T. Rowe Price: AI Investment Faces Financing and Valuation Test in 2027

Artificial intelligence investment is entering a new phase in which financing costs, earnings growth and the ability to generate returns will become increasingly important for investors, according to executives at T. Rowe Price.
Speaking at the firm’s 2027 Global Market Outlook press briefing on November 17, executives highlighted the growing role of capital markets in financing AI infrastructure, alongside investment opportunities emerging beyond major technology companies.

Jodi Love, Lead Portfolio Manager at T. Rowe Price, said the AI investment cycle is moving beyond its initial phase, when major technology companies largely financed infrastructure spending through internal resources. “The AI build out is real, but it’s entering the phase where it has to be financed, and then it has to pay returns,” she said.
Citing estimates from J.P. Morgan, Love said approximately $5.5 trillion in cumulative AI and data center funding is expected between 2026 and 2030. Only about 18%, or $1 trillion, is expected to come from hyperscalers’ own cash flow, with the remainder requiring financing through debt and equity markets.
“This is a profound change for companies I have covered for several decades. The first phase of the build out was funded internally, but the next depends on capital markets,” she said.
Love also emphasized the importance of distinguishing access to financing from the ability to generate investment returns. “Financing is not a substitute for paying customers, and access to capital is not the same as creating shareholder value,” she said.
Love explained that higher financing costs could affect future investment decisions even if existing debt remains sound. “Credit doesn’t have to break for the cycle to slow,” she said.
Love cautioned that “technology can succeed without every investment in it succeeding.”
Arif Husain, Head of Global Fixed Income and Chief Investment Officer of the Fixed Income Division at T. Rowe Price, also identified opportunities in hyperscaler debt following a repricing of bonds. He said longer-dated securities could be particularly attractive, as future AI-related issuance is expected to be concentrated in maturities of 10 years or less.
AI Valuations Remain a Live Debate
Sebastien Page, Co-Head of Global Investments and Chief Investment Officer at T. Rowe Price, said the firm’s investment teams continue to debate whether current equity valuations adequately reflect the growth potential of AI.
Traditional valuation measures suggest that U.S. equities are expensive. Page noted that the cyclically adjusted price-to-earnings ratio is in the 99th percentile of its historical range, while the Buffett indicator, which compares stock market capitalization with GDP, is approximately 240%.

However, adjusting valuations for expected earnings growth produces a different picture, he said: “Markets are cheap if we adjust the price earnings ratio by expected analyst earnings growth.”
The question, he explained, is whether those earnings expectations are realistic. Page identified AI token consumption and pricing as indicators worth monitoring when assessing the commercial development of the technology.
Citing data from Ramp covering approximately 22 months, he said token prices had declined 44%, while total token spending had increased 64-fold over the same period.
Page linked the increase in spending to demand for AI computing resources, including AI agents, which can consume significantly more tokens than non-agent-based models.
Page noted that J.P. Morgan estimates AI agents can consume 10 to 100 times as many tokens as non-agent-based models. However, he also acknowledged that token spending had been stabilizing since July, highlighting the uncertainty surrounding future AI revenues.
“We have portfolios that are positioned more defensively on the AI trades, others more aggressively,” Page said.
AI Spending Broadens Equity Opportunities
Love said the AI investment cycle is also creating opportunities beyond the technology companies making the largest infrastructure commitments.
She identified power generation, grid equipment, electrical infrastructure, semiconductor supply chains and industrial companies as areas benefiting from AI-related capital expenditure. “Capex is a cost to the spender and revenue to the supplier,” Love said.
Meanwhile, equity market performance in 2026 has already shown signs of broadening beyond major technology stocks. According to Love, the Russell 1000 Growth Index was down approximately 5% year-to-date as of November 17, compared with a 25% gain for the Russell 1000 Value Index.
“Returns this year were actually earned by profits, not by promises or rising valuations,” she said.
Love also pointed to the concentration of the S&P 500, where the 10 largest stocks account for approximately 40% of the index, compared with a long-term average of 23%.
She argued that historical periods of similarly elevated concentration have been followed by stronger relative performance among the remaining constituents, although this does not necessarily imply a collapse in mega-cap technology stocks.
Tim Murray, Capital Markets Strategist at T. Rowe Price, also addressed AI’s broader economic impact, noting that information technology investment has become a major contributor to U.S. growth.
Murray expects AI investment growth to moderate in 2027 but does not anticipate a significant slowdown in infrastructure development over the next couple of years.
He also argued that AI adoption has so far been gradual, with companies often using the technology to expand output without proportionately increasing employment rather than undertaking widespread layoffs.
Beyond the implications for financial markets, Page discussed how T. Rowe Price is incorporating AI into its own investment research processes. He said the firm’s historical research database, including approximately 10,000 analyst research notes produced annually, provides an opportunity to train AI systems using proprietary investment information.
Rather than relying solely on commercially available AI models, the firm is exploring whether its own research data can help generate investment ideas that differ from conventional human analysis. “We run AI, we train AI on some of our processes, and it’s recommending things that sometimes human would not have thought of,” Page said.