E-Paper
Sign In
945 TWh by 2030: Why Lado Okhotnikov's space solar idea persists
With data centre demand set to double by 2030, the feature assesses the engineering, economic, and governance realities behind the idea
- PUBLISHED: Wed 25 Mar 2026, 3:56 PM
In December, an idea that had long circulated quietly within academic and aerospace research briefly surfaced in the public domain. Lado Okhotnikov, founder of Holiverse, suggested that the long-term energy demands of artificial intelligence may eventually exceed what Earth-based systems can reliably provide. Among the possibilities he referenced was the concept of harvesting solar energy in space and transmitting it to Earth to support AI infrastructure. He also noted that his team is exploring this direction.
The statement was not presented as a proposal or a prediction. There was no suggestion of near-term feasibility or deployment. Instead, it functioned as a signal — a way of illustrating how sharply the conversation around AI and energy is evolving. It raises a central question: if serious technologists are willing to discuss space-based energy at all, does that reflect visionary foresight, or does it reveal how constrained existing systems may become?
Recommended For You
When earth’s energy systems begin to strain
AI infrastructure places demands on power systems unlike most industrial loads. Data centres operate continuously, often at high and variable intensity, with little tolerance for disruption. In 2024, they consumed around 415 terawatt-hours, or roughly 1.5 per cent of global electricity demand. The International Energy Agency projects this could rise to approximately 945 TWh by 2030, driven largely by AI and expanding digital services, making data centres one of the fastest-growing sources of new power demand in advanced economies.
The trend is especially visible in the United States, where data centres accounted for about 4 per cent of national electricity consumption in 2023, a share expected to reach around 9 per cent by the end of the decade. Individual hyperscale facilities can draw as much electricity as tens of thousands of homes, while AI-optimised servers — far more energy-intensive than conventional hardware could account for around 44 per cent of total data-centre power consumption by 2030, with electricity use projected to rise nearly fivefold between 2025 and 2030.
Renewables already supply around 27 per cent of data-centre electricity, but intermittency, transmission constraints and grid congestion mean supply is not always available where and when demand is rising fastest.
Energy from orbit: Straightforward in theory
The concept of space-based solar power is quite simple to understand. Solar panels deployed in orbit would receive continual sunlight, regardless of night cycles or atmospheric conditions. The gathered energy would then be wirelessly delivered, most typically via microwave or laser systems, to receiving stations on Earth, where it would be transformed into electricity and fed into the grid.
The underlying science is well established. Wireless power transmission has been demonstrated at limited scales, and space agencies have explored the concept for decades. In principle, it is feasible.
In practice, however, the scale required to meaningfully support AI infrastructure introduces a different level of complexity. Supplying even a fraction of global data-centre demand would require orbital systems far larger and more sophisticated than anything currently deployed.
The cost of building above the planet
Generating power in orbit would demand infrastructure well beyond today’s capabilities. Solar arrays would need to be manufactured, launched and assembled at unprecedented scale. Autonomous or semi-autonomous construction in space would most likely be required, bringing considerable technical risk.
Maintenance is another difficulty. Space conditions are harsh: radiation, micrometeoroids, and material deterioration all have an impact on system longevity. Repairs are not simple nor cheap, even as launch costs fall.
On Earth, receiving stations would also require a huge investment. Integrating huge, continuous energy flows into existing grids requires changes to transmission, balancing, and safety systems. When considered together, these criteria underline the economic uncertainty that surrounds any large-scale deployment.
Reliability and new dependencies
AI infrastructure is extremely vulnerable to interruption. Training big models and executing real-time inference necessitate persistent uptime, which places high demands on energy reliability.
Introducing orbital energy into the supply mix would create new dependencies. Power delivery would rely not only on terrestrial infrastructure but also on systems operating far beyond Earth’s surface, exposed to space weather, orbital debris and technical failures that are difficult to service quickly.
Terrestrial grids already face reliability pressures from extreme weather and ageing infrastructure, particularly in regions with rapid data-centre expansion. Whether an energy system that includes space-based components could match or exceed the resilience of ground-based generation without extensive redundancy remains unresolved.
Governance beyond earth
Beyond engineering and economics, space-based energy raises unresolved governance questions. Who owns orbital power infrastructure? Who regulates it? How is access prioritised if AI becomes a dominant consumer?
Unlike national grids, which operate within established regulatory frameworks, orbital systems would likely require new international agreements. Issues of safety, coordination, competition and dispute resolution would need to be addressed before such infrastructure could function at scale.
At present, these governance structures do not exist.
A thought experiment with weight
There is no roadmap today for powering AI from space. No deployment timeline. No certainty that the economics will ever align. Space-based solar energy remains a theoretical, long-term research direction.
Yet its appearance in public discourse is significant. As AI-driven electricity demand continues to rise potentially doubling by the end of the decade — technology leaders and energy planners are being forced to think beyond incremental improvements. More efficient hardware, advanced cooling, improved grid integration and better demand management will all play essential roles. Whether they will be sufficient on their own is less clear.
In this context, the idea of harvesting solar energy in space does not function as a prediction. It highlights the growing tension between AI’s trajectory and the pace at which Earth-bound energy infrastructure can adapt.
Whether AI data centres will ever be supported by energy collected in orbit remains uncertain. The technical, economic, regulatory and reliability challenges are substantial. What is clear is that space-based solar power for AI is theoretical and long-term — and that the idea entered broader conversation through figures such as Lado Okhotnikov, who has said his team is exploring this direction.
The outcome is not certainty, but suspension: an invitation to reassess the limits of current systems and to consider how far beyond them the future of AI may eventually reach.