THE BUILDOUT
Intelligence is becoming infrastructure.
AI is a demand shock propagating through the physical economy.
Every increment of machine intelligence creates a bill that is settled downstream, in physical capacity: the semiconductors and compute systems that run it; the power generation, transmission, and cooling that feed it; the data centers, and the land, water, and grid interconnects beneath them; the robots and autonomous systems through which intelligence acquires hands; the defense and space platforms being rebuilt around autonomy; the advanced manufacturing that builds all of the above; and the market infrastructure that clears an increasingly machine-speed economy.
These sectors form one supply chain, viewed at different depths. Demand originates in intelligence and cascades down the stack, and capital, talent, and pricing power are migrating to the layers where capacity is scarcest.
The companies at those choke points share three traits. They are disproportionately private, and many remain private through important stages of their growth. They are disproportionately oversubscribed, because the same observation is visible to everyone. And they are disproportionately closed: their cap tables are managed, their rounds are allocated among insiders, and their shares change hands only through negotiated transfers.
Three further shifts make this moment unusual. AI demand is drawing a growing share of private capital toward physical-layer companies. A decade of extended private tenure has built a structural overhang of employee and early-investor positions that need liquidity — sellers whose timelines are independent of company quality. And the repricing of late-stage entry after rate normalization has punished undisciplined buyers while rewarding negotiated entries. Access plus pricing discipline is the trade this moment rewards.
Which is why the binding constraints in this market are access and structure: getting into the room, and building a form of ownership that actually clears — through rights of first refusal, board consents, transfer windows, and the other machinery that stops most buyers. Those two constraints are the firm. Carbyne exists to earn positions in the buildout’s scarcest companies through direct relationships.
THE CHAIN
How demand moves down the stack.
01 · AI & DEEP TECH
Where demand originates.
Frontier models and applications set the demand for everything beneath them. The chips, interconnect and inference that serve them feel it first, and AI-native science carries it into biology and the brain.
models · chips · inference · AI-native science
02 · ENERGY
The binding constraint.
Advanced nuclear and its fuel supply, generation, storage and transmission. Compute scales only as fast as the power behind it, and power is the slowest layer to expand.
nuclear · fuel · generation · grid
03 · ROBOTICS
Where intelligence acquires hands.
Robot foundation models, autonomous vehicles and drones. Intelligence leaves the data center and enters the physical world.
robotics · autonomy · drones
04 · SPACE & DEFENSE
Where the buildout leaves the planet.
Launch, commercial stations, in-space manufacturing and orbital compute, where power and cooling are abundant. Alongside it, defense rebuilt around unmanned and autonomous systems.
launch · stations · orbital compute · defense
05 · MARKET INFRASTRUCTURE · ADJACENT
Clearing at machine speed.
Exchanges, custody and settlement for an economy that moves faster every year. An adjacent layer, where we invest selectively.
settlement · custody · exchanges
THE CHOKE POINTS
Scarcity forms where capacity is slowest to grow.
Capital, talent and pricing power move toward those positions, from the frontier labs to the launch pads. Carbyne looks for position there, one company at a time.
WHERE WE LOOK
Four core sectors.
AI & Deep Tech
Frontier models, agents and applications, and the semiconductors, optical interconnect, data platforms and inference capacity that serve them. It extends into AI-native science: drug discovery, protein design and brain-computer interfaces, where models compress years of laboratory work.
Space & Defense
The buildout leaves the planet: launch, commercial stations, in-space manufacturing and orbital compute, where power and cooling are abundant. Alongside it, an industrial base for defense rebuilt around unmanned and autonomous systems.
Energy
Intelligence is an energy problem. Advanced nuclear and its fuel supply, generation, storage and transmission are becoming the binding constraint on compute, and the slowest layer to expand.
Robotics
Intelligence acquiring hands: robot foundation models, autonomous vehicles and drones, and the advanced manufacturing that builds the rest of the buildout.
ADJACENT Market infrastructure: exchanges, custody and settlement for a machine-speed economy, pursued selectively. Communications infrastructure on the same terms.
ALSO
The same work, from other angles.
Company-approved liquidity
When a company wants to offer structured liquidity to early employees or investors, Carbyne can take part with the company’s approval and within its transfer provisions. For companies & lead investors →
Special situations
Occasionally a real position comes with a complicated path to it: layered entities, unexercised options, a consent chain nobody has mapped. We take these on only when the company clears the same bar — and we price the complexity openly.
WHY DEDICATED VEHICLES
Built for the size of these rounds.
Many of the largest private rounds in this buildout exceed what any single fund holds on its own balance sheet. Lead investors and companies fill the difference with dedicated vehicles, one company per vehicle. Carbyne is built for that work: we source allocations alongside lead investors, underwrite each company on its own terms, and run each vehicle from close through exit with clear reporting.
The thesis
Why the AI buildout is one supply chain, and where scarcity forms along it.
Read the thesis →For companies & lead investors
How Carbyne works as a shareholder, and how we come into a round.
Companies & lead investors →