The median is down 53%: Morgan Stanley (MS.US)'s 93-page report sets the tone for AI buying on falling

According to Woofun AI, Morgan Stanley (MS.US) released a 93-page global themed strategy report on November 8, with the title “Buy the Drop” (Buy the Drop), which clearly characterized the impact of the US midterm elections on the AI sector as a rhythm adjustment rather than a loss of demand, and called on investors to take advantage of the market panic layout.

In the 2026 midterm elections, the Democratic Party won both houses of Congress and won the governors of Texas, Ohio, and Georgia at the state level. Gina Hinojosa, the governor designate of Texas, proposed a suspension of approval for the new data center, and Dr. Amy Acton, the governor designate of Ohio proposed a conditional suspension, and market panic spread. According to the report, projects already under construction will not be affected, and the 2027 production capacity forecast will remain unchanged. The core variable is in 2028: Approximately 5 GW of IT capacity (approximately 20% of the original US 2028 projects and 10% of related projects worldwide) was delayed from Texas and Ohio until 2029.

This means that elections are changing the pace of construction, not AI requirements themselves. The report maintains the 2027/28 capital expenditure forecast for hyperscale cloud vendors totaling $1.6 trillion/$1.8 trillion. According to data compiled by Woofun AI, the new capacity of US data centers in 2028 was lowered from 24.9 GW previously predicted to 19.9 GW, and the year-on-year growth rate changed from positive to -4%; however, delayed production capacity did not disappear; instead, it was stacked up until 2029, causing the increase to jump from 22.7 GW to 27.6 GW in that year, reaching 38% year-on-year growth.

This “postponement rather than cancellation” pattern brings benefits along with a smoother rebalance between supply and demand in the supply chain, improved credit market prospects, and higher computing power rental prices. At the global level, total capacity was fine-tuned from 224 GW to 219 GW in 2028, and remained unchanged at 271 GW in 2029 and beyond; the electricity gap widened from 38 GW to 44 GW, accounting for 38% of the electricity needed, highlighting the scarcity of electricity.

Damocar disassembled the “buy when falling” strategy into three baskets. The first basket was for hyperscale cloud vendors. Analyst Brian Nowak gave an increase rating for META, GOOGL, and AMZN, and Adam Wood increased MSFT because the geographical layout of these companies was scattered and capital expenditure guidelines were not adjusted. As of November 5, META closed at $605.08, target price of $820, and upside +35.5%; GOOGL closed at $282.57, target price of $375, space +32.7%; AMZN closed at $260.21, target price of $335, space +28.7%; MSFT closed at $535.93, target price of $650, and space +21.3%.

The second basket is the select data center supply chain. It is recommended to buy power and infrastructure companies that have already fallen out of value during the pullback, and be wary of the risk of inventory accumulation in 2028. High-confidence benchmarks include Bloom Energy (BE.US), Solaris (SEI.US), Vistra (VST.US), Talen Energy (TLN.US), and Constellation Energy (CEG.US). Even if data center growth falls short of expectations, their risk-return benefits are still beneficial. In terms of hosting data centers, Equinix (EQIX.US) has only 9 operating centers in Texas, accounting for about 3% of the world's 280 seats. It has limited exposure and may benefit from higher pricing.

The third basket is software supporters. The report predicts that value creation is shifting from semiconductors and infrastructure to software and application layers. Cybersecurity and infrastructure software is the preferred entry point, with additional targets including Palo Alto Networks (PANW.US) and CrowdStrike (CRWD.US); the SaaS side added Snowflake (SNOW.US), Datadog (DDOG.US), and MongoDB (MDB.US).

The semiconductor sector has already declined deeply, and the report believes it can be picked up. As of November 6, the median semiconductor stocks covered by Damo fell 53% from the 52-week high, with an average decline of 48%. The report highlights that the core pillar (demand) of AI spending has not been damaged, and token usage is still growing exponentially. Compute is the highest confidence sub-sector; NVDA and AVGO are about 9 times CY28 P/E. NVDA was the first choice for three reasons: it has a wider customer base than AMD/ASIC; it can grow by increasing revenue per watt in a power-limited environment; the revenue per GW of the next-generation Feynman architecture is expected to rise from Rubin's approximately $40 billion to over $50 billion; and its leading position in tokens/watt metrics. Storage preferences are MU and SNDK. The report specifically mentions SanDisk (SNDK.US). Based on the closing price of $835 on November 6, if the repurchase scale is around 50% of revenue, more than 50% of tradable shares can be repurchased by the end of 2028.

In addition, NXPI and ADI are preferred for analog chips, and ALAB and MRVL are preferred for optical communication.

The value of AI adoption is moving from experimentation to quantifiable rewards. Citing AlphaWise's analysis of approximately 18,000 financial reports and meeting minutes, the report found that 44% of companies in the S&P 500 mentioned at least one quantifiable AI benefit in the past 12 months, compared to 32% a year ago. The share of companies that mentioned cost savings or AI-driven revenue growth rose from 16% to 26%. The financial industry is at the forefront: 67% of financial firms mentioned tangible improvements brought about by AI, compared to only 42% a year ago. Based on this, the report suggests broadening investment horizons from AI builders to AI adopters, focusing on banks, insurance, and consumer companies. This is a key signal of the migration of market logic to the application layer and value delivery layer following the pullback of the infrastructure layer.