The constraint has moved.

The venture model that worked from roughly 2010 onward was built around a single insight: software had near-zero marginal cost, so whoever reached customers first and cheapest compounded fastest. Capital went into go-to-market because go-to-market was the constraint.

That advantage has largely been competed away. Building and shipping software is now cheap enough that a product lead no longer lasts long, and the channels that once produced cheap growth are saturated and priced accordingly.

What is genuinely scarce now sits upstream: firm power, interconnection capacity, advanced packaging, settlement throughput, qualified production lines, and security clearances. None of them respond to a growth budget.

A data center is a power contract with a building attached. A defense program is a procurement path with software attached. A tokenised fund is a transfer agent with a ledger attached. In each case the binding constraint is an institution or a piece of physical infrastructure, and the timeline is measured in years rather than sprints.

This is unglamorous work, and it is where the durable margin is accumulating. Companies that secure a scarce physical input early hold an advantage that a better-funded competitor cannot simply buy its way past.

Most venture firms are not built to underwrite this. The diligence required is closer to project finance and engineering review than to analysing a growth funnel: reading an interconnection queue position, understanding a power purchase agreement, assessing whether a program of record will actually open, or judging whether a custody architecture survives a regulator's examination.

Firms that cannot do that work tend to either avoid the category or price it on narrative. Both create opportunity for investors willing to do the specific, unscalable work of verifying the claim.

We do not run five sector practices. We hold one view — that supply-side constraints now determine outcomes — and the five areas are where that view is currently expressed.

The connections are not theoretical. An AI infrastructure company's gating item is a power contract. A defense autonomy company's gating item is software talent and a procurement path. A private-markets platform's gating item is the operational reality of fund accounting. We invest across all five precisely so that we see both sides of those dependencies before we underwrite either one.