Davos this year produced a lot of predictions about AI. One of them was about energy.
Nvidia’s CEO Jensen Huang said four words that reframe everything. “AI is infrastructure. Just like electricity. Just like roads.”
At the Morgan Stanley Technology, Media & Telecom Conference, weeks later, he went further. He said software is not running on neutral hardware. It is a five-layer cake: energy at the base, then chips and computing infrastructure, then cloud data centers, then AI models, and finally the applications built on top of them. Every layer depends on the one below it. Pull out any layer and the whole stack fails.
“AI is a factory because factories are power-limited. It doesn’t matter how many plants you have. Each plant is still 100 megawatts or a gigawatt, and therefore tokens per watt is the single most important thing for the top line of companies. CEOs have to make those decisions very, very carefully.”
That framing deserves to sit with you for a moment.
According to the IEA’s Electricity 2026 report, global electricity demand is forecast to grow at 3.6% annually through 2030, adding roughly 1,100 terawatt-hours each year. That is 50% more than the average annual increase recorded over the previous decade. The grid is being asked to absorb growth it was never designed to handle at this pace.In the US, data center grid power demand is projected to reach 75.8 gigawatts this year alone, and nearly double to 134 gigawatts by 2030. The median wait time to commercial operation now approaches five years. From 2000 to 2024, only one in eight projects that applied to connect to the US grid actually made it to commercial operation.
Cheap solar and batteries could fix that problem. But the grid isn’t ready.
The grid was designed around a simple assumption: large, centralised generators push electrons outward in one direction to millions of passive recipients. That assumption made sense for most of the twentieth century.
It does not describe the grid of today, and it is deeply inadequate for the grid of the next decade.
Households and buildings now generate power. Batteries store it and release it on schedules of their own. Electric vehicles add millions of new load points that move around. AI infrastructure adds load so concentrated and constant that it reshapes regional market dynamics by itself.
The electrons are no longer moving in one direction. The economics are no longer simple. But the coordination infrastructure, the market mechanisms, the settlement logic, and the visibility tools are still largely built for the world that existed before solar was cheap.
That mismatch is where most of the waste lives. Global renewable curtailment volumes rose roughly 55% in 2024 meaning clean electricity that was generated simply went unused because the grid had nowhere to put it. In California alone, over 2,700 gigawatt-hours of wind and solar were curtailed in just the first five months of 2025, discarded because the coordination layer wasn’t there to route it.
The same pattern is appearing across several countries such as China, Germany, Brazil and the UK with the number of hours recording negative electricity prices surging during peak solar generation. Batteries dispatch according to price signals that don’t necessarily help the grid. Communities with surplus rooftop solar have nowhere to send it beyond the feed-in tariff, which in most markets has been steadily cut. The energy is real. The waste is real. The coordination layer was never built.
Local energy as the answer, and the deployment gap that keeps blocking it
The logic of local energy markets is straightforward. If the issue is matching generation with demand in real time across a grid that was never designed for bidirectional, distributed flows, then push the matching closer to where the electrons are. Energy communities. Microgrids. Embedded networks. Coordination at the local feeder.
The value case is well understood. Less congestion on transmission infrastructure. Better utilisation of distributed assets. Lower prices for community members because value stays local rather than flowing to a retailer’s margin. Price signals for new generation exactly where it’s needed. More resilience when the centralised supply is under stress.
Utilities understand this. Regulators are moving toward it. But between understanding and deployment sits a barrier that has quietly killed more pilots than any technical limitation ever has.
Enterprise-grade P2P energy trading carries substantial infrastructure prerequisite: real-time smart meter connectivity, bidirectional communication systems, participant-facing apps, billing integrations, automated settlement, and dedicated operational teams. Each of those makes sense once a project has proven its economics and earned the budget for a full build. At the pilot stage, they represent a commitment that hasn’t been justified yet. You need the infrastructure to prove the value, but you can’t justify the infrastructure until you’ve proven the value.
What Transactive Lite is actually for
Powerledger’s Transactive Lite is built to break that loop. Not to replace a full P2P trading platform, but to be the layer that comes before one is warranted.
It works with existing meter data. Projects don’t need new smart meter infrastructure, real-time connectivity, or participant apps built before they can run a single trade. Feed in standardised interval data and the platform applies market-based matching algorithms, generates carbon accounting at the meter level, records every transaction on the Solana blockchain, and produces client-ready reporting: trading summaries, energy statements, and emissions reduction documentation suitable for regulatory submissions and ESG reporting.
It uses the same trading logic that powers Powerledger’s transactive platform, extracted into a batch-based model that can be live in days rather than months.
An energy community can use it to coordinate local solar and battery sharing without upfront IT investment. A university or hospital can use it to demonstrate verifiable sustainability outcomes before committing to a full platform build. A regulatory sandbox can use it to generate the evidence base that policy reform actually requires. An EV charging network can use it to prove the economics of matching charging demand against local renewable supply before scaling the infrastructure to support it.
The Solana audit trail is worth dwelling on. Every transaction recorded, every matched trade transparent and verifiable in a way that can be independently confirmed. That is what turns a pilot result into regulatory-grade evidence: the difference between a presentation and a proof that can actually move a policy conversation forward.
Why this moment matters
The grid will not be rebuilt from the top down fast enough to absorb what is coming at it. The interconnection queue numbers, the data center demand projections, the capacity market prices: they all point to the same conclusion. The system is already behind the demand curve, and the gap is widening.
The pressure has to be relieved from below. Distributed generation, local storage, local markets — none of these are alternatives to the central grid. They are complements that reduce the load on it, add flexibility to it, and keep value closer to where it is created. But that only works if the market infrastructure can actually be deployed, and it can only be deployed if the complexity of infrastructure needed for the early proof-of-concept stage stops killing projects before they mature.
The constraint on the AI economy is the grid. The constraint on the grid is market infrastructure that can coordinate distributed assets at scale. The constraint on building that infrastructure has been the deployment gap at the proof stage, the place where good projects stall before they can prove anything.
That is the specific problem Transactive Lite addresses. A path from concept to deployment, with the evidence that earns the next step.
Powerledger’s Transactive Lite is available now for pilot programs, regulatory sandboxes, and early-stage commercial deployments. Get in touch →
