No borrower will escape the surge in market-based interest rates unleashed by the Middle East war and resulting energy supply shock. But for U.S. tech firms planning to spend over US600-billion this year in the artificial intelligence arms race, spiking borrowing costs could not have come at a worse time.
The AI capex surge is unprecedented. The expected US630-billion in capital expenditure from Big Tech this year, mostly on AI data centers, chips, and cloud computing, is worth more than 2 per cent of GDP. The projected spend of more than US800-billion next year is closer to 3 per cent of GDP.
Big Tech historically used cash to fund expansion. And they still have lots of it: some estimates put the big five hyperscalers’ combined cash and equivalent holdings at over US350-billion. That’s partly why Apple and Microsoft still enjoy credit ratings higher than those of the U.S. government.
But they’re burning through it.
At the end of last year, about 60 per cent of hyperscalers’ operating cash flow was used to fund capex, according to Apollo Global Management. That’s now almost 70 per cent. At this pace, almost every dollar earned may soon be earmarked for capex.
Analysts at Morgan Stanley say Big Tech’s cumulative capex this year and next will be about US1.4-trillion, consuming nearly 90 per cent of expected operating cash flow of US1.6-trillion.
Tech giants must therefore increasingly turn to the credit markets. Bank of America analysts estimate that hyperscalers’ debt issuance this year will total US175-billion, up from US121-billion last year and more than six times the annual average of US28-billion in the preceding five years.
Widening the lens, the scale of borrowing across the sector is even greater. MUFG analysts reckon that investment-grade issuance by tech and AI companies last year totaled US245-billion. That’s not too far off the cumulative US298-billion amassed over the previous decade.
Last week I outlined the bullish U.S. tech narrative, which is rooted in the idea that hyperscalers are relatively well-positioned to weather this exogenous shock. Remarkably, since the Iran war broke out four weeks ago, tech earnings growth estimates have risen faster than any other sector, even energy, according to Capital Economics.
But investors are clearly skeptical that the AI investment and borrowing being undertaken will generate adequate returns. The Roundhill “Magnificent Seven” exchange-traded fund fell 5 per cent last week, putting its monthly loss around 10 per cent and its fall from the October high near 20 per cent.
There is cause for concern. The use of leverage to finance AI projects will put additional pressure on hyperscalers’ balance sheets, even as the scale of investment will make every incremental dollar of profit harder to achieve. That pessimism will only grow if market-based interest rates continue rising.
The 10-year U.S. Treasury yield is up 45 basis points this month, its biggest rise since October 2024. If it adds another few basis points by March 31, this will mark the steepest monthly rise since October 2022.
This hasn’t caused major ructions in the investment-grade corporate bond market, where spread widening over the same period has been a fairly tame 15 basis points.
But that might change. Big Tech faces a potential double whammy - higher rates and growing debt obligations on one side, the prospect of squeezed profits and sliding share prices on the other.
For the wider market and economy, the consequences of this are potentially significant, given how central these companies are to overall U.S. earnings and growth.
If the eye-watering capex binge - one of the largest collective investments ever in a single sector - comes to fruition, it’s difficult to see how the economy could tip into recession. But if rising yields and falling share prices stymie these plans, the perfect storm of increasing inflation, tighter borrowing costs and an already weak hiring market may be too much for the U.S. economy to bear.