Strategics Drove a Record H1. Anthropic's Compute Lease Has a Kill Switch.

Throughout the year we've covered the rising gap between deal value and deal volume, the trend that describes the majority of 2026 M&A value concentrated in a few, very high-value transactions. What we've talked about less is where these deals are actually happening, and what's driving the large transactions.

From the outside looking in, those numbers would seemingly indicate the private equity comeback that has been forecasted since the Fed started cutting rates in 2024, but that's not the case. There was roughly $2.85 trillion of M&A in H1 2026. Sponsors did participate in the rebound, but strategic buyers drove the headline megadeals and the majority of overall value.

There were 47 megadeals (transactions above $10 billion) in the first half of the year, and together they carried a value of about $1.3 trillion, nearly half the global figure concentrated in fewer than fifty transactions. PwC's read is the same: the Americas produced 61% of global deal value on just 28% of volume, led by US megadeals. Technology led every sector, with roughly $649 billion in announced deals in the first half (LSEG), the single largest concentration of value in the market. These are corporate buyers using scale, stock, balance-sheet cash, and readily available debt financing to pursue transformative combinations.

Sponsors are in a different position entirely. Private equity deal count was essentially flat in early 2026, and US buyout funds were sitting on roughly $1.13 trillion of dry powder as of June, capital raised but not deployed. In 2025, fundraising hit its weakest level since 2020, IPO exits stayed narrow, and hold periods have stretched into six and seven years.

Continuation vehicles and GP-led secondaries are increasingly becoming a way for PE firms to return liquidity to investors who may be feeling like their money's been wrapped up just a bit too long. Continuation vehicles, where a firm sells a portfolio company out of an old fund into a new fund which it also manages, are mainstream at this point. GP-led secondary volume rose 53% in 2025, to about $115 billion, according to Jefferies, while GCM Grosvenor estimates that continuation vehicles now represent roughly 19% of sponsor-backed exit volume. Eighty-three of the world's 100 largest buyout sponsors have used one.

When strategics dominate a cycle, more of the strategic thesis, internal alignment, synergy underwriting and integration planning sits with corporate development teams. That should be supportive of in-house hiring, particularly for professionals who can execute a transaction and help carry it into integration.

Safety commitments in AI have always felt a little abstract to me. If robots were taking up arms against us it would be one thing (could still happen), but we're often talking about less cinematic risks, possibilities of cyber incidents, unaligned objectives, and the spread of misinformation. How exactly do we police the major AI labs to enforce safety regulations?

Contractual clauses could be a start, and we're seeing a major one between Anthropic and SpaceX.

Back in Issue 11, we discussed Anthropic's leasing of SpaceX's Colossus supercomputer. The compute contract was disclosed in SpaceX's S-1 ahead of its June IPO. Anthropic pays SpaceX roughly $1.25 billion a month through May 2029 for dedicated access to the Colossus I and II clusters, more than $40 billion if maintained over the full term, across roughly 325,000 Nvidia GPUs.

An interesting detail within the contract is that SpaceX can reportedly reclaim the compute if Anthropic's models are found to engage in actions harmful to humanity.

Neither company has confirmed its exact wording or status, and it has been widely reported rather than published. But the structure is still worth noting. If Musk's restriction is reflected in the contract, and not simply a public statement, it would turn safety conduct into a condition attached to critical infrastructure access. Safety shifts from a promise a company makes about itself to a condition a counterparty can act on.

As with anything, it's important to look at motives behind the trigger. Is there a situation where it could be in SpaceX's best interests to enforce the clause, even if it means letting go of a revenue stream? Private contracts can create operational consequences for unsafe behaviour, but they are a poor substitute for independent governance when the standard is vague and the enforcer is also a competitor.