AI: Valuation signals

By Joseph Wu, CFA

Valuation reset

The narratives around AI continue to oscillate between enthusiasm over its transformative potential to concerns about the scale of capital expenditures (capex), monetisation and disruption risks.

With capex by eight of the largest U.S. tech companies expected to exceed US$1 trillion in 2027, according to Bloomberg, we view skepticism around the AI investment theme as healthy and warranted. But it is equally important to recognise that markets have not remained static. Valuations have already adjusted meaningfully to reflect various uncertainties and risks.

A broad basket of tech and AI-related companies now trade at roughly 20x forward earnings, down almost 30 percent from 28.5x in Oct. 2025 and well below its 24.5x average since 2015. The relative valuation correction has been just as pronounced. From 2015 to 2025, this AI cohort commanded an average forward price-to-earnings premium of 30 percent to the S&P 500. That premium has fallen sharply, bringing the group close to parity with the broader market.

Valuation premium has almost vanished

Broad AI basket versus S&P 500

• Relative forward P/E
• Average (2015—2022)
• Average (since 2015)
• Average (2023—present)

Note: Broad AI basket includes the S&P 500 Info Tech sector (70 percent) and equally weighted Amazon, Alphabet, and Meta Platforms (30 percent). The different averages in the chart contrast two distinct periods around the AI cycle. The 2015–2022 average premium (1.33x, 33 percent) captures the period before the start of the AI investment cycle. The 2023–present average premium (1.25x, 25 percent) captures the period after the start of the AI investment cycle.

Source – RBC Wealth Management, Bloomberg; data through 8/21/26

This is a line chart showing the relative forward price-to-earnings (P/E) ratio of a broadly defined AI basket of stocks comprised of 70% S&P 500 Info Tech sector and 30% equally weighted Amazon, Alphabet, and Meta Platforms vs. the S&P 500. From 2015 to 2025, this AI cohort commanded an average forward P/E premium of 30% to the S&P 500. That premium has fallen sharply over the past 12 months, with the group now trading close to parity with the broader market on a forward P/E basis.

Trillion-dollar themes

The potential drivers behind this valuation reset can be broadly distilled into several interconnected themes. The first concerns the capex cycle runway and financing. For much of the past three years, the largest tech companies were able to fund their investment programs from internally generated cash flow. That equation has changed as capex has grown much faster than cash flows. Major hyperscalers retain strong balance sheets and ample access to capital, but rising capital intensity means debt and other external financing are becoming more important to the AI buildout. This makes the investment cycle more sensitive to the cost of capital and the willingness of investors and lenders to provide funding. According to one of our third-party research providers, year-to-date AI capex debt issuance has already surpassed $250 billion, nearly doubling last year’s pace.

Next is monetisation. For the capex cycle to remain durable, several things have to happen. Hyperscalers need to generate faster revenue growth and attractive returns from data centres (cloud computing services) that exceed their cost of capital. Innovations in AI products and services need to continue apace in order to drive adoption and increase compute demand. And enterprises need to convert AI experimentation into measurable improvements in sales, costs or productivity. We think the Q2 2026 earnings season provided some encouraging signs that the demand side remains healthy. Microsoft, Amazon and Alphabet – the three largest cloud infrastructure providers and the biggest AI spenders – all reported accelerating revenue growth, strong margins and record backlogs.

We think sustained momentum in financial results will be crucial to bolster confidence that AI spending can generate adequate returns on invested capital. This has implications for the biggest initial beneficiaries of the AI boom, the infrastructure suppliers – particularly semiconductor and memory chip makers, which have captured an outsized share of the economic benefits so far through extraordinarily high profitability. But because tech hardware industries have a long, cyclical history of steep corrections after periods of peak demand, the market is understandably assessing whether current profitability represents a durable increase in earnings base or a temporary windfall that could fade as supply constraints ease.

Can Tech companies keep their return on equity advantage?

• Broad AI basket
• S&P 500

Note: Return on equity is a measure of how well a company uses equity capital, calculated by dividing net profits by average shareholder equity.

Source – RBC Wealth Management, Bloomberg consensus estimates; data through 8/21/26

This is a bar chart showing the calendar year return on equity (ROE) of a broadly defined AI basket of stocks comprised of 70% S&P 500 Info Tech sector and 30% equally weighted Amazon, Alphabet, and Meta Platforms vs. the S&P 500 since 2023. The broadly defined AI basket maintains a higher ROE vs. the S&P 500, but this gap is expected to narrow over the next two years, according to Bloomberg consensus figures. In 2023, the broad AI basket had a ROE of 29% vs. the S&P 500’s 18%; 2024: 30% vs. 18%; 2025: 33% vs. 19%; 2026 estimate: 40% vs. 23%; 2027 estimate: 37% vs. 23%; and 2028 estimate: 32% vs. 22%.

Finally, disruption risks will likely linger. Software has been at the epicentre of early concerns and valuation pressure recently, but AI could reshape business models across a much wider range of industries over time. Meanwhile, the unit economics of AI products and applications remain unproven, with lower-cost models from Chinese developers adding another layer of complexity for markets to digest.

Burden of proof

We believe AI has the potential to become a foundational technology for the economy, with applications spanning consumers and businesses. But the optimism that has propelled AI stocks higher since 2023 has given way to a more discerning and disciplined environment.

Despite strong adoption and growth trends, valuations for the AI equity complex have compressed as market participants have become more conscious of the enormous capital requirements, uncertain monetisation paths and wide dispersion of potential long-term outcomes associated with a rapidly improving technology still in its early stages. Markets tend to price such uncertainty by demanding a larger margin of safety, and we think that adjustment appears to have taken place to some extent. Whether today’s valuations, which appear more reasonable to us, have adequately discounted those risks and the probability of less favourable outcomes will take time to determine.

As the AI investment theme enters the “show me” phase, we believe each quarterly reporting season should help provide additional evidence on whether capital spending is producing commensurate revenue growth and cash flows. For valuations to stabilise or return to a premium, we think markets will need to see the companies most directly linked to AI continue to demonstrate durable demand momentum, improving monetisation and an ability to sustain above-market profitability over the long term.