Morgan Stanley reviews the AI crash: It's the valuation that's being slashed, not the profits! But don't rush to bottom-fish.

Odaily Finance APP notes that this global correction in AI stocks is less about a collapse in fundamentals and more about a pure “valuation bubble squeeze.” In a report released on November 8, Morgan Stanley’s Quantitative Equity Research team reached a rather counterintuitive conclusion: this round of decline in the AI sector was driven almost entirely by valuation contraction—while consensus earnings expectations have not been revised downward but are instead steadily climbing.

Intensity and Structure of the Decline

From the close on September 30 to November 5, the AI & Tech Diffusion Index compiled by Morgan Stanley dropped a cumulative 26.6%, while the MSCI ACWI Global Index fell only 3.7% during the same period; extending the calculation to Japan and Asian-listed constituents through the close on November 6, the group’s loss expanded to 28.4%, representing a 33% drawdown from its historical peak on October 6. The correction unfolded in two waves: the first in mid-October, and the second starting at the end of October, which deepened significantly after the US midterm elections on November 3.

Valuation contraction “accounted for more than all the declines.” The report decomposed the prices for 91 constituents with positive earnings in both periods: the equal-weighted expected FY2 EPS in each region rose by about 3 log points, while price-to-earnings ratios collectively contracted by about 37 log points (ranging between 26 to 48 log points across regions). In other words, the decline in valuations not only explains all of the declines, but even more than that. This suggests investors may have priced in the risk of a slowdown in the AI infrastructure’s growth before it was reflected in consensus forecasts.

Regional Divergence: Japan’s Valuation Cuts Are the Deepest 

On a percentile basis from their own valuation history (since September 2019), US and European AI stocks have fallen from the middle of their ranges to the lowest quintile (median percentiles of 17 and 13, respectively), while Japan plummeted from the 82nd to the 28th percentile, making it the most typical sample of “resilient earnings, valuation contraction.” Conversely, Asia/emerging markets remain near their historical median (53rd percentile), but have the highest share of downward revisions—7 of 26 constituents (27%) saw a downward revision of more than 0.5% in FY2 forecasts.

How Thick Is the Valuation “Safety Cushion”?

Morgan Stanley conducted a stress test: assuming a hypothetical 20% cut in 12-month forward EPS, the share of constituents trading above their own historical median P/E would jump from the current 27% to 55%. The “breakeven downward buffer” for median EPS across regions is roughly: US 24%, Europe 23%, Japan 16%, while Asia/emerging markets, with over half already above their medians, have a negative median buffer—practically no safety cushion.

“Adopters” Outperform, but Earnings Momentum Remains Upstream

Since the end of September, the AI Adopters basket returned -3.2%, while Enablers dropped -29.0%—a nearly 26 percentage point difference, which remains substantial even after stripping out market beta.

It is important to note, however, that this style shift is “valuation-driven” rather than “earnings-driven”: Enablers’ median forward P/E fell from 27.4x to 18.6x, almost in line with Adopters at 18.1x, but their net breadth of earnings upgrades remains at +48%, well above Adopters at +2%, with 35% of Adopters seeing downward revisions in FY2 forecasts.

Capital Expenditures Come into Focus

The report also finds that stocks with the largest earnings upgrades fell the hardest (rank correlation -0.42); companies with heavy capital expenditures experienced an average “valuation cut” of about 40 log points after earnings, significantly higher than others at 25. Controlling for unchanged earnings upgrades, every 10% upward revision in capital expenditure forecasts led to approximately a 4 percentage point relative underperformance (t-value -2.8), reflecting that the market is increasingly scrutinizing the returns on incremental AI investments.

Three Core Strategies 

Morgan Stanley’s central conclusion is: valuation resets have improved AI’s entry points, but earnings realization remains key. Strategically: first, avoid indiscriminately bottom-fishing upstream AI stocks, and be wary of delayed impacts of greater regulation and project delays on datacenter orders and profits;

Second, selectively overweight AI Adopters, prioritizing names with resilient earnings forecasts rather than adding across the board;

Third, focus on regions with thicker valuation cushions (US and Europe are the thickest, Japan has seen the deepest valuation cuts), and keep a close eye on earnings revision trends—especially among the 33 constituents whose latest reports have yet to be included. On the macro front, Iran conflict and energy market volatility have amplified risk-off sentiment, and the renewed upward trend in yields remains an additional valuation risk, independent of earnings.