the curve. Its August refunding included $42 billion of 10-year notes and $25 billion of 30-year bonds, with regular note and bond auctions continuing as existing debt matures and new deficits have to be financed.

Government borrowing is only one source of securities arriving in investors’ portfolios. Corporate issuance linked to the AI investment boom has expanded sharply, with U.S. AI-related corporate bond sales exceeding $220 billion by late August, compared with roughly $12.5 billion over the comparable period a year earlier. Some of the issuers are among the world’s strongest corporate credits, but high ratings do not eliminate the need to offer investors a return above Treasuries for taking credit and liquidity risk. The rapid issuance has already forced large technology companies to pay larger concessions in some offerings, while American hyperscalers have also turned to euro-denominated bond markets as their capital programmes grow.

Treasuries and technology-company bonds are not interchangeable assets. Regulations, credit mandates, collateral requirements and liquidity needs can place them in different parts of an institutional portfolio, and it would overstate the mechanism to describe governments and hyperscalers as bidding directly for the same dollar. They nevertheless meet inside the broader balance sheets of pension funds, insurance companies and asset managers deciding how much duration, credit and currency risk to hold. A corporate bond is priced from a government yield curve, and its spread has to compensate an investor who can choose among other securities rather than accepting every new issue at the prevailing price.

The investment behind much of the new corporate financing also has a much longer physical life than the software products associated with AI. Data centres require land, buildings, servers, cooling systems and electricity connections, while semiconductor production depends on fabrication plants, advanced packaging facilities and specialized equipment. Electricity infrastructure can take longer still. The International Energy Agency estimates that more than 2,500 gigawatts of generation, storage and large-load projects are waiting in grid-connection queues worldwide; a data centre can often be built within one to three years, while grid infrastructure commonly takes five to fifteen.

Annual investment in electricity grids is already about $400 billion globally and would need to increase by roughly half by 2030 under the IEA’s assessment of expected demand. Five large technology companies spent more than $400 billion on capital expenditure in 2025, and the agency expects their combined spending to rise another 75 percent in 2026. The supporting infrastructure cannot all be financed by those technology companies themselves. Utilities, power producers, semiconductor companies and other suppliers have to commit capital to assets that may take years to build and even longer to earn back their cost.

Some of that financing comes from retained earnings, bank credit, equity or project finance rather than bonds, which is why there is no single pool of “AI capital” that can be measured against Treasury issuance. The common feature is a large increase in spending on long-lived productive assets whose financing decisions occur before their eventual economic returns are known. Investors are being offered more ways to earn income by financing that buildout just as major sovereign borrowers continue to place debt of their own.

Long-term yields still cannot be explained by supply alone. Expected Federal Reserve policy, inflation, oil prices, fiscal credibility, growth expectations and the term premium all affect Treasury rates, and the global bond selloff at the start of September reflected several of those forces simultaneously. AI can complicate the picture further because enormous investment spending can increase the present demand for capital while successful productivity gains could raise potential growth and equilibrium real rates later. The eventual magnitude of that productivity effect remains uncertain, but the capital expenditures and securities being issued to finance the buildout are already occurring.

For Japanese institutions, the timing matters. More American public and corporate securities are reaching markets at a moment when Japan’s domestic yield curve offers an increasingly credible alternative, while Korean government bonds are acquiring a benchmark-driven place in international portfolios. The marginal buyer of an American long-term security has not disappeared; that buyer simply has more attractive alternatives against which to price the next issue.

AI capital comes back through semiconductors

Korea encounters the same investment cycle through a second route that points in the opposite direction. U.S. and global spending on AI computing creates demand for high-bandwidth memory, advanced packaging and other semiconductor products in which Korean manufacturers occupy critical positions, turning capital expenditure abroad into export revenue, corporate profits and investment inside Korea. When the Bank of Korea raised its 2026 growth forecast from 2.6 percent in May to 3.3 percent in August, it cited a semiconductor sector that had performed substantially better than expected and projected exports and investment to maintain high growth.

SK hynix made the financial connection unusually visible in July. The company issued 17.79 million new common shares underlying American Depositary Shares through Citibank, raising $26.5071 billion from overseas investors in an offering whose proceeds were designated for facilities. A week later, its board approved ₩7.0931 trillion in construction investment for the P&T7 advanced-packaging facility in Cheongju, with the investment period extending through the end of 2032 and the stated purpose of securing production capacity in response to global demand for AI memory semiconductors.

The two transactions should not be treated as though every dollar raised in the ADR sale was earmarked directly for P&T7, and the offering was equity rather than the corporate debt being issued by U.S. hyperscalers. Their connection lies further upstream. Capital markets were willing to finance a Korean producer building physical capacity because global AI investment had increased the expected demand for what that capacity would produce. A technology investment cycle centred heavily in the United States had become large enough to generate multiyear plant construction in Cheongju and to bring tens of billions of dollars of foreign capital directly onto a Korean corporate balance sheet.

