Q2 2026 M&A: record value, the thinnest deal count since 2010

Three totals for one half-year

H1 2026 global M&A value has three credible numbers and they do not reconcile. Every figure below is a first cut published within two weeks of quarter-end, and first cuts revise upward, so where providers disagree we name both.

• US$2.8 trillion, per Reuters on LSEG data, 1 July. Up 48% year on year and the highest first half since LSEG records began in 1980.
• US$2.85 trillion, per a slightly later LSEG cut. Up 50%. Same provider, different cut-off date.
• US$3.16 trillion, per Mergermarket's 8 July league tables. Up 44%. Mergermarket counts assumed debt and includes transactions LSEG excludes, so the roughly US$300 billion spread is definitional.

The gap survives down to a single deal: NextEra Energy's combination with Dominion Energy is US$66.8 billion on LSEG's equity basis and US$118.8 billion on Mergermarket's. Cite one provider per number, and name it.

Record value, fewest deals since 2010

Q2 2026 was roughly US$1.6 trillion, up 31% on Q1 and a record quarter on LSEG data. Deal count collapsed at the same time.

• 10,309 transactions in Q2, the lowest quarterly count since 2010, per A&O Shearman.
• About 24,000 deals in H1, down 9% year on year and a six-year low, per Reuters on LSEG data.
• 47 deals above US$10 billion carried more than US$1.3 trillion between them, close to half of all global value and the highest concentration on record.
• Technology led at US$649 billion, up 90%. The US reached US$1.5 trillion, 54% of global value; Europe more than doubled to US$676 billion.

More capital moving through far fewer processes.

Why were sponsors absent?

Corporates could pay and sponsors could not.

• Rates went the wrong way. The Fed held at 3.50% to 3.75% on 17 June and the ECB raised all three key rates 25 basis points on 11 June, its first increase since 2023.
• Investment grade stayed open. Global IG corporate issuance hit US$3.4 trillion in H1, the highest first half on LSEG's record.
• Buyout debt closed. PitchBook LCD records sponsored M&A loan volume down 54% quarter on quarter and recent LBO spreads at S+509 against S+474 in Q1.
• US PE value fell to US$177.3 billion, down 37.5% sequentially, with exit value down 46.3% (PitchBook). Sponsors kept transacting. They transacted smaller.

Strategic acquirers took 78% of volume, per Mergermarket.

What this changes in the data room

Longer timetables and stock consideration break the assumption every legacy VDR is built on: one seller, disclosing one way, to many bidders.

Rooms now stay open for years. Bain puts announcement to close above US$10 billion at about seven months; NextEra and Dominion guide to 12 to 18 months across six regulatory tracks. CogniSuite resolves access per folder, layering user overrides over org grants over inherited parents, so a room survives turnover without a rebuild.

Stock consideration makes diligence two-way. Bain counts stock plus cash megadeals at a historical high and all-cash at a cyclical low. CogniSuite knows which side of the table you sit on: set the mandate buy-side and the seller becomes the external party, blocked from the internal room by default.

More counsel, more jurisdictions, narrower slices. Folders carry a visibility label, including one readable by a single named org. View-only folders can never be exported, in any format or bulk ZIP. Watermarks are burned in at serve time with the viewer's name, org and timestamp. Bankers can preview the room as a specific buyer, resolved through the real permission engine.

AI exposure became its own workstream. In PitchBook's Q2 survey, AI disruption ranked with rates and geopolitics as a top macro factor, and software was the sector most cited as pulled back from. Every CogniSuite AI answer retrieves only documents the asker may read, and counterparty-facing drafts are grounded only in what that side may read, with citations verified against source text before anyone sees them.

What we do not do

• No reconciliation engine. Our AI is instructed to surface conflicting figures with a source and date. Nothing extracts numbers across a whole room and flags contradictions on its own. Ask any vendor claiming otherwise to prove it on your files.
• Watermarking is visible provenance, not forensic marking. PDF, Word, images, Excel and PowerPoint all carry the mark, and a format that cannot be marked is refused rather than served clean, but a recipient who photographs or retypes a page defeats it.
• The audit trail is oversight, not evidence. It is readable by the advising firm rather than mutually, and it is not cryptographically tamper-evident.

Our posture is documented at /security.