By James Aspinwall, co-written by Alfred Pennyworth (my trusted AI) — March 4, 2026, 09:55
Fraser Anderson, managing director at Link Ventures, sat down with Alex Wissner-Gross at Davos on January 21, 2026. The conversation covered GPU geopolitics, space-based compute, AI futures markets, and what startups should be building right now. Below is a breakdown of the key ideas and why they matter.
Who Is Alex Wissner-Gross?
Alex is a physicist, inventor, and investor who builds agentic AI systems. He’s the kind of person who normally lives inside his digital layer in Massachusetts and rarely leaves. The fact that he traveled to Davos — through Paris, Germany, Austria, and Liechtenstein — signals that the conversations happening at the intersection of AI policy and geopolitics are too important to have remotely.
His highlight from Davos: a conversation with US Under Secretary of State Sarah Rogers about GPU diplomacy — the idea that compute distribution across the Earth’s surface is becoming a foreign policy issue on par with energy and trade.
GPU Diplomacy: Compute as Geopolitical Leverage
Fraser frames the context well. Over the past 18 months, realpolitik has reasserted itself in global business. The physical layer of intelligence — GPUs, data centers, power infrastructure — determines which countries develop fastest, which maintain strategic trade advantages, and which get left behind.
Alex acknowledges two clichés and then moves past them:
- “GPUs are the new oil” — compute is to the 21st century what petroleum was to the 20th.
- “The world is dividing into spheres of influence with independent tech stacks” — US, China, EU, and others are building parallel AI ecosystems.
These are now conventional wisdom. What comes next is more interesting.
The Dyson Swarm Is Coming
Alex’s core thesis: data centers are going to space, and this is now a foregone conclusion.
A Dyson swarm — a constellation of solar-energy-collecting structures orbiting a star — sounds like far-future science fiction. But the logic is straightforward:
- Earth has limited power, limited cooling, and political boundaries.
- Space has unlimited solar energy, natural cooling (radiative dissipation), and no zoning boards.
- The explosion of intelligence infrastructure needs somewhere to go. Up is the answer.
The geopolitical implications shift dramatically. Alex points to a friend running one of the first lunar data center companies. On the Moon, there’s no established defense perimeter. No sovereignty frameworks. No territorial law enforcement. It’s the Wild West.
He references the Brad Pitt film Ad Astra, which depicts commercialization hitting the Moon in violent, chaotic ways. That’s not metaphor — it’s trajectory.
What this means: The current scramble for GPU access and data center real estate is a preview of a much larger scramble for orbital and lunar compute territory. Countries and companies that think only in terms of terrestrial infrastructure are thinking one dimension short.
Compute Futures: Let the Market Decide
Fraser asks whether the GPU market will see the classic supply-squeeze-to-glut cycle. Nvidia’s market cap has grown enormously. Google is making progress with custom silicon. Will supply catch up?
Alex sidesteps with a meta-answer. He has a financial interest in a company called ORE that builds futures markets for compute. His argument: don’t ask any single person to predict GPU prices. Instead, build a market and let the collective intelligence of participants with insider knowledge establish prices through futures contracts and options.
This is now reality — as noted in the March 3 briefing, ORE and Cali have launched the first CFTC-regulated H100 price contracts. Compute is officially a traded commodity, like oil or wheat.
Alex’s personal position: long-term bull on the intelligence explosion. He wants intelligence too cheap to meter. Whether that’s good or bad for any single player like Nvidia, he’s agnostic. The goal is to “tile the solar system with compute.”
What Startups Should Build Now
Fraser asks for calls to action. Alex’s answer is precise and contrarian:
Not enthusiastic about: SaaS for [fill in the blank]. The traditional software-as-a-service model is being eaten by AI agents that can replicate most SaaS functionality on demand. Building another dashboard or workflow tool is a shrinking opportunity.
Very enthusiastic about (potential multi-trillion dollar companies):
- AI to redesign coastlines and create new islands — climate adaptation and land creation as engineering problems.
- AI for solving physics — using AI not just to simulate but to discover new physical laws and materials.
- AI for transforming physical service economy labor — reshaping how service-based work categories function.
His rule of thumb: “If you saw it in a science fiction novel and it hasn’t been built yet, now is the time to build it.”
The window is simultaneously closing (AI capabilities are advancing so fast that today’s breakthrough is tomorrow’s commodity) and wider than ever (the tools to build science-fiction-grade projects are now accessible to small teams).
The Sci-Fi Reading List
Alex recommends three novels that map the pre-, trans-, and post-singularity experience:
-
Accelerando by Charles Stross — A single family going through the singularity. Economics, posthumanism, and the acceleration of change told across three generations.
-
Rainbows End by Vernor Vinge — The pre-singularity era as experienced by ordinary people. Augmented reality, biotech, and the social disruption that precedes the main event.
-
Diaspora by Greg Egan — Post-singularity existence as seen by post-human entities. What happens after biological constraints are left behind.
Alex’s timeline: 10 years to Star Trek-level future, maximum. He’s not speaking metaphorically.
The Takeaway
This conversation captures a specific moment in early 2026 when the abstractions became concrete:
- GPU diplomacy is now an actual State Department function.
- Compute futures are now traded on regulated exchanges.
- Lunar data centers are real companies with real funding.
- SaaS is dying; science fiction is the new business plan.
The thread connecting all of it: intelligence infrastructure is becoming the primary substrate of civilization, and the competition to control it is moving from boardrooms to orbit.
Build the sci-fi future. The clock is running.
Disclosure: Alex Wissner-Gross disclosed financial interests in ORE, PSI, and 021T.
James Aspinwall is the developer of WorkingAgents, an AI consulting firm specializing in agent integration and access control for medium-size companies.