The proceeds then crossed another market. South Korean foreign-exchange authorities purchased roughly $20 billion of the dollars SK hynix repatriated from the ADR offering through over-the-counter transactions conducted by the Foreign Exchange Stabilization Fund. The operation allowed a very large conversion to be handled without sending the entire flow through the spot won market and became part of the authorities’ reserve and currency management. Investor capital raised in New York had moved through a Korean semiconductor company, into domestic facilities and equipment, and then into a foreign-exchange transaction large enough to involve the state.

Korea was therefore not simply importing higher U.S. bond yields while benefiting from an unrelated technology export boom. Both originated partly in the same expansion of long-lived capital formation abroad. AI infrastructure was adding to financing activity in American corporate markets while producing demand for Korean semiconductors; U.S. Treasury yields were influencing Korean long-term rates while American and other foreign investors were supplying equity to Korean chipmakers. The two channels reached different balance sheets and could strengthen the Korean economy and tighten its financial conditions at the same time.

That combination helps explain the setting in which the Bank of Korea raised rates in August. Strong semiconductor exports and investment had lifted growth and improved income conditions, contributing to a recovery in domestic demand. The central bank projected consumer inflation of 2.7 percent in 2026 and core inflation of 2.5 percent, both above its 2 percent target, while Seoul and surrounding housing markets remained strong and household borrowing continued to rise. Household credit had reached ₩2,019.8 trillion at the end of June after increasing ₩25.9 trillion in the second quarter.

A 3 percent policy rate acts on those domestic conditions, but the economy around it is financed through several markets simultaneously. Foreign benchmark investors are buying government bonds, international equity investors are funding semiconductor capacity, exporters are earning dollars from the AI cycle and Korean long-term yields are responding partly to conditions in U.S. markets. Monetary policy has to operate inside those flows rather than separating Korea from them.

Korea occupies a different position from either of the two large economies at the ends of this story. It does not issue the global reserve asset or a Treasury market capable of setting the principal international risk-free benchmark, and its savings do not play the same overseas creditor role as Japan’s enormous institutional portfolios. Yet its government debt is becoming a more regular holding in the portfolios through which Japanese and other global investors allocate sovereign capital, while its leading industrial companies are drawing another form of international capital because of the productive investment those portfolios are helping to finance elsewhere.

The country is increasingly involved in both the allocation of long-term savings and the industrial income produced by what those savings build.

Cross-border finance has not contracted while these developments have taken place. Bank for International Settlements data show that cross-border bank credit increased by $1.7 trillion in the first quarter of 2026 to $39.5 trillion, 11 percent higher than a year earlier, with credit expanding to both banks and non-bank borrowers. The measure includes loans, deposits and banks’ holdings of debt securities rather than a homogeneous pool of money available for 10- or 30-year financing, so it cannot tell us whether long-duration capital is abundant by itself. It does make a simple explanation based on the world running out of money difficult to sustain.

A short-term dollar loan, a Treasury bill and a 30-year sovereign bond all transfer financial resources, but they require very different commitments from the institution on the other side. A long-term bond exposes its owner to years of changes in inflation, interest rates and market prices, while an international buyer may also have to bear exchange-rate risk or pay repeatedly to hedge it. Pension liabilities, insurance obligations, regulation, benchmark rules and internal risk limits determine which balance sheets can hold those exposures and what return they demand before doing so.

Large pools of savings can therefore coexist with expensive long-term borrowing. The amount of capital in the system says little by itself about the currency into which the capital can move, the length of time for which its owner is prepared to commit it or the risk that has to be absorbed along the way. Japan’s recent changes make that separation particularly visible because the savings themselves have not disappeared; the return available without sending them abroad has improved.

Korea’s WGBI inclusion changes another part of the allocation without creating new global savings. It gives benchmark investors a reason to redirect some existing capital toward Korean sovereign debt. The United States, meanwhile, continues to issue government securities and finance an investment boom that is generating both new corporate securities and new productive capacity. Capital remains available to all of them, but the terms on which investors are prepared to supply it are being renegotiated against a more competitive range of alternatives.

The evidence does not establish that long-term interest rates must remain permanently high. Inflation can decline, fiscal borrowing can change, AI investment can slow if expected returns disappoint, and institutions elsewhere can replace purchases that Japanese investors make less readily. Japanese pensions and insurers may continue to keep large foreign portfolios even with higher yields at home because diversification and currency conditions still matter. The recent shift is clearer in relative incentives than in any forecast of the eventual level of yields.

For most of the period when Japanese sovereign yields hovered near zero, global fixed-income markets could rely on a powerful recurring incentive: one of the world’s largest pools of savings had good reason to leave home in search of income. That incentive has weakened without disappearing. Korean sovereign debt has acquired a formal place in the global benchmarks through which some of that money travels, while American governments and companies continue to ask investors to finance programmes extending many years into the future.

Korea now sits close to each part of that movement. Japanese and other foreign money is entering its government-bond market, U.S. long-term rates continue to influence the price of Korean debt, and international capital is arriving through semiconductor companies building capacity for the same AI expansion helping to increase financing demand abroad. The next change in a central-bank policy rate will alter those comparisons, but it cannot recreate the portfolio environment that existed before the alternatives changed. Long-term capital remains abundant enough to move among Tokyo, Seoul and New York; what is becoming less uniform is the set of terms on which its owners are prepared to leave it there.