One thing I’ve noticed from interviewing thousands of people is that everyone has huge areas of vulnerability. A man who is strong in one situation is helpless in another situation.<\/p>\n
There might be some parallel between individuals and countries. <\/p>\n
Oh the humanity!<\/p>\n
The “squishy underbelly” is the US government’s dependence on cheap borrowing, especially its dependence on foreign investors and allied governments continuing to hold and buy US Treasury debt.<\/p>\n
The column’s argument<\/a> is that American power rests on a contradiction. The US can pressure other countries on trade, currencies, defence and geopolitics, but it also runs enormous fiscal deficits and needs those same countries to keep financing them.<\/p>\n Japan makes the vulnerability unusually visible. If Japan wants to strengthen the yen, two obvious options are:<\/p>\n Raise Japanese interest rates sharply. That makes Japanese government bonds more attractive relative to Treasuries, encouraging Japanese money to come home. Either route can raise American borrowing costs.<\/p>\n That matters because the US federal government is already borrowing so heavily. A movement from roughly 4 percent to 4.6 percent on the 10-year Treasury, with the 30-year above 5 percent, is not merely a Wall Street inconvenience. Higher yields progressively increase the government’s interest bill, worsen the deficit, require still more borrowing, and potentially force harder choices about taxes, spending and monetary policy.<\/p>\n So Martin’s point is stronger than “America has too much debt.” It is:<\/p>\n American geopolitical freedom of action depends partly on the willingness of foreigners to finance American fiscal policy.<\/p>\n That gives creditors and major holders of dollar assets a kind of latent leverage. They cannot casually dump Treasuries without hurting themselves, which is why the dollar system remains extraordinarily powerful. But neither can Washington behave as though foreign demand for its debt is irrelevant.<\/p>\n The yen episode exposes this neatly. Washington ostensibly intervened to help Japan. Yet the intervention was structured so that Japan could support the yen while limiting immediate Treasury sales. In other words, the US was helping Japan solve its currency problem in a way that protected the US Treasury market.<\/p>\n That is the tell.<\/p>\n The deeper “squishy underbelly” is therefore the bond market. More precisely, it is the combination of very large structural fiscal deficits, rising interest costs, dependence on continued global demand for Treasuries, and the need to preserve confidence in the Federal Reserve and dollar system.<\/p>\n You could compress the whole column into one sentence:<\/p>\n The US can bully much of the world economically only so long as much of the world remains willing to lend it money cheaply.<\/p>\n Americans have been warned about an impending fiscal reckoning for roughly half a century. There was the inflation and deficit anxiety of the 1970s, Reagan’s “twin deficits” in the 1980s, Ross Perot’s debt clock in 1992, the bond vigilantes, warnings that China could dump Treasuries in the 2000s, post-financial-crisis warnings about quantitative easing, and then another explosion of debt after Covid. Yet the predicted funding crisis repeatedly failed to arrive.<\/p>\n So I would separate three propositions.<\/p>\n First, “the US depends on foreigners buying its debt” is not a new discovery. It is actually somewhat less true than it once was. Foreign investors held about 32 percent of marketable Treasury debt at the end of 2025. That share peaked around 55 percent in 2008 and has been roughly stable since 2020. The FT’s rhetoric about America needing “the world to buy its bonds” is therefore true in a loose sense but overstates the specific dependence on foreign governments.<\/p>\n Second, there really is something different about the present fiscal arithmetic. CBO estimates debt held by the public at about 101 percent of GDP in 2026, compared with deficits averaging just 3.8 percent of GDP over the previous 50 years. The 2026 deficit is projected at 5.8 percent of GDP despite the economy not being in recession. Net federal interest costs are about $1 trillion, 3.3 percent of GDP, versus a 50-year average of 2.1 percent. CBO projects them reaching $2.1 trillion and 4.6 percent of GDP in 2036.<\/p>\n That is the change I would take seriously. The problem is no longer simply “the debt is big.” The US is running something resembling recession-sized deficits during ordinary economic conditions while refinancing an enormous stock of debt at substantially higher interest rates.