- According to Bloomberg, Big Tech’s debt for AI infrastructure has doubled to $350 billion.
- Bank of America estimates that AI-related bonds will reach $270 billion in 2026, up from $136 billion in 2025. Hyperscalers make up $194 billion of this amount.
- Hyperscalers went from issuing no non-dollar bonds in 2024 to making it a significant part of their funding by 2026, which has led US tech companies to seek financing in European credit markets.
Bloomberg reports that Big Tech’s total AI-related debt has doubled to $350 billion — a trend that Tech Funding News has closely tracked as hyperscalers burn through cash faster than they can replace it.
Amazon, Alphabet, Microsoft, and Meta plan to spend a combined $725 billion on capital expenditure in 2026, up 77% from last year’s record $410 billion, according to Yahoo Finance. Even with more than $1 trillion in combined cash, the four can’t cover that spending from cash flow alone.
Alphabet alone generated $174 billion in operating cash flow over the 12 months to March 2026, and still went out and raised $80 billion in equity, the largest single equity raise in US corporate history, on top of more than $85 billion in debt across six currencies in the past year.
Hyperscalers issued about $121 billion in US corporate bonds in 2025, more than four times their 2020–2024 annual average of $28 billion, according to Investing.com. Issuance in 2026 is tracking even higher. As a result, companies are looking further afield for funding. Alphabet set borrowing records in yen, Canadian dollars, Swiss francs, and sterling within a single year, and issued a 100-year bond in February as part of a $31.5 billion global offering, following an earlier $20 billion bond sale that drew more than $100 billion in investor orders. Morgan Stanley projects that hyperscaler euro borrowing could reach €50 billion in 2026, making US tech the largest single source of corporate debt in the eurozone, ahead of France.
This matters beyond Wall Street
It isn’t only a US credit-market story. When hyperscalers compete for the same euro-denominated debt that European scale-ups and infrastructure funds rely on, financing conditions shift here too, even for companies with no direct AI exposure.
It also sharpens a question for the AI-infrastructure and compute-adjacent startups TFN covers regularly, from Munich’s QuantumDiamonds, which just raised €91 million to build out chip inspection tech that the AI supply chain depends on, to Sheffield’s Pixel-Flo, which raised £5.25 million to fix a bottleneck in microLED manufacturing.
Does hyperscaler debt-funded expansion mean cheaper, more available compute downstream for companies like these, or will providers prioritise their biggest enterprise contracts first?
The market’s patience is wearing thin
Amazon’s own bond sale illustrates the shift: a $25 billion offering on 7 July was, per Bank of America, the softest hyperscaler bond launch since Meta’s $30 billion sale in October 2025, coming after Meta was already reported to be on a quest to raise $29 billion through private credit for similar data centre expansion.
Traders have also been selling existing tech bonds — Amazon, SpaceX, Alphabet, Nvidia, Meta, and Oracle among them — to make room for new supply, widening secondary spreads by 7 to 13 basis points in some cases. As Bloomberg notes, investors are selling one bond to buy another because they’re running out of room, not because they’ve lost confidence in the sector.
Share buybacks have quietly become the casualty. Of the four biggest AI spenders, only Microsoft bought back stock in the first quarter, and at $3.4 billion, that was its smallest buyback in nearly a decade. For now, the capital-light tech giant is gone.
Whether $350 billion in AI debt is smart leverage for a generational infrastructure buildout, or the first sign of strain in a story investors have funded on faith, will be answered by the next few bond auctions.
