AI infrastructure spending is projected to reach about $1.1 trillion by 2027. Morgan Stanley estimates US hyperscalers will spend more than $800 billion on capex in 2026 alone.
The $800 billion figure is described as roughly matching what the entire non-tech group in the S&P 500 spent in the prior year. It is forecast to be nearly double 2025 levels and triple 2024 levels.
Capex Volume Not Just Price
Rising costs play a part, but the main driver is higher volumes of chips, power, memory, and compute. Over four years, chips, memory, and clustered compute are said to have become four to seven times more capable.
Google reported processing 16 billion tokens per minute, up 60% quarter on quarter. Higher-capacity model training is expected to test how well recent capex converts into deployable systems and revenue.
The buildout has supported semiconductor shares, while also increasing demand for debt funding. Investment-grade issuance is running well ahead of last year, with more borrowing pushed further out in maturity.
Higher capex can support revenues, while credit markets must absorb steady new supply that can pressure spreads. If spending slows, the impact could spread across risk assets, including credit.
Positioning For The Next Phase
The AI infrastructure race is not just continuing; it is accelerating into mid-2026. Looking back at the Q1 earnings reports from April, we saw hyperscaler capital expenditure guidance once again surprise to the upside, reinforcing this dominant theme. For the coming weeks, our strategies must be aligned with this relentless spending cycle and its consequences.
The most direct approach is to maintain bullish exposure to the direct beneficiaries, primarily the semiconductor sector. Data released by the Semiconductor Industry Association on May 1st showed that global chip sales for the first quarter of 2026 grew over 25% year-over-year, confirming that the demand is real. We should consider using call spreads on semiconductor ETFs to capture further upside while defining our risk, as outright call options remain expensive due to high implied volatility.
However, we must respect the awkward asymmetry this spending creates. This investment is being fueled by massive debt issuance, which is already pressuring credit markets, with data showing that tech’s investment-grade bond supply is tracking 40% ahead of the pace we saw in early 2025. Buying puts on investment-grade bond ETFs like LQD could serve as an effective hedge if credit spreads continue to widen under this supply pressure.
This market is priced for perfection, and any sign of a slowdown in spending could cause a sharp reversal across asset classes. The VIX has been hovering near its yearly lows around 14 for the past month, making downside protection relatively inexpensive. Buying out-of-the-money puts on the Nasdaq 100 is a prudent way to hedge the risk that this capital-intensive narrative eventually falters.
We also have to consider the second-order effects of this buildout, particularly the strain on the power grid. Recent reports have highlighted that energy consumption from new data centers is exceeding forecasts, creating a clear bottleneck. A less crowded trade would be to look at long call options on the utilities sector or specific companies involved in power generation and grid infrastructure.
