TECHNOLOGY SECTOR — Buy the Drop: The Post-Midterm AI Playbook and Its Best Ideas

From The Heat Map · 11:54

Show by Eric

A major bank's global thematics team published "Buy the Drop" on November 8, 2026. Its thesis: the market's reaction to the US midterms is overdone. Hyperscaler cap ex projections are unchanged in aggregate — about $481B in 2025, $951B in 2026, $1,641B in 2027 and $1,760B in 2028 — but a new data center review process shifts roughly 20% of US projects (about 5 GW) from 2028 into 2029. US builds go roughly flat 2027 to 2028, then up about 40% in 2029, and the powered data center shortfall exceeds 30% through 2028 (44 GW in 2028). The report argues the push-out elongates the cycle: smoother supply/demand rebalancing, a better credit outlook, and higher compute lease rates — all favoring the largest, best-capitalized scaled leaders. Its best ideas: Overweight hyperscalers (AMZN, META, GOOGL, MSFT, plus SPCX); compute as the highest-conviction semi bucket with NVDA as top pick ($300 target) alongside AVGO ($505) and TSMC; selective memory; de-rated hardware (STX, WDC, LITE, COHR, ANET, KEYS); time-to-power names (EQIX and the powered shell providers, behind-the-meter BE, SEI, GEV, LBRT, VST, TLN, CEG) and Asia power winners; and a rotation call broadening alpha from semis and infrastructure into software (SNOW, DDOG, NET, PANW, CRWD, NOW among the named) and early AI adopters. Its named risks: a 2028 supply-chain air pocket, regulatory tightening on new builds, and Asia cloud-semis upstream suppliers exposed to the 2028 push-out over the next two to four quarters. The desk overlay, from the desk's own AI underwriting card (debt and stability, normalized growth, five-year underwriting returns): agreement on NVDA as number one, and on AVGO, TSMC, GOOGL and MSFT. Three plain divergences: (1) memory — the report's largest target upsides are MU ($1,200) and SNDK ($1,750); the desk carries MU on its still-have-downside list (low stability, commodity earnings stream, about 5.7%/yr mid-cycle underwriting return) and reads memory as where a 2028 air pocket lands first; (2) ANET — the report is Overweight at $230; the desk carries it on still-have-downside (most concentrated revenue in its underwritten set, 54–58x normalized multiples); (3) software — the desk's only software name in this frame is PANW, buying its remaining software economics inside MSFT and AVGO instead. Desk synthesis: own the earnings that are contracted, backlogged and scarce — the grid and the tolls — and treat the report's memory and ANET enthusiasm as the trade, not the underwrite. For informational purposes only. This is not financial advice. Please do your own research.

More episodes of The Heat Map

More episodes

PodSearch is a free, independent podcast search engine. Read about PodSearch, the privacy policy or contact us.