NextFin News - Nouriel Roubini, the economist who built a reputation on calling crises, is not worried about the surge in US bond yields. His reason cuts against the grain of the rout: he sees the rise not as a warning signal but as the price of an artificial-intelligence investment boom that should lift growth and stocks alike. That call puts him at odds with a market that has pushed the 10-year Treasury yield above 4.8%, its highest level since January 2025, and with Federal Reserve Chair Kevin Warsh, who told the Jackson Hole symposium last Friday that inflation is running above the Fed's 2% target and that the central bank still has work to do.
The tension is stark. Bond yields have climbed to levels unseen since before the global financial crisis, oil has surged past $90 a barrel on renewed Middle East fighting, and traders have swung from pricing a rate cut to pricing a better-than-even chance of a rate hike at the Fed's mid-September meeting. Yet Roubini's argument is that the US is different from the other advanced economies caught in the rout, because the driver of its yield rise is private-sector credit demand for AI capital expenditure, not fiscal panic or inflation expectations. The rest of the developed world, in his telling, is not so lucky.
What the Bond Market Is Actually Saying
The starting point is that higher yields are not one thing. Roubini separates the recent rise into three possible causes, only one of which is benign. First, supply shocks — protectionism and war-driven oil disruptions — can push nominal yields higher while cutting growth and equities, a stagflationary mix. Second, unsustainable fiscal deficits can lift yields either because markets expect central banks to monetize the debt, stoking inflation, or because investors demand a sovereign-risk premium, which can crowd out private spending and trigger a debt crisis. In both of those cases, higher yields are bad news for growth and stocks.
Then there is the third channel: a private-sector investment boom. When firms borrow heavily to finance data centers, chips, and power infrastructure, credit demand pushes nominal yields up without any rise in inflation expectations. In that case, higher yields are a signal of stronger future growth, which is good for equity prices. Roubini argues this third factor has been a major driver of US yields and stocks for the past few years. Since the launch of ChatGPT in 2022, he notes, US yields have climbed from a pandemic-era bottom of 1% to current levels around or above 5%, while US equities have chalked up double-digit returns. Even into mid-August, the rise in yields coincided with new all-time highs for the S&P 500, which closed at a record 7,798.99 on August 14 — its 27th record close of the year.
The mechanism is straightforward, and it is why the usual bond-market alarm bells may be ringing for the wrong reason. A rise in real yields is what one would expect when tech firms and hyperscalers borrow on a massive scale to finance data-centre construction. The scale of that borrowing is no longer theoretical: Goldman Sachs estimates AI capital expenditure of roughly $765 billion in 2026, rising to $1.6 trillion annually by 2031, with about $7.6 trillion of cumulative spending across compute, data centers, and power over the six-year span. US government debt has surpassed $40 trillion — $40.05 trillion as of August 19, according to the Treasury Department, more than double the $19.4 trillion level a decade ago — yet Roubini contends the US fiscal deficit has plateaued near 6% of GDP and is likely to fall in the next few years as higher potential growth and additional tariff revenues kick in.
One therefore should assume that in high-innovation economies like the US, the recent rise in bond yields is mostly driven by the rise in real yields rather than by an increase in inflation expectations, which remain well anchored slightly above 2%.
That attribution matters because it determines whether the yield rise is a tax on growth or a symptom of it. If inflation expectations are anchored and the rise is in real yields, then the discount-rate hit to equity valuations is being offset by higher expected earnings growth — the 1990s productivity-boom pattern rather than the 2022 inflation-shock pattern.
Why the US Is Different From the Rest of the Developed World
Roubini's argument is explicitly not universal. The same bond rout is hitting the UK, Japan, Germany, and France, where yields have reached multi-decade highs, but the diagnosis differs. In those economies, wide fiscal deficits and enlarged debt burdens are doing more of the work. German and French long-bond yields hit 15-year highs in early September, French OAT futures traded at their lowest since the contract launched in 2012, and Japan's 30-year yield has neared a record above 4%, having touched an all-time high of 4.21% earlier this month. Britain's 30-year gilt yield climbed to around 5.85% on September 1, its highest level since 1998.
The distinction is between a yield rise funded by productive private investment and one funded by public dissaving. In the US, the AI capex boom creates future productive capacity — data centers, power grids, and computing infrastructure that can raise potential output. In the European and Japanese cases, higher yields reflect debt-sustainability concerns with no offsetting supply-side payoff. That is why, in Roubini's framing, the rout will not hit US stocks and growth the way it hits those of other advanced economies. The exorbitant privilege of the dollar, in his view, is not over either, which keeps US funding conditions comparatively favorable.
The productivity premise is the linchpin. Hudson Bay Capital research authored by Roubini sketches a plausible path in which AI adds 0.5 to 1.5 percentage points to annual productivity growth, pushing US potential growth toward 4%. That is a large claim: it would roughly double the trend growth rate the US economy has managed since the global financial crisis. It is also the exact claim that the bond market's skeptics attack.
The Counter-Case: What If the Productivity Never Shows Up?
The strongest argument against Roubini is simple: the productivity boom is priced in, but the productivity itself has not arrived. The San Francisco Federal Reserve's measure of utilization-adjusted total factor productivity grew just 0.07% over the four quarters ending in the first quarter of 2026 — close to zero and well below the post-Golden Age era. On that measure, the AI revolution has so far produced a capex boom without a corresponding efficiency gain, which is the worst of both worlds: higher credit demand today with no offsetting supply expansion tomorrow.
