So far, the Artificial Intelligence (AI) revolution has sparked debate about the possible repercussions on employment, productivity increases, or the risk of a bubble in stock markets. But its impact has just opened a new front in the debt market.
Read more Using AI to do homework and assignments reduces learning by 20%
For many years, the American tech giants Microsoft, Google, and Meta have grown thanks to software, algorithms, intellectual property, and digital networks. However, AI requires data centers, semiconductors, power grids, cooling systems, land, and infrastructure. Just like the railroad companies of old, borrowing is necessary.
Amazon, Alphabet, Meta, and Oracle have already issued nearly $194 billion in bonds in 2026, 79% more than the previous year, according to Reuters. AI-related investments, including Microsoft, could exceed $730 billion this year according to Goldman Sachs, and a third will have to be financed with debt. These companies represented barely 3% of the U.S. debt stock at the end of 2025, but already account for more than 15% of all new issuance so far in 2026. What does this mean?
That now sovereign states and multinational tech companies have to simultaneously turn to the same investors to ask for huge amounts of money. Governments, which also have to borrow to sustain their expenses (defense, energy transition, welfare state), have found a new competitor, the big big tech, which threaten to suck up the scarce available capital by offering better returns.
A pension fund, an insurance company, or a sovereign wealth fund now have more options. According to the law of supply and demand, they invest where they have better returns. This ends up making financing more expensive, not only for the same companies but also for states. The U.S. Treasury 30-year yield reached 5.337% this week, the highest since 2007. In Japan, 10-year debt is the most expensive since 1996, and in Germany, the highest in 15 years.
The key is that, at parity of maturity, hyperscaler firms currently offer better yields than an equivalent U.S. government bond. Depending on the term, the difference ranges from 40 to 118 basis points. If U.S. debt at 20 or 30 years now yields around 5.2%, a corporate bond from the big tech could reach 6.3-6.4%. What do you think the investor will choose?
Tech companies are already a competitor to sovereign debt and that pushes yields up
Carl Weinberg, founder of High Frequency Economics, considers that the large capital injections destined for AI unleash a competition to obtain resources that “come from the same pool of savings that finances the public deficit, as well as the investment of all other companies in the economy.” In his opinion, a new “superdebtor, the collective AI company,” has emerged, which is displacing investment. A February document from the Dallas Federal Reserve considers that the issuance of these tech companies this year would be equivalent to one-eighth of the ten-year debt issued by the Treasury. This avalanche could exert extra pressure on longer maturities, push yields up, and increase the final cost for the entire credit system.
The argument is not shared by all analysts. Goldman Sachs considers that, for now, the huge bond placement by the big tech companies is being absorbed by the market without causing a significant displacement of other issuers.
The problem is that AI has become a new capital demander at a time when the U.S. government also needs record amounts of financing. Public debt in the U.S. has exceeded $40 trillion, as confirmed by the Treasury Department on Thursday. Ten years ago, it was $19.4 trillion.
This is a threatening milestone for an economy that rests on precarious fiscal foundations after decades of borrowing to finance the rising costs of defense, social protection programs, and the tax cuts established by President Donald Trump, trapped by spending on the Iran war, lower tax revenue, and tariff refunds.
“To put it in perspective, it took the United States almost 200 years to reach a gross debt of one trillion dollars for the first time in 1981,” explained Maya MacGuineas, president of that committee, in a statement. “A debt of 40 trillion does not exist only in the accounting books. It is perceived throughout the economy and affects citizens’ pockets. The more we borrow, the more we fuel inflation, displace other budget priorities, and expose ourselves to vulnerability in the face of internal emergencies and foreign turmoil,” she added.
The fiscal situation in the U.S. has led the Treasury to announce yesterday that it will increase the volume of its debt buybacks in the long end of the yield curve. A patch to control the cost of financing, when debt interest has reached almost $1.2 trillion this year and constitutes the second budget expense above Defense and Medicare.
Trump’s promises have not been fulfilled. And the AI boom makes it even harder to balance the numbers.
Read more A magnitude 7.2 earthquake shakes the Andean region of Ayacucho, in southern Peru
