The sharp sell-off on 5 November rattled investors in technology stocks. According to Edmond de Rothschild, it mainly reflects a forced reduction of positions rather than a fundamental trend reversal. The risk, however, is shifting towards second-tier AI suppliers.
On Thursday 5 November, the Nasdaq fell 5.5%, the SOX semiconductor index dropped 12% and the S&P 500 lost 4.4%. A one-day decline of that size is rare: since 2000 the Nasdaq has suffered one only 42 times, on fewer than 1% of trading days.
Midterms as accelerator – not as trigger
According to the Swiss bank, the US midterm elections accelerated existing doubts about artificial intelligence. Those doubts had been building for weeks: since the 10-year US Treasury yield peaked on 5 October, defensive sectors have led the market while technology and data- centre stocks have lagged.
The bank sees three drivers: uncertainty about the pace and profitability of AI investment, fears for growth if Democratic lawmakers tightly restrict data-centre construction, and heavy selling by systematic strategies, amplified by margin calls.
Political risk – more local than federal
According to a tally by The Washington Post , around 30 local governments voted during the midterms on moratoriums or restrictions on AI data centres. These could affect roughly 45% of planned US expansion capacity. Even so, Edmond de Rothschild considers it premature to expect a structural rejection of AI investment. The Democrats have regained control of both chambers of Congress, but lack the two-thirds majority needed to override a veto by President Trump. The bank expects delays and stricter building conditions rather than a halt to the investment cycle.
Valuations – a cushion for leaders, risk for suppliers
Edmond de Rothschild tested around 100 technology stocks against scenarios in which hyperscalers' AI investment through 2030 comes in 10% to 50% below expectations. Consensus forecasts capital expenditure by Microsoft , Amazon , Alphabet , Meta and Oracle of $847 billion in 2026 and $1,345 billion in 2030, up from $379 billion in 2025. Even a 30% shortfall would leave 2030 spending at 2.5 times its 2025 level.
Share prices of direct beneficiaries already reflect a marked slowdown. To justify current prices, investment would have to fall by 80% for memory makers, 60% for hyperscalers and 52% for AI chip designers. Semiconductor equipment (3%), electrification (13%) and data-centre infrastructure (-1%), by contrast, are priced as if everything goes to plan. For these second-tier suppliers, estimated intrinsic value turns negative if investment falls by just 10% to 15%.
A low multiple is not automatically a bargain, though. Nvidia and Broadcom trade at around 10 times expected 2030 earnings because the market doubts their sustainability. Memory makers appear to be valued at the peak of their cycle. And neoclouds such as CoreWeave, Nebius and Iren remain hard to value given their dependence on external financing.
Cautious short term – constructive medium term
In the short term, the risk of further deleveraging remains significant. The VIX futures curve is inverted, leverage in AI strategies is still substantial and the Nasdaq has slipped below its 200- day moving average. Historically, however, such declines have often been followed by recoveries: three months after a one-day drop of more than 5%, the Nasdaq was on average around 4% higher. Since 2010, the three-month return has been positive every time, averaging 19%.
Edmond de Rothschild advocates selectivity. High-quality technology companies with solid balance sheets and good revenue visibility could offer attractive entry points. Companies that rely heavily on external financing and on a continued acceleration in infrastructure spending carry a higher risk profile. In the bank's view, the correction may not mark the end of the AI cycle, but rather the end of a phase of unchecked expansion. The market now demands greater financial discipline and proof that investments can generate returns.
