Goldman: M&A up, $700B AI capex

Goldman’s M&A desk has flagged a near-term uplift in dealflow with projections that global announced volumes could rise from roughly $3.1 trillion toward as much as $3.9 trillion in 2026—figures cited alongside Dealogic data by Bloomberg’s market coverage. (bloomberg.com) Goldman Research calculates that hyperscaler capital spending would need to reach about $700 billion in 2026 to mirror the peak intensity of the late‑1990s telecom investment cycle, while the consensus capex estimate for the large hyperscaler group was recently revised to roughly $527 billion. (goldmansachs.com) Aggregate guidance from the big cloud platforms—Alphabet (Google), Microsoft, Meta and Amazon—has pushed combined 2026 capex expectations close to the $700 billion scale, according to CNBC reporting that tallied company outlooks. (cnbc.com) Goldman’s 2026 Global M&A Outlook explicitly labels AI as an “innovation supercycle” that is expanding the strategic rationale for transformational deals and increasing private‑market sponsor activity in larger transactions. (goldmansachs.com) Goldman warns that supply‑chain bottlenecks and shifts in investor appetite, rather than hyperscalers’ cashflow or balance‑sheet capacity, are the more probable ceilings on deploying capex at the $hundreds‑of‑billions scale. (goldmansachs.com) Practical project blueprint: assemble AI‑tagged M&A timestamps from Dealogic, daily equity returns from CRSP and daily Fama‑French factor series to compute abnormal returns with an estimation window of −250 to −31 trading days and an event window of −1 to +30 days. (iongroup.com) Estimate abnormal returns using a market‑model and Fama‑French 3/5‑factor specifications, then run cross‑sectional regressions of cumulative abnormal returns on deal size, acquirer hyperscaler status, and PE‑sponsor involvement following canonical event‑study protocols (MacKinlay 1997; Brown & Warner 1985). (researchgate.net)