AI ammunition is far from exhausted! Goldman Sachs predicts that tech giants' capital expenditures will reach $1.4 trillion by 2027.

“Expectations for hyperscaler capital expenditures in 2027 are overly conservative.”

On June 12, Goldman Sachs economist Ryan Hammond published a report titled "More AI Capex, More Volatility," significantly raising forecasts for hyperscaler capital expenditures next year: the base scenario is $1.1 trillion, and the extremely optimistic scenario reaches $1.4 trillion.

The report comes amid a wave of Q1 earnings reports and capex guidance from hyperscalers, i.e., tech giants such as Microsoft, Google, Amazon, and Meta. According to Morgan Stanley, forecasts for these companies’ combined 2027 capex have already jumped from $950 billion in Q4 2025 to over $1.1 trillion—a single-quarter adjustment of over 30%.

Goldman's Three-Tier Forecast: From $920 Billion to $1.4 Trillion

Mainstream analysts currently forecast hyperscaler capital expenditures in 2027 at about $920 billion, with growth rate slowing sharply—from 84% in 2026 to 22%.

Goldman disagrees with this figure.

Hammond’s modeling framework: If incremental AI infrastructure investment reaches 2% to 3% of GDP—analogous to historic construction cycles in railways and automobiles—then capital expenditures in 2027 will reach about $1.1 trillion, corresponding to a growth rate of 45%.

In an even more extreme optimistic scenario, factoring in hyperscalers’ own cash flow generation and the financing capacity of the investment-grade credit bond market, capex could hit $1.4 trillion—a growth rate of 89%.

Valuation at Highest Since Launch of ChatGPT

The upward revision of capex expectations has directly driven valuation expansion in the AI infrastructure sector.

Their data show that the median P/E ratio for AI infrastructure stocks has reached 26x, the highest since ChatGPT launched. Semiconductor and power (non-utility) sectors’ median P/E ratios have continued to rise this year, whereas valuation expansion for hyperscalers themselves and memory chip stocks has been relatively limited.

Hammond’s assessment: “Capex above expectations means that profits and stock prices for AI infrastructure beneficiaries still have room to rise in the near-term.”

But Hammond also gave a warning: Recent valuation expansion and portfolio changes mean volatility ahead will intensify. Investors need to find balance between “capex above expectations,” “potential capex growth slowdown,” and “profit sustainability in doubt.”

AI Adoption Data: High Praise, Low Usage

In the Q1 earnings season, about 54% of companies mentioned AI and productivity on earnings calls. But only 11% actually quantified productivity improvements in specific use cases; only 2% quantified AI productivity gains at the profit level—the previous quarter these figures were 10% and 1%, almost no substantial progress.

More direct data from user surveys: Only 12.6% of people use AI daily, up just 2 percentage points from a year ago. Self-reported cumulative productivity gains for all workers were 1.6% a year ago, now 2.2%—an increase of only about 0.5 to 0.6 percentage points in one year.

In other words, trillions in capex are pouring in, but actual user engagement and productivity returns remain quite limited.

"Terminal Value" Debate Among Software Stocks

Hammond also discussed the valuation logic of the software sector.

Last year, the median P/E ratio for software stocks peaked at 39x, dropped to 21x in March this year, and has now rebounded to 25x, though there is wide differentiation among subsectors.

Hammond calculated using a discounted cash flow model: At the start of the year, about 85% of the software sector’s present value came from "terminal value," that is, discounted expectations for profits in the distant future. This means that even “modest” changes to market assumptions about software companies’ long-term growth rates and profit margins will cause significant valuation swings—which explains this year’s volatility in software stock valuations.

The core debate: Is AI a tool to empower software companies, or a disruptor? Will the emergence of low-cost competitors continue to suppress revenue and profit growth for existing software companies? This debate will continue to drive divergence within the sector.

Token Price Wars: Another Variable Not to Be Ignored

At the time of this report’s release, another thread is unfolding in the market: the “race to cut prices” between OpenAI and Anthropic in token pricing.

The price war’s backdrop: Some companies have already faced the problem of "tokenmaxxing," i.e., excessive token use—for example, Uber depleted its entire annual AI budget in one quarter. As token prices continue to drop, revenue pressure on large model vendors will grow further, and their ability to support massive capex commitments will be tested.

This means there may be a substantial gap between forecasted capex figures and actual results.

Risk Disclaimer and Exemption ClauseThe market carries risks, and investment should be approached cautiously. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situation or needs of any individual user. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Investment based on this article is at your own risk.