TechCrunch published a contrarian investment thesis this month: the best AI investment in 2026 might not be an AI company at all, but an energy technology company. The argument is not complicated. AI's explosive demand for compute has turned data centres into the fastest-growing load on electrical grids worldwide. Goldman Sachs projects data centre power consumption will rise 175% by 2030. That demand is driving capital into grid infrastructure, battery storage, and renewable generation at a scale and speed that the energy transition alone was not producing. The companies that benefit most from this buildout — grid software firms, storage manufacturers, transmission developers — are currently trading at a fraction of the valuation of the AI companies they serve.
The conventional framing of AI's energy footprint is that it is a cost — to the climate, to ratepayers, to the credibility of tech companies' sustainability commitments. That framing is accurate as far as it goes: AI data centres do use enormous amounts of electricity, much of it currently supplied by fossil fuels, and the political backlash explored in previous coverage is real. But CNBC's January reporting on the energy transition offers a different frame: fossil fuels are a "crutch" that the AI buildout is actually hastening the abandonment of, because the scale of demand AI creates is precisely what makes large renewable projects — utility-scale solar, offshore wind, long-duration storage — economically viable to build. The AI boom may be accelerating the energy transition even as it strains the grid.
The mechanism is a self-reinforcing cycle that energy analysts have started to map explicitly. Massive, predictable data centre loads make renewable energy projects financeable — lenders will fund a solar farm if a hyperscaler has signed a 15-year power purchase agreement for its output. Those renewable projects drive down the cost of clean electricity. Cheaper clean electricity makes electrification of other sectors (transport, heating, industry) more economical. More electrification increases the value of grid intelligence and storage. AI companies, needing more renewable power for their own sustainability commitments, fund more renewable development. Each step accelerates the next.
Scientific American's analysis of AI's net climate impact reaches a similar conclusion but with important caveats. The modelling suggests AI systems could cut global climate pollution by up to 5.4 billion metric tons annually over the next decade — primarily through grid optimisation, industrial efficiency, and accelerated materials discovery for clean energy technology. That potential reduction exceeds the expected increase in emissions from running AI data centres. But the word "could" is doing significant work: realising those gains requires that AI actually be deployed in climate applications at scale, which requires deliberate policy choices and investment decisions that are not automatic consequences of the AI buildout.
NVIDIA's partnership with Emerald AI on grid-flexible data centres points at one concrete mechanism. The idea is that data centres, which have large controllable loads and often on-site generation capacity, can operate as grid assets — curtailing consumption when grid stress is high, increasing it when renewable supply exceeds demand. A data centre that participates in grid balancing markets is not just a consumer of electricity; it is part of the grid infrastructure that makes variable renewable energy dispatchable. If data centres at scale adopted this model, they could meaningfully improve renewable integration without requiring new transmission or storage investment. That is an interesting idea that is still mostly theoretical at the deployment scale needed to matter.
The investment implication that TechCrunch's piece draws out is worth taking seriously even if the climate optimism requires qualification. The energy technology companies that are building the infrastructure AI demands — grid software, battery storage, interconnection capacity — are benefiting from AI spending without carrying AI valuation multiples. They are, in the language of investing, picks-and-shovels plays on a gold rush. Whether AI's net effect on climate is ultimately positive or negative depends on choices that are still being made. What is less uncertain is that the energy transition is being accelerated, at least in part, by the same demand curve that makes it look like a problem.