AI’s Power Problem Moves Off Earth
The hosts:
Patrick Tucci, Vsquared Ventures
Karol Szubstarski, OTB Ventures
What is this Ripple about?
As AI compute collides with terrestrial constraints — grid capacity, energy cost, cooling, water use, land availability and permitting — orbital data centres are moving from science fiction into serious infrastructure conversations. This Ripple asks what would need to hold true for space-based compute to become CREDIBLE climate infrastructure: a new energy layer for a bottlenecked world, with AI as the first use case.
Why this topic timely or urgent in 2026
AI’s energy demand is forcing the data-centre conversation into new territory: the IEA projects global data-centre electricity consumption to more than double to around 945 TWh by 2030, with AI as a major driver. At the same time, Google has launched 'Project Suncatcher', exploring solar-powered satellite constellations equipped with TPUs and free-space optical links, with prototype satellites planned for early 2027. The topic became even more timely with reports that Google and SpaceX are discussing orbital data-centre infrastructure. Whether SpaceX becomes the dominant enabler, a launch provider, or a competitor to startups, the signal is clear: hyperscalers, launch platforms and space-infrastructure companies are beginning to treat orbital compute as a potential part of the future AI and energy stack. 2026 is the right moment to debate whether earth-based solutions — cleaner grids, faster permitting, better cooling, nuclear, storage, efficiency and new chips — are enough, or whether entirely new infrastructure layers will be needed. It is also the right moment to ask what the attempt itself might unlock: Apollo was not only a moonshot that landed humans on the Moon; it also accelerated integrated circuits, with MIT working closely with Fairchild Semiconductor on the Apollo Guidance Computer, and space and defence procurement becoming an important early demand source for the semiconductor industry.
Key questions or challenges being explored
What would need to be true for orbital data centres to make technical, economic and climate sense?
Power generation, thermal management, radiation hardening, optical links, launch costs, servicing, latency, lifetime and de-orbiting all matter.
Who will push the ecosystem forward — hyperscalers, SpaceX, startups, defence customers or governments?
If orbital compute becomes real, the winning stack may be shaped as much by market power and launch access as by technical merit.
What could this moonshot unlock, even if orbital data centres are not the final outcome?
Apollo helped push the semiconductor frontier; a space-compute push could similarly accelerate advances in optical communications, radiation-hardened chips, thermal systems, autonomous operations, in-orbit servicing, space-based solar, in-space manufacturing and defence infrastructure.
The key tension or debate at the heart of this discussion
The central tension is fear of missing out vs. fear of being fooled. - Investors and operators can see why orbital compute might matter if AI power demand keeps exploding and terrestrial infrastructure remains slow, bureaucratic and capacity-constrained. But the category is still early enough that it could become either generational infrastructure, an expensive techno-optimist distraction — or something more subtle: a moonshot whose most valuable outputs may be the technologies, capabilities and strategic positions created along the way.
Who would get the most value from attending this Ripple?
This Ripple is for deep-tech VCs, climate infrastructure investors, space-tech founders, data-centre operators, semiconductor and photonics founders, energy/grid experts, defence and sovereign-compute stakeholders, and LPs interested in frontier infrastructure. It is especially relevant for people asking whether orbital infrastructure is an investable category, and where startups can still win if hyperscalers and launch giants move first.