Journey of the US 10Y Treasury Yield (2016-2026)

Annual average yield for 2016–2026; 2026 shows the latest available observation. Source: Board of Governors of the Federal Reserve System (DGS10), via FRED.

The current AI-driven equity boom arguably began in May 2023, when Nvidia’s earnings release transformed generative AI from a compelling technological narrative into an investable earnings story. The company guided second-quarter revenue more than 50% above Wall Street expectations, sending its shares 24% higher in the following trading session.1 What followed was a sustained surge across public equity markets, increasingly interpreted as evidence that generative AI could deliver productivity gains and economic value significant enough to outweigh a difficult global backdrop in the markets.

That optimism has persisted despite one of the most geopolitically uncertain environments in decades. Trade tensions, geopolitical fragmentation and conflict have done remarkably little to derail equity markets, reinforcing the belief that the economic potential of generative AI can offset many of these headwinds. Yet this narrative risks overlooking a more uncomfortable longer-term macroeconomic problem.

Government borrowing and public debt remain historically elevated across many of the world’s largest developed economies, while inflation has proved stubborn and interest rates remain restrictive. If inflation refuses to fall decisively, how much further can the US economy absorb higher rates? More importantly, if AI-fuelled equity gains continue to loosen financial conditions, increase household wealth and support demand, could the stock-market boom itself make the task facing central banks more difficult? The question is no longer simply how far AI can take equity markets, but whether markets can continue rising without eventually forcing monetary policy to push back.

How Can Warsh Battle Soaring Borrowing Costs?

Federal Reserve Chairman Kevin Warsh faces the problem that there is no painless way to bring borrowing costs down. In the short term, the Fed could lean on its balance sheet: buying longer-dated Treasuries, or shifting reinvestments towards the long end, would remove duration from the market and compress the term premium. Previous Fed asset-purchase programmes materially lowered long-term yields.2 But doing that while inflation remains above target risks sending the opposite message. The message would be that the Fed is easing before price stability has been restored.

The longer-term fix is less convenient. If investors are demanding a higher premium to fund persistent deficits, the Fed cannot solve that with monetary policy alone. Warsh can anchor inflation expectations, but Washington has to convince the bond market that the trajectory of debt and issuance is sustainable. Otherwise, lower policy rates could simply be offset by a higher fiscal and inflation risk premium at the long end.

And this is not uniquely American. Borrowing costs are rising across the largest developed economies. Japan’s 10-year government bond yield has reached 3% for the first time since 1996,3 while UK 10-year gilts recently neared 5.3%, their highest since June 2008.4 The global bond market is sending a consistent message that capital is no longer cheap. The question for Warsh is whether the Fed can lower the price of money without convincing investors that inflation, or fiscal discipline, is being sacrificed.

AI Hyperscalers: The Culprits for Warsh’s Bond Yield Headache

The uncomfortable part of the AI boom is that every blockbuster earnings season brings another spending plan. Alphabet, Amazon, Microsoft and Meta are expected to spend around $700 billion this year as each races to build the chips, data centres and power infrastructure needed to stay ahead. The equity market rewards that ambition, but the financing has to come from somewhere.

Increasingly, it is coming from debt. AI-related global debt issuance is forecast to approach $570 billion in 2026, while Amazon, Alphabet, Meta and Oracle had already issued roughly $194 billion of bonds by early July.5 That wave of supply is forcing investors to absorb corporate debt while governments are issuing heavily too. The result is a battle for capital that pushes yields higher and makes cheap money harder to recover.

This is the culprit behind why Warsh’s headache is not uniquely American. Bond yields are at multi-year or multi-decade highs across the US, UK, Germany and Japan, while the ECB and Bank of Japan are being pushed towards tighter policy as inflation risks persist. AI may eventually deliver the productivity gains markets are pricing in. But before it lowers costs, the race to build it is raising the price of capital.

Footnotes

  1. Reuters, Chip giant Nvidia nears trillion-dollar status on AI bet (opens in a new tab), 26th May 2023.

  2. Federal Reserve Board, The Effect of the Federal Reserve’s Securities Holdings on Longer-term Interest Rates (opens in a new tab), 20th April 2017.

  3. Reuters, Japan's benchmark bond yield rises to 3% for first time in 30 years (opens in a new tab), 1st September 2026.

  4. Reuters, UK bond yields hit fresh 19-year high, adding to pressure on Healey (opens in a new tab), 2nd September 2026.

  5. Reuters, Hyperscaler debt binge pushes yields up as investor demand cools (opens in a new tab), 29th July 2026.