Market skepticism over AI capital expenditure has persisted for over a year, but fresh data is beginning to shift that narrative.
For the past twelve months, investors have worried that tech giants pouring hundreds of billions into AI infrastructure might face a "high spending, low return" problem. However, with cloud demand accelerating, mounting evidence suggests AI capital outlays are converting into stronger order backlogs and future revenue growth, rather than the feared overinvestment.
Data shows that cloud backlog orders at hyperscale providers have surged more than 150% year-over-year, reaching approximately $1.7 trillion in total. This growth rate far outpaces the roughly 80% increase in capital expenditure over the same period. JPMorgan notes that this significant gap indicates the potential revenue returns from AI infrastructure investments are exceeding market expectations, suggesting that the valuation pressure on tech giants may be nearing its end.
Against this backdrop, JPMorgan points out that the market is reassessing the return potential of AI infrastructure spending, and the valuation multiples for the Mag 7 may have already bottomed out.
AI Spending Shifts from Cost to Revenue Source, Cloud Demand Beats Expectations
Signals from the second-quarter earnings season indicate that cloud computing demand is materializing rapidly.
JPMorgan analyst Mark Schilsky states that the growth rates of cloud backlog orders and net new annual recurring revenue (ARR) are both significantly outpacing capital expenditure growth. This implies that future revenue growth is likely to cover the current large-scale infrastructure investments.
During Amazon's second-quarter earnings call, CEO Andy Jassy unusually raised the long-term outlook for AWS. He noted that the company previously expected AWS to grow into a business generating hundreds of billions in revenue, but now believes that figure could at least double, potentially becoming a $1 trillion annual revenue business in the future.
Jassy also stated that the company has already seen demand levels for 2028 that are "staggering," while the large-scale adoption of AI inference services by enterprise clients is still in its early stages.
Management teams at Microsoft, Alphabet, and Meta have also issued similar signals: the commercialization of AI applications is still in the early expansion phase, and enterprise demand is far from reaching maturity.
Earnings Forecasts Rise, But Valuations Fall to Historic Lows
Despite improving AI fundamentals, tech stock valuations have experienced significant compression.
Following the market correction in July, the forward price-to-earnings (P/E) ratio for the S&P 500 Information Technology sector has fallen to around 20x, near its lowest level in the past year. This places it at the 1st percentile of its historical valuation range, below the average of roughly 23x over the last decade. This indicates a rare divergence in the tech sector: earnings expectations are steadily improving, while valuation multiples continue to decline.
JPMorgan notes that the forward P/E ratio for large-cap tech stocks (excluding semiconductors) is currently more than two standard deviations below their historical average since 2018. If valuations recover to one standard deviation below the historical mean, it would imply an upside of roughly 30%. A return to the long-term average suggests a potential upside of around 56%.
Meanwhile, the performance of hyperscale cloud providers relative to the S&P 500 has fallen to near three-year lows. Historically, such positions have often been followed by strong mean-reversion opportunities.
Institutional Allocation Still Lags, Tech Stocks May See Catch-Up Buying
Another side of the valuation slump is that institutional capital allocation has not yet fully caught up with the changing fundamentals.
Data from Deutsche Bank shows that despite significant improvements in earnings growth and earnings forecasts for large-cap tech companies, institutional investor positioning in the sector remains only slightly overweight, notably lower than the allocation levels seen during past strong earnings cycles.
Additionally, capital flows have been heavily concentrated in the semiconductor sector this year, while positioning in large-cap tech stocks (excluding semiconductors) remains relatively underweight.
JPMorgan believes that if the core market narrative around AI shifts from "is capital expenditure excessive?" to "investment returns are materializing," the next phase of tech stock gains may be driven more by internal sector rotation rather than solely relying on continued gains in chip stocks.
From a technical perspective, the MAGS index has rebounded nearly 10% from its recent low, reclaiming its 200-day moving average and approaching the long-term uptrend line established since April last year. JPMorgan notes that the 200-day moving average is currently flattening, suggesting the market is undergoing an extended consolidation phase. Historical experience indicates that the longer the sideways movement, the stronger the breakout tends to be once a direction is chosen.
Comments