PV for AI, AI for PV – Part 1 of 4
On Monday 14 September, the global PV community gathers in Rotterdam for EU PVSEC 2026. Its scientific opening plenary will not be about new cell efficiency records but rather focused on PV in the Energy System. One of the three opening talks will feature Jef Poortmans of imec and is titled “AI for PV and PV for AI: Linking Nanoscale to TeraWatt(h)”. In a series of posts, we, at Awendio Solaris, propose to review the timely connection between solar innovation and the infrastructure powering artificial intelligence.
That relationship deserves our attention. Solar can help power AI, while AI can help us discover better solar materials, improve manufacturing and operate energy systems more effectively.
The Electricity Challenge
The IEA’s updated outlook projects global data-centre electricity consumption doubling from 485 TWh in 2025 to around 950 TWh in 2030, with consumption at AI-focused facilities tripling in that period.
Electricity attracts enormous attention because a 1 GW campus is an extraordinary physical load on the grid and a cost just under $600mio per year. A gigawatt-scale data centre costs somewhere around $50 billion to build, and roughly $35 billion of that is chips. Given the expensive and rapidly depreciating nature of GPUs, GPUs dominate the owner's annual economics, contributing over 60% of the total annual costs. Electricity is a small line item by comparison, commonly cited at 5 to 10 percent of total cost. The counterintuitive conclusion is that AI data centres are far more capital-intensive than energy-intensive economically. So, when people say AI is power-hungry, the problem they are describing is not really a cost problem.
It is a speed problem. The constraint is how quickly a developer can get connected and energised. In the United States, the wait to connect a new generation project to the grid has stretched from roughly 15 months two decades ago to something approaching 45 months. Load-side queues are lengthening too. ERCOT, the Texas grid and one of the fastest-moving in North America, peaks at around 80 GW and is sitting on interconnection requests representing well over 200 GW of new load. Most of those requests are speculative and will never be built, but they still clog the process for the ones that are real.
The honest diagnosis is that the shortage is not in generation. Grid connections, permitting and electricity equipment are key factors constraining expansion. Access to reliable power increasingly determines how quickly new computing capacity can become productive.
Constrained by slow grid connections, data centre developers in the United States are pushing forward projects with onsite natural gas-based power generation which has caused a crunch for gas turbines which are effectively sold out for years. Nuclear power plants and SMRs are getting a lot of interest but their implementation is measured in decades while their cost will be very high.
The Solar Opportunity
For solar, this creates an opportunity to compete on speed to power as well as cost.
PV’s modularity, short construction lead times and competitive economics make it a strong candidate to supply this growth, alongside storage and the wider electricity system. Solar and storage do not fix the grid. What they do is offer the one thing nobody else is offering right now: a project you can energise in about twelve months.
The scale is becoming tangible. In Abu Dhabi, Masdar and EWEC are developing 5.2 GW of solar with 19 GWh of batteries, designed to deliver 1 GW around the clock. The project reached financial close in July and is expected to operate in 2027.
Geography still matters. Batteries can shift daytime generation into the evening; prolonged low sunlight and winter conditions require a broader supply strategy. Data centres need dependable electricity every hour.
The greatest opportunity lies in designing solar, storage and data-centre demand together from the outset. That means choosing locations with suitable energy resources and infrastructure, securing long-term supply agreements, and delivering reliable power on a credible timetable.
For the solar industry, AI creates a demanding new customer… and a reason to accelerate innovation across the entire PV value chain.
How much faster could we scale AI if energy availability shaped data-centre location from day one?
Next: how AI can accelerate PV materials and cell research.
