UBS says tech-stock volatility is at its highest since the dot-com bust
A new report from UBS research unit HOLT says volatility in the global technology sector has risen to its highest level since the collapse of the dot-com bubble. Investors are reassessing whether the cash-flow returns from AI-driven hyperscale data centers and semiconductor companies can be sustained. HOLT Unit, also known as the HOLT Team, is UBS’s internal core quantitative group focused on value-driven investment research. It is not an independent legal entity but a specialized team within UBS’s global investment bank or wealth-management research operations. The bank’s analysis says AI infrastructure spending is creating increasingly heavy pressure. Although profit margins have remained stable, the massive investments by large hyperscale data-center operators are weakening their asset efficiency and are expected to drive down their cash-flow return on investment, or CFROI, by 2028. UBS estimates that the five largest hyperscale data-center operators—Microsoft, Meta, Alphabet, Amazon and Oracle—will face a combined funding shortfall of $227 billion next year, leaving them unable to meet their operating and financing commitments. Historical data shows that, since 1998, there have been about 650 major capital-spending surges. Sixty percent of those surges resulted in a permanent decline in CFROI, with the effect most pronounced among companies whose initial returns were already high. The semiconductor industry faces a different set of risks, UBS said. The sector’s returns have nearly tripled to about 30%, a level reached by fewer than 1% of companies since 1990. Current valuations assume that these high returns will continue for five years, an assumption UBS says runs counter to a typical competitive landscape. The report also cites Chinese AI developers DeepSeek and Moonshot AI as examples suggesting that the industry’s so-called moats may not be insurmountable, given China’s tendency to prioritize market share over profitability. Meanwhile, valuations for software, enterprise-data and services stocks have been reset amid concerns about AI-driven disruption. Their aggregate price-to-book ratio has fallen by about 40% over the past 18 months. UBS found that, historically, 80% of stocks that experienced a decline of a similar magnitude failed to recover their previous valuation levels within 10 years. The price-to-book ratio, or P/B and PBR, is the ratio of a stock’s price per share to its book value per share. It reflects how much common shareholders are willing to pay for each 1 yuan of net assets and indicates the market’s assessment of a company’s asset quality. Looking beyond the technology sector, UBS said value stocks and low-volatility stocks were the strongest-performing style factors during past selloffs triggered by technology stocks. Value stocks performed well in all six major selloffs since 2004, while low-volatility stocks performed well in four. However, UBS warned that both approaches have limitations in the current environment. Since 2023, the traditional relationship between value stocks and the economic cycle has weakened. Low-volatility stocks also tend to perform poorly outside selloff periods unless they have strong fundamentals.
HOLT Unit, also known as the HOLT Team, is UBS’s internal core quantitative group focused on value-driven investment research. It is not an independent legal entity but a specialized team within UBS’s global investment bank or wealth-management research operations.
The bank’s analysis says AI infrastructure spending is creating increasingly heavy pressure. Although profit margins have remained stable, the massive investments by large hyperscale data-center operators are weakening their asset efficiency and are expected to drive down their cash-flow return on investment, or CFROI, by 2028.
UBS estimates that the five largest hyperscale data-center operators—Microsoft, Meta, Alphabet, Amazon and Oracle—will face a combined funding shortfall of $227 billion next year, leaving them unable to meet their operating and financing commitments.
Historical data shows that, since 1998, there have been about 650 major capital-spending surges. Sixty percent of those surges resulted in a permanent decline in CFROI, with the effect most pronounced among companies whose initial returns were already high.
The semiconductor industry faces a different set of risks, UBS said. The sector’s returns have nearly tripled to about 30%, a level reached by fewer than 1% of companies since 1990. Current valuations assume that these high returns will continue for five years, an assumption UBS says runs counter to a typical competitive landscape.
The report also cites Chinese AI developers DeepSeek and Moonshot AI as examples suggesting that the industry’s so-called moats may not be insurmountable, given China’s tendency to prioritize market share over profitability.
Meanwhile, valuations for software, enterprise-data and services stocks have been reset amid concerns about AI-driven disruption. Their aggregate price-to-book ratio has fallen by about 40% over the past 18 months. UBS found that, historically, 80% of stocks that experienced a decline of a similar magnitude failed to recover their previous valuation levels within 10 years.
The price-to-book ratio, or P/B and PBR, is the ratio of a stock’s price per share to its book value per share. It reflects how much common shareholders are willing to pay for each 1 yuan of net assets and indicates the market’s assessment of a company’s asset quality.
Looking beyond the technology sector, UBS said value stocks and low-volatility stocks were the strongest-performing style factors during past selloffs triggered by technology stocks. Value stocks performed well in all six major selloffs since 2004, while low-volatility stocks performed well in four.
However, UBS warned that both approaches have limitations in the current environment. Since 2023, the traditional relationship between value stocks and the economic cycle has weakened. Low-volatility stocks also tend to perform poorly outside selloff periods unless they have strong fundamentals.