Q2 2025 M&A: value up, deal count down, and longer merger reviews

Why the headline Q2 2025 numbers do not agree with each other

There is no single reconciled global figure for Q2 2025. Cite PitchBook's own report page. If only the secondary summary is retrievable, keep the current attribution wording but add: "We could not retrieve the underlying report and are relying on that summary." so the sourcing weakness is visible to the reader. Mergermarket and Dealogic, in M&A Highlights 1H25, report first-half value of USD 2tn, up 25% year on year, on 16,663 deals, down 16% and the lowest count since the first half of 2005. PwC, in its Global M&A Industry Trends 2025 Mid-Year Outlook, frames the same half as values up 15% and volumes down 9%.

These are not the same measurement. Mergermarket counts large minority stakes and financings that other providers exclude. We have not attempted to reconcile them and no one should quote a single blended global Q2 number as if it existed.

Two caveats worth carrying through the rest of this piece. The Mergermarket data has a cut-off of 23 June 2025, so its second quarter is incomplete and the final figures are higher. And most of what it reports is half-year, not quarter. Where a datapoint below is genuinely Q2-specific, we say so.

What the sources do agree on is the shape. Value rose. Count fell. Mergermarket counted 33 deals above USD 10bn in the first half, the most since the second half of 2020, while the mid-market between USD 200m and USD 1bn recorded USD 330.8bn across 776 deals, down 3.2% in value and 1.4% in count against the same period of 2024. Fewer processes ran, and the ones that ran were larger transactions.

Sector leadership is only available on a half-year basis. Mergermarket puts technology first at USD 478bn, 24% of the global total and up from 19% a year earlier, with financials second on the back of Chinese state-led bank recapitalisations including a USD 22.7bn placement in Bank of China, and healthcare third at USD 176.6bn. We found no verified Q2-only global sector table.

Where the April tariff announcement shows up in the data

The 2 April tariff announcement shows up in the monthly series. Mergermarket records EMEA deal volume falling 39% month on month in April to USD 64.2bn. North American volume also dipped in April, then rose 46% month on month in May to USD 207.2bn.

The regional split held for the whole quarter. EMEA volume in Q2 2025 was down 24% on Q1 2025, and its first-half total of USD 430.5bn across 6,451 deals was the lowest deal count in twelve years, cutting its share of global activity to 21% from 25%. North America grew value 11% to USD 969.8bn but its deal count fell 17% to 5,038, the lowest since 2009. Asia-Pacific was the outlier, with volume up 97% to USD 572bn and share up to 28.4% from 18%, driven by Japan up 3.7 times to USD 158.7bn and mainland China up 2.3 times to USD 231.1bn.

Financing cost does not explain the European weakness. The ECB cut its deposit facility rate to 2.25% on 23 April 2025 and to 2.00% on 11 June 2025, while the Federal Reserve held its target range at 4.25% to 4.50% through the entire quarter, per the FOMC rate-change record. Europe eased and European buyouts still fell. That points to uncertainty rather than the cost of borrowing as the binding constraint.

We could not find verified Q2 2025 data on leveraged loan issuance, high-yield supply, private credit volumes, or average purchase multiples. Policy rates are not the same as deal financing conditions and we have not substituted one for the other.

What happened to sponsor buyouts and exits in the quarter

Private equity showed the same pattern as strategic M&A. Mergermarket reports global buyout volume up 35% to USD 359bn on a 7% fall in deal count to 1,398. Six buyouts above USD 10bn totalled USD 101bn, against three worth USD 34bn a year earlier, and four of the six were signed in Q2 2025. The largest was the USD 24bn consortium bid for Santos by XRG, ADQ and Carlyle, announced 16 June 2025.

Exits tell the harder story. Global sponsor exit volume rose 45% to USD 279bn while the number of exits fell 7% to 613. Trade sales were 64% of exit volume and secondary buyouts 33%. Mergermarket's wording is 'Only 19 IPOs took place in 1H25, with 18 seen in the same period last year. Just nine IPOs priced in 2Q25, down from 11 in 2Q24.' The source uses 'priced' only for the nine in Q2; the 19 'took place'. Change to 'Only 19 sponsor-backed IPOs took place in the half, and just nine priced in Q2 2025, down from 11 in Q2 2024.' Continuation funds absorbed 27 companies year to date against 26 a year earlier, which is flat. So exit value went up because a few very large assets cleared, not because the exit market reopened.

Within North America the quarter cooled sequentially. Sponsor-led buyouts were USD 83.6bn in Q2 2025, down 14% on Q1, and sponsor-led exits were USD 92.4bn, down 12%.

How merger review timelines and remedies changed

This is the hardest timeline evidence in the quarter. The Dechert Antitrust Merger Investigation Timing Tracker for Q2 2025 puts the average duration of significant US merger investigations at 13.6 months, 20% above the 11.3-month average for full-year 2024, with an H1 2025 average of 12.6 months. EU Phase I remedy cases concluded in the quarter averaged 13.8 months, roughly four months longer than the 2011 to 2022 average.

The offsetting development was that settlements returned. Dechert counted five consent decrees in Q2 2025, more than the previous nine quarters combined and the most in a single quarter since Q2 2022. McDermott Will & Emery notes the FTC reinstated early termination of the HSR waiting period, with more than 100 requests granted since the programme restarted. The net effect was longer investigations for contested deals, faster clearance for uncontroversial ones, and a workable path to a negotiated fix where there had been very little for several years.

What the quarter meant for diligence practice

Tariff exposure does not appear in a target's historical financials, so it had to be diligenced directly. Mayer Brown set out what that meant in practice: raw materials, components and supplies, cost of goods sold, demand impact, alternative sourcing, and a contract-by-contract read of supplier and customer agreements to work out who bears an increased cost. The same note calls the exercise "extremely challenging and somewhat speculative". Sellers pushed for tariff exclusions from MAE definitions, buyers pushed for tariff-specific closing conditions, representation and warranty insurers added broad tariff exclusions, and earnouts became more common.

The workload consequence is specific. Contract review moved from a sampled exercise to a population exercise, because the answer to "who bears the tariff" lives in individual pass-through, price adjustment and force majeure clauses rather than in a summary schedule. Link each figure to the publisher's own page for the specific publication, or drop the Pulse Survey figures. Verify that the linked page contains the 30% and 51% numbers before publication; if it does not, remove the sentence entirely rather than re-point it. The deals that continued absorbed the extra scope.

We could not find any survey that measured due diligence duration during Q2 2025 itself. The frequently cited structural timeline data, summarised by Goodwin from Boston Consulting Group analysis, covers deals through 2022, not this quarter. It is context, not measurement.

Two things follow for how a data room should work under these conditions. Requests arrive as spreadsheets in whatever layout the requesting side uses, and a longer sign-to-close window means the same question gets asked again by a different party months later. CogniSuite handles the first by having the model produce a reading plan for the spreadsheet, which columns and rows mean what, and then executing that plan mechanically against the raw cells, so a request cannot be paraphrased or invented in transit, with coverage warnings when rows may have been dropped. It handles the second by scoring new requests against existing ones and, on the deal team's approval, attaching the already-confirmed answer documents.

Where this stops is worth stating. There is no engine that reconciles a figure in one document against a conflicting figure in another. That behaviour rests on the model's instructions and on both documents happening to be retrieved together, which is not a control. Scanned PDFs without a text layer are not readable by the ingestion pipeline. And each document is represented by a single embedding over its opening text, so retrieval points at a document rather than at a clause inside a long supply agreement. For the tariff work described above, that means the software narrows where to look. A person still has to read the contract.