<\/p>\n Third, even that does not mean a Treasury crisis is imminent. The United States possesses advantages that countries featured in traditional sovereign-debt crises do not. It borrows in its own currency. The dollar is the principal reserve currency. Treasuries constitute the basic safe collateral of much of the international financial system. The Treasury market is huge and liquid. If Japan sells a billion dollars of Treasuries, it doesn’t remove a billion dollars of American “financing.” It lowers the price until somebody else finds the yield attractive enough to buy them.<\/p>\n That distinction matters enormously.<\/p>\n The real danger is not:<\/p>\n “One day Japan and China stop lending us money and America goes broke.”<\/p>\n It is:<\/p>\n “The marginal buyer increasingly demands a higher yield to absorb enormous and continually growing Treasury issuance.”<\/p>\n Then the adjustment occurs through price. Treasury yields rise. Mortgage and corporate borrowing costs rise. Federal interest expense rises. The government has to issue still more debt to pay the interest. Eventually fiscal policy begins competing with monetary policy.<\/p>\n And that is where today’s situation becomes more interesting than the warnings of 1975 or 1985.<\/p>\n There’s an especially nasty feedback mechanism:<\/p>\n higher debt \u2192 greater Treasury issuance \u2192 higher required yields \u2192 higher federal interest expense \u2192 larger deficits \u2192 still greater Treasury issuance.<\/p>\n For decades, declining interest rates concealed much of the deterioration in the government’s balance sheet. Washington could increase debt without anything proportionate happening to the interest bill. That trick becomes much less effective when refinancing occurs around 4 to 5 percent rather than 1 to 2 percent.<\/p>\n There is another important qualification to Martin’s argument. Foreign official institutions actually appear to be gradually reducing some Treasury exposure already. Treasury securities held in custody at the Fed for foreign official accounts fell from about $2.91 trillion in 2022 to $2.67 trillion in May 2026. Nothing catastrophic happened. Other buyers absorbed the debt.<\/p>\n So I wouldn’t read the yen intervention as America suddenly “baring its financial weak spot.”<\/p>\n I would read it as a small piece of evidence for a much narrower proposition: US policymakers are becoming increasingly sensitive to anything that could push long-term Treasury yields higher.<\/p>\n The interesting threshold is not when foreigners “stop buying America.” It is when maintaining an equilibrium between gigantic Treasury supply and investor demand consistently requires yields that are politically or economically painful.<\/p>\n We haven’t conclusively crossed that threshold. But compared with most of the deficit scares of the past 50 years, we are considerably closer to finding out where it is.<\/p>\n US debt burden relative to the productive capacity, tax base, wealth, population, technological leadership and financial attractiveness of the United States.<\/p>\n And then there is another comparison that deficit hawks often neglect:<\/p>\n We need context when examining US fiscal deterioration versus US relative economic performance.<\/p>\n Imagine two countries. Country A increases its government debt from $10 trillion to $20 trillion while its economy stagnates. Country B increases its debt from $10 trillion to $20 trillion while becoming dramatically richer, more productive and more important to global capital markets. Those are not remotely the same credit story.<\/p>\n Something very much like that distinction has occurred between the US and its advanced-economy peers.<\/p>\n There is an extraordinary historical irony here. For most of the 2010s and 2020s you could have read endless commentary about American fiscal decline, political dysfunction and impending loss of economic supremacy. Meanwhile the American economy was pulling further ahead of Germany, France, Britain, Italy and Japan in precisely the economic capacity from which sovereign debt ultimately gets serviced.<\/p>\n China’s enormous rise during the 2000s drove America’s share downward. It reached roughly 22 percent by 2014 and then began climbing again.<\/p>\n So the pattern is roughly:<\/p>\n 2004: 28%<\/p>\n 2014: 22%<\/p>\n 2019: 24%<\/p>\n 2025: 26%<\/p>\n That is fascinating because the “American decline” story was much more empirically plausible around 2004-2014 than it has been during 2014-2025.<\/p>\n The US was losing global GDP share rapidly as China industrialized.<\/p>\n Then something changed.<\/p>\n China slowed. Japan stagnated. Europe stagnated relative to America. The US generated extraordinary technology companies, attracted capital and skilled immigrants, produced stronger productivity growth, developed the shale energy revolution and came out of Covid much stronger economically than most other advanced countries.