Caspian Vey
I build companies the world will need before it knows it does.
Most of what I work on now would have looked unbuildable five years ago. That is, on reflection, the only kind of company worth starting.
I write here for the operators, investors, and would-be partners who'd rather skip the pitch deck. What follows is what I'm doing, what I believe, and who I'm looking for.
Building Greylight
Partner, Aperture Capital
Director, six boards
Twelve years in. Three companies built. One thesis I keep refining.
I started my first company at twenty-six because the alternative — five more years of consulting — felt like a slow erasure. It sold to a strategic in 2019 for more than I'd ever expected to see in a single number. The second built quietly through the pandemic and runs today with a CEO better suited to it than I am.
The third is what I'm doing now. It is harder than the others combined, and probably the only one I'll be proud of in twenty years. The rest of my time goes to people building things they'd struggle to do without a check, a board seat, or a hard conversation at the right moment.
$340M
$82M ARR
19
Companies I have founded, run, or sold.
I include the one that did not work. Anyone who tells you their record without it is editing the truth.
Greylight
Software for the operators who decide, in real time, where electricity goes. We started in stealth in 2024 with a former utility CTO and two physicists. We are presently deployed at one of the larger US ISOs and another in Iberia. This is the company I expect to spend the next ten years on.
Linnaeus
A B2B platform for life-sciences contract research. Built quietly through the pandemic. Reached $80M ARR in 2024 without a single growth-marketing hire. I handed it to Eleanor at the end of last year because she is materially better at running it than I was, and the company deserved that.
Stratum
Logistics software for mid-market freight forwarders. Bootstrapped to $11M ARR, then raised one institutional round to underwrite the expansion into the EU. Sold in 2019 for $340M to a strategic that has, to its credit, continued investing in the product. The team and I were locked out for two years. Many of them are still there.
Halten
Consumer hardware in a category that, in retrospect, was incompatible with consumer behavior in 2013. We sold 4,200 units, learned a great deal about supply chains, and returned 62¢ on the dollar. I would not undo the decision to try, but I would compress the time-to-decision from eighteen months to six. Halten is the reason I now prefer companies whose customers are paid to think about the problem.
The next decade will reward those who pick problems by their physics, not their margins.
Written in March 2026. This is my working version. I will keep refining it.
For most of the last fifteen years, the dominant story of value creation was that you could pick almost any consumer behavior, route it through a phone, and capture an exceptional margin in the middle. I built two companies in that era. I am grateful for both. I do not believe the rest of my life should look like more of them.
The companies that compounded fastest in 2020 were companies of distribution. The companies that will compound fastest in 2035 will be companies that resolve a constraint the physical world is presently failing to resolve on its own — power, materials, biology, computation, atoms moving through space. Each of these is a market with thinner margins on day one and structurally better margins by year ten, because the alternative is unworkable rather than merely worse.
Pick the problem by its physics. The margins will follow.
I am not, to be clear, against software. Greylight is software. So is most of what I back. What I am against is software whose only defensibility is a network effect that, on inspection, is a habit. Habits break. Constraints don't.
The most useful question I ask now, in any room, is: what does the world physically need more of in 2040, and what is presently impossible about supplying it? The answers are not subtle. They are unevenly distributed across very few people willing to spend a decade on them. Those are the people I want to be in business with.
Companies I have backed, joined the board of, or quietly helped over years.
Through Aperture, my personal account, and board service. Selected by the founder's tolerance for hard conversations.
Coronado Nuclear
Small-modular reactor company, Series B. I joined the board in 2024 because their COO is the best operator I have ever sat across a table from. The siting work alone is a decade of value.
Marrow Diagnostics
Single-cell sequencing at the price of a chest X-ray. The founders walked me through the chemistry on a flight. By landing I had wired their first check and asked them to let me join the seed round at any size.
Heliodyne
A photonic computing startup whose CEO I met at a wedding. She asked me to advise on commercial structure; I asked her to teach me physics. Both of those exchanges remain ongoing.
Foundry Atelier
A small specialty alloys company outside Pittsburgh that supplies aerospace. Not glamorous. Profitable from year two. I'm chairing the board through their next round of capital and a clean succession.
Argosy
Inference scheduling for regulated industries — health, energy, defense. Two founders who refused to take VC money for the first eighteen months and are now glad they did the regulatory work first.
The Coastland School
A nine-year-old independent secondary school in northern Portugal. The work that matters least to my résumé and most to my conscience. I serve as a trustee, no fee, no equity, just letters and the occasional flight.
Long-form interviews, not soundbites.
Selected. I decline more interviews than I accept. The ones I take are with people I would have talked to anyway.
On building during a structural change
What I learned giving away the CEO seat
The unfashionable companies
Three founders, three exits, one regret
Caspian is the person I phone before any board meeting where the answer is not obvious. He will not flatter you, he will not theorize at you, and he will not tell you what your team already told you. He will reframe the question in a way that makes the choice easier to live with — whichever way you choose.
Mara Solberg
I have written checks alongside Caspian for eight years. He is the rare investor whose returns are explainable: he picks people, then he stays away unless asked, and when he is asked, he is exceptionally useful. He is, in a category that mostly performs being helpful, actually helpful.
Idris Whitlock
Not employees. Not founders to back, exclusively. Partners.
A co-founder, late.
Greylight will need one. Someone with deep experience selling to grid operators or utilities — ideally an operator from inside that world who wants to build the tool they always wished they had. Equal partner. Real equity. Direct line to me.
A second GP for Aperture.
An operator-investor who has built and sold at least once, who can hold a thesis without overfitting, and who wants to underwrite the next decade of physical-world companies. We are raising Fund II in 2027. The bench is small on purpose.
Companies to back, early.
Founders with deep domain conviction in energy, biology, materials, or compute. I write between $250k and $2M at pre-seed and seed. I respond to introductions in under a week. If you're cold-emailing, lead with the constraint you're trying to break.
If you've read this far, you already have the address.
Plain text, please. Lead with the constraint you are trying to break, the people in the room, and the worst version of the question you would ask me on a phone call. I read everything that reaches me. I reply, on average, within three days — sometimes shorter than you'd like, but always honest.