That gap between spending and measured output is what makes the bond rout dangerous. If AI capex turns out to be mostly speculative — if the data centers are built but revenue per watt falls short, if the models plateau, if the power constraints bind harder than expected — then the real-yield rise was not justified by future growth. In that scenario, yields stay high because of fiscal risk, not investment demand, and the equity market, particularly the long-duration tech names that are most sensitive to the discount rate, has priced a payoff that never materializes. The trade has been framed precisely by skeptics of the AI boom: indebted governments are betting that a productivity surge from AI will pay the bills, but it could be a while before the boost comes.
There is also the monetary-policy channel that Roubini's view must survive. Warsh's Jackson Hole speech moved markets sharply: the probability of a 25 basis-point rate hike at the September meeting rose to 55.7%, about 20 percentage points higher than a day earlier, according to the CME Group's FedWatch tool. The policy-sensitive 2-year Treasury yield jumped nearly 8 basis points to 4.31%, its highest since late July. A Fed that is still fighting inflation above target is not a Fed that will look kindly on an overheated capex cycle, and a rate hike would hit the same long-duration assets that are carrying the equity indices.
Warsh was explicit about the inflation backdrop in his prepared remarks. "The Fed's price-stability objective of 2 percent, as measured by the personal consumption expenditures (PCE) price index, is a firm, fixed target," he said, before adding that the 12-month change in the PCE price index stands at 3.7% and the six-month change at 4.1%. "None of these measures are perfect, but they all tell a similar story: Inflation is running above our 2 percent target. So the Fed's predominant focus right now should be on prices." His standard for easing is demanding: "We must be confident that underlying inflation is moving to our objective, clearly, and at sufficient speed. Otherwise, we have work to do."
Roubini anticipates part of this objection. He notes that the positive correlation between yields and stocks can turn negative when inflationary shocks force tighter policy, as it did in 2022, when the S&P 500 fell about 15%, the Nasdaq fell more than 20%, and many individual tech and growth stocks dropped more than 30%. His answer is that the current environment is not 2022: inflation expectations are anchored slightly above 2%, and the yield rise is real-yield driven. But that answer depends entirely on the productivity numbers arriving. If they do not, the 2022 dynamic — yields up, stocks down — reasserts itself.
What This Means for Markets: Beneficiaries, the Exposed, and the Time Horizon
If Roubini is right, the beneficiaries are clear. Hyperscalers and the AI supply chain — chips, data-center construction, power equipment, and the utilities that can deliver reliable electricity — benefit from a multi-year capex cycle funded at real yields that are high but sustainable. The US dollar and US equities outperform their developed-market peers, because the growth differential widens. Long-duration tech stocks, despite their sensitivity to yields, continue to outperform because their earnings growth outruns the discount-rate headwind.
The exposed are the mirror image. Economies running wide deficits without a productivity engine — the UK, France, Germany, and Japan in this framing — face rising debt-service costs with no growth offset. Within the US, firms with weak balance sheets and no AI exposure face higher borrowing costs without the revenue growth to cover them. And if the productivity payoff disappoints, the long-duration tech complex that has led the index higher becomes the most vulnerable asset class, since its valuations embed the very growth assumptions that would be revised down.
The time-horizon split is the cleanest way to frame the trade. In the short term, sentiment and liquidity dominate: oil spikes, Fed rhetoric, and Treasury market functioning will drive volatility, and the recent intervention in the bond market suggests policymakers themselves are uneasy about the speed of the yield move. In the medium term, fundamentals decide: the next four to eight quarters of productivity and earnings data will show whether AI capex is converting into output. In the long term, the structural question is whether AI is a general-purpose technology on the scale of electricity or the internal combustion engine — in which case Roubini's 4% potential-growth call is conservative — or a narrower automation wave that boosts measured productivity only modestly.
Three signals would falsify the rosy view. First, if core inflation prints at or above 0.3% month over month for two consecutive months while the unemployment rate holds near 4.1%, the Fed's inflation fight is not over and the real-yield story gives way to a policy-tightening story. Second, if utilization-adjusted total factor productivity remains below 0.5% annualized over the next four quarters, the capex boom is not translating into efficiency, and the 2022 playbook returns. Third, if the 10-year Treasury yield breaks decisively above 5.0% — roughly a quarter-point above the level at which it has been trading in recent sessions — while the S&P 500 fails to reclaim its August record near 7,800, the market itself is telling investors that the yield rise has turned from growth-positive to growth-destructive.
The base case is that US yields grind higher but stocks hold up, because the AI capex cycle has years to run and inflation expectations stay anchored. The upside case is that productivity data begin to confirm the 0.5-to-1.5 percentage-point boost, potential growth rises toward 4%, and the 1990s-style combination of falling inflation, rising real yields, and surging equities repeats. The downside case is that capex spending peaks before revenue arrives, fiscal deficits reassert themselves as the dominant yield driver, and the correlation between yields and stocks flips negative just as it did in 2022.
Roubini's bet is not that higher yields are harmless. It is that they are a symptom of a real investment boom, and that a boom funded by productivity is very different from one funded by debt. The bond market will settle the argument — but only after the productivity data arrive, and only if the Fed believes inflation is beaten before it reaches for the hike button.
Explore more exclusive insights at nextfin.ai.