<\/p>\n The US therefore regained much of the relative economic ground it had lost.<\/p>\n This matters enormously for the FT column we started with. There are really two stories competing with one another.<\/p>\n One story says:<\/p>\n America is borrowing itself into dependence on its creditors.<\/p>\n The other says:<\/p>\n The world’s richest and most productive large economy keeps increasing its economic lead over almost every plausible alternative destination for enormous pools of capital, which helps explain why investors continue lending it colossal sums.<\/p>\n The second story helps explain why the first story has failed to produce the predicted crisis for 40 or 50 years.<\/p>\n And there is a deeper mechanism here. The US doesn’t merely have debt. It produces the asset that the rest of the financial system wants to hold. A Treasury bond is simultaneously an American liability and one of the world’s principal stores of liquid wealth. The larger and more productive the US economy becomes relative to Europe and Japan, the harder it is to identify a comparably deep alternative market into which $5 trillion or $10 trillion can move.<\/p>\n That doesn’t make deficits harmless. Eventually the arithmetic can overwhelm even extraordinary economic advantages. If interest expense permanently grows faster than nominal GDP and primary deficits remain enormous, debt dynamics deteriorate.<\/p>\n But it suggests that the question deficit hawks have spent decades asking may be slightly wrong.<\/p>\n Instead of asking “How can America possibly sustain $30 or $40 trillion of debt?”, ask:<\/p>\n “How much debt can an economy representing 26 percent of nominal world GDP, possessing the dominant reserve currency, the deepest capital markets and substantially better productivity growth than its principal rich-country competitors sustain?”<\/p>\n Nobody knows the answer.<\/p>\n But the experience of the last 50 years strongly suggests that the number is much larger than deficit alarmists repeatedly assumed.<\/p>\n the United States has an unusually powerful combination of long working hours, very high productivity per hour, and a disproportionate position at the technological frontier. AI could strengthen all three of those advantages.<\/p>\n On “Americans work harder,” the hours data support you if we mean the other large advanced economies. OECD estimates for 2025 are about 1,800 hours per worker in the US, versus 1,716 in Italy, 1,533 in Britain, 1,498 in France and only 1,332 in Germany. Japan was about 1,617 in 2024. South Korea is an important exception. Koreans work even longer hours.<\/p>\n On productivity, the US is not literally number one per hour. But among really large advanced economies it is now the leader. OECD puts US GDP per hour at $84.10 in constant 2020 PPP dollars in 2024. A handful of smaller countries rank above it, notably Norway and Denmark, plus some unusual cases such as Ireland and Luxembourg where multinational accounting inflates the figures. But the ranking has the US ahead of Germany, France, Britain, Italy, Canada, Japan and Korea.<\/p>\n And the trajectory may matter more than the level. The OECD’s new 2026 productivity report says explicitly that “the United States has steadily pulled further ahead of other major economies.” US output per hour rose from $50.80 in 1995 to $84.10 in 2024, about 1.8 percent annually. The EU-27 and Japan managed about 1.1 percent. The EU went from roughly 90 percent of the US productivity level in 2000 to 75 percent in 2024. Japan went from 73 percent to 62 percent.<\/p>\n That is an enormous fact. It means the US advantage isn’t simply that Americans accept fewer vacations. America is getting more output from each hour and supplying more hours.<\/p>\n AI is where the story potentially becomes much bigger.<\/p>\n The US presently has an extraordinary concentration of the inputs required for an AI productivity boom. Stanford’s 2026 AI Index puts US private AI investment at $285.9 billion in 2025, more than 23 times China’s reported private investment. It counted 1,953 newly funded American AI companies. The US also has 5,427 data centers, more than ten times the number in any other country.<\/p>\n More importantly, we’re starting to get evidence that AI is doing more than generating impressive demos. OECD’s review finds sizeable productivity effects at the task level in areas including customer service, business problem-solving and medical screening. A 2026 study covering 12,000 European firms estimated that AI adoption raised short-run firm-level labor productivity by about 4 percent. US industry data since 2022 also show faster productivity improvement in industries with greater AI adoption, although the effect is not yet economy-wide.