Five areas, one thesis.

Compute is the scarce input. We fund the layers that decide whether it gets used well.

We are less interested in model development than in the infrastructure and applications around it: serving and inference economics, data pipelines for industries that cannot send records to a public endpoint, evaluation, and systems that remove a line item from a company's budget rather than adding a tool to its stack. We underwrite cost curves and switching costs, not benchmark scores.

Settlement is being rebuilt. We fund the plumbing, not the tokens.

Our interest is in what institutions need in place before they move size on-chain: custody and key management, clearing, collateral mobility, and transfer agency for assets that are currently tracked in spreadsheets. We do not underwrite token price. We underwrite fee-generating infrastructure with identifiable counterparties and a regulatory path that someone has actually walked.

Private markets still run on spreadsheets and PDFs.

Capital moved private faster than the software supporting it. Fund accounting, capital calls, valuation, waterfall math, and LP reporting are still substantially manual at firms managing billions of dollars. We back the operating systems for that work, and the payments and credit infrastructure underneath it, where the underwriting is real and the data is proprietary by construction.

A data center is a power contract with a building attached.

Interconnection queues, not chip supply, increasingly set the pace of the compute buildout. We invest in firm generation, storage, and the software that moves flexible load to where power already exists — including behind-the-meter generation for large industrial consumers who cannot wait years for a utility to connect them.

Procurement is changing faster than the incumbents are.

Programs of record are opening to companies that ship on commercial timelines. We back dual-use systems with a credible path through acquisition — autonomy, sensing, resilient communications, electronic warfare — and domestic production of components that currently have a single foreign source. We expect to underwrite contracting risk directly, not around it.