AI capex cycle likely to end with market pushback, not spending cuts: Jefferies
New Delhi, July 5
The AI investment cycle is most likely to end not because US hyperscalers cut spending, but because markets start pushing back against the lack of returns, Global Brokerage firm Jefferies said in a report.
The report argued the turning point will come when investors focus on what it calls a massive wealth transfer from hyperscalers' balance sheets to North Asia. The combined market capitalisation of Korea and Taiwan has more than tripled from US$3.2 trillion at the start of 2023 to US$9.8 trillion, reflecting how much of the AI capex is flowing to chipmakers and suppliers in the region.
That dynamic is already showing up in relative performance. The four major US hyperscalers - Microsoft, Alphabet, Amazon and Meta - are up 180 per cent since the start of 2023 and outperformed the S&P 500 by 44 per cent. But they have declined 8.7 per cent since late May and underperformed the index by 10.2 per cent from the early-May relative high. "GREED & fear" says this is the key chart to watch going forward.
The brokerage added that the four hyperscalers have issued US$144 billion in bonds so far this year, vs US$83 billion in all of 2025, suggesting more of the capex is being debt-funded. It warns of "massive capital destruction, or malinvestment" for the hyperscalers if returns fail to materialise.
Outside AI, geopolitics remains a risk markets are ignoring for now. The note flags questions around the Iran MoU and the deepening involvement of NATO in the Ukraine conflict, and continues to recommend owning some energy stocks as a hedge.
In Australia, the brokerage sees rising odds of a downturn. The RBA has hiked rates to 4.35 per cent to fight inflation at 4.0 per cent YoY in May, and Sydney home prices have fallen 3.7 per cent over the past five months. A new federal budget adds further pressure from July 2027: capital gains will be taxed at marginal rates of up to 47 per cent, and the negative-gearing deduction will be removed for new investment properties except newly built homes. The brokerage expects this to cause a sharp drop in property transactions and investor lending, which was still 43.4 per cent of new NSW mortgages in 1Q26.
Macro headwinds are compounded by weak productivity - real GDP per hour worked is down 5.1 per cent since 1Q22 - and stretched affordability, with the national dwelling value at 8.4x median income and 45.9 per cent of income needed to service a mortgage.
On markets, the note favours an overweight in resources over banks. Resources have outperformed banks by 54 per cent since June 2025. It also notes the concentration risk in Asia: tech hardware now accounts for 49 per cent of MSCI AC Asia ex-Japan, making it harder for diversified portfolios to beat the benchmark.
— ANI
Reader Comments
As an investor in Indian mutual funds with significant US tech exposure, this report is sobering. The $144bn in bonds issued by just 4 companies in 2025 vs $83bn in all of last year - that's 73% more debt in half the time! If returns don't materialise, we're looking at a potential correction that will ripple through global markets. Time to diversify into energy stocks as they suggest? 🤔
The wealth transfer to North Asia is staggering - market cap tripling from $3.2T to $9.8T in just 2 years. For Indian chip design firms and semiconductor supply chain companies, this is actually a massive opportunity if they can position themselves as alternatives to Taiwan. We need to push harder on our own semiconductor mission. 📈
The Australia data is concerning - rising rates, falling property prices, and new tax changes from 2027. It's like they're trying to engineer a soft landing but the numbers say otherwise. For NRIs with Australian investments, the message is clear: diversify out of property and into resources. RBA rate hikes won't help when productivity is down 5.1% since 2022.
One must question: is AI really creating value or just moving wealth from US balance sheets to Asian chipmakers? The Jefferies report calls it "malinvestment" - strong words from a global brokerage. For India, the lesson is to focus on AI applications that solve real problems (agriculture, healthcare, education) rather than just chasing the hype cycle. We don't have $144bn to burn. 💡
J James Anderson