<\/p>\n The macro estimates remain extraordinarily uncertain. Daron Acemoglu’s skeptical estimate amounts to only about 0.12 percentage points added to annual US labor-productivity growth over a decade. OECD modeling for the G7 gives a much wider and potentially transformative range of about 0.2 to 1.3 percentage points annually.<\/p>\n That upper half of the range really would be a game changer.<\/p>\n Suppose the underlying US productivity trend is around 1.5 to 2 percent and AI eventually adds something approaching another percentage point. You’re no longer talking primarily about replacing secretaries with chatbots. You’re talking about a general-purpose technology affecting software, law, accounting, finance, medicine, engineering, logistics, advertising, customer service, research, administration and eventually physical production through robotics.<\/p>\n And the US may benefit disproportionately because it already possesses the complementary assets that general-purpose technologies require: huge technology companies, deep capital markets, enormous venture-capital capacity, flexible labor markets, highly profitable firms capable of making gigantic capital investments, enormous quantities of computing infrastructure and a business culture relatively willing to reorganize firms around new technologies.<\/p>\n This last part is crucial. Owning the best AI model isn’t the same thing as realizing productivity gains from AI. Companies have to redesign workflows, fire or reassign people, change organizational structures, retrain workers and invest in complementary software and capital. The OECD explicitly says those complementary investments and managerial adjustments determine how much of the theoretical AI gain appears in actual productivity.<\/p>\n That may favor the US over Europe substantially.<\/p>\n There is a warning sign, however. Stanford finds that while America dominates AI investment and much of AI production, it doesn’t currently lead in broad adoption. Its estimate places US generative-AI adoption at only 28.3 percent, 24th internationally. So American companies still have to convert technological leadership into widespread organizational transformation.<\/p>\n This brings us directly back to the deficit.<\/p>\n The standard fiscal story focuses relentlessly on debt and deficits. But sovereign debt dynamics depend critically on economic growth. If AI causes American productivity and nominal output to grow substantially faster than expected, today’s terrifying-looking nominal debt numbers become much less terrifying relative to the future tax base.<\/p>\n And if the effect is asymmetric, it becomes even more important geopolitically.<\/p>\n Imagine that over the next 15 years AI raises American productivity substantially while Europe and Japan adopt it more slowly. The US might simultaneously:<\/p>\n grow its tax base faster,<\/p>\n increase its share of advanced-world output,<\/p>\n make American companies more valuable,<\/p>\n attract still more global capital,<\/p>\n strengthen demand for dollar assets,<\/p>\n and increase the productive capacity underlying Treasury debt.<\/p>\n In that world, America’s “squishy underbelly” could remain squishy for another generation without actually rupturing.<\/p>\n That is why I think the most interesting question isn’t “Is $40 trillion or $50 trillion of debt sustainable?”<\/p>\n It is whether America’s fiscal deterioration is happening faster or slower than the improvement in America’s productive capacity and relative economic position.<\/p>\n The answer so far is surprisingly favorable to the US. The IMF now says strong, broad-based productivity growth has “set the U.S. economy apart from its peers.”<\/p>\n AI could reverse that conclusion. It could disappoint. China could capture much more of the technology than current investment figures imply. Productivity gains could be concentrated rather than economy-wide.<\/p>\n But if AI really is a general-purpose technology comparable to electrification or computing, the country entering the transition with America’s combination of capital, computing, entrepreneurial firms, long working hours and already-high productivity is in a remarkably strong position. That possibility belongs near the center of any serious discussion of America’s long-run debt capacity, not as an afterthought.<\/p>\n","protected":false},"excerpt":{"rendered":" One thing I’ve noticed from interviewing thousands of people is that everyone has huge areas of vulnerability. A man who is strong in one situation is helpless in another situation. There might be some parallel between individuals and countries. Oh … Continue reading
\nSell some of Japan’s enormous holdings of US Treasuries and use the proceeds to buy yen. That directly pushes Treasury prices down and US yields up.<\/p>\n