Where we spend our attention.
Chosen because they compound into each other. Intelligence needs infrastructure; infrastructure needs securing; security needs autonomy; all of it eventually needs to be paid for. We would rather hold one view of that whole picture than seven unrelated ones.
Models, tooling and the applied systems that put them to work inside real operations.
The interesting question is no longer whether a model can do the task, but what has to change around it — data, evaluation, cost, accountability — before an organisation will let it. We look for teams working on that edge, in both horizontal infrastructure and deep vertical applications.
Identity, data and infrastructure protected against an attack surface that keeps widening.
Every technology we back elsewhere becomes something for somebody to attack. We invest across identity, cloud and data security, detection and response, and the emerging problem of securing autonomous systems and the agents now acting on their behalf.
Autonomy and machines that do physical work in unstructured environments.
Labour shortages, reshoring and cheaper perception are moving robotics out of the fenced cell. We look for hardware disciplined enough to be manufactured and software good enough to make the hardware ordinary.
Compute, networks and storage — and the power and platforms underneath them.
The constraint on the current technology cycle is physical: energy, land, silicon supply and the interconnect between them. We invest in the layers that make capacity usable, and in the companies that make existing capacity go further.
Systems of record and workflow that become hard to remove once installed.
We favour software that owns a process end to end rather than decorating one, in markets where the incumbent is an incumbent by default rather than by merit.
Payments, markets and the plumbing that moves and settles money.
Infrastructure over interfaces: rails, risk, compliance and settlement, including the places where regulated finance and new asset types are being made to work together.
Fields still forming, where the first credible teams set the direction.
A deliberate open category. Some of the best positions come from areas nobody had a thesis on eighteen months earlier, and we would rather keep room for them than pretend the list is finished.
What makes something a fit.
A hard technical core
Something that took real work to build and would take real work to copy. We are comfortable with problems that stay difficult after the first version ships.
A buyer with a reason to move
Regulation, threat, cost or capacity — a forcing function that means the decision gets made this year rather than eventually.
Founders who intend to stay
Our horizon only works if it is shared. We look for teams building a company rather than a position.
Room for us to be useful
A partnership to open, a licence to structure, an introduction that changes the year. If capital is all that is needed, we are probably not the right holder.