Software Could Save the Grid, and Startups Are Betting Big on It

The US electrical grid is straining under surging demand from AI-powered data centres, prompting a wave of software startups to pitch digital solutions as cheaper, faster alternatives to physical infrastructure expansion.

The idea that software can significantly bolster the US electrical grid is gaining real-world traction in 2025, as electricity demand surges in tandem with the growth of data centres. Electricity prices in the US are up 13% this year, driven in large part by power-hungry AI models and experimental infrastructure projects like repurposed supersonic jet engines for data processing and space-based solar power trials. With data centre energy consumption expected to nearly triple over the next decade, the electrical grid has become a flashpoint for both environmental activism and economic anxiety.

Against this backdrop, software startups are stepping into a traditionally hardware-dominated sector, offering tools that promise to do more with what’s already there. Companies like Gridcare and Yottar are among those making the case that the grid doesn’t necessarily need to be rebuilt from the ground up; it just needs better visibility and smarter utilisation.

Spare capacity already exists, but it needs better coordination

Gridcare, for instance, has compiled data on everything from transmission lines and fibre-optic routes to extreme weather patterns and community sentiment. The company uses this data to identify underutilised parts of the grid that could accommodate more load. According to Gridcare, several viable sites have already been identified that utilities had previously overlooked. Yottar takes a more targeted approach, helping mid-sized energy users quickly connect to available capacity.

The company’s software identifies overlap between existing grid supply and rising demand, particularly from AI data centres, and streamlines the connection process for companies that need immediate access.
These solutions may help relieve pressure on utilities that are struggling to build new power plants fast enough to keep up with demand. Concerns about AI’s runaway electricity consumption have led some environmental groups to call for a moratorium on new projects, raising the stakes for utilities trying to keep rates stable and outages under control.

Virtual power plants are already in the wild

In parallel, several startups are taking a different approach, aggregating distributed batteries into so-called virtual power plants (VPPs) that can be dispatched to support the grid. These software-managed battery fleets act as flexible, on-demand energy sources that kick in during peak usage or grid instability.
Base Power is building one such fleet in Texas, offering leased batteries to homeowners at a lower cost. Homeowners benefit from backup power during outages, while Base earns revenue by selling the combined capacity back to the grid.

Terralayr is doing something similar in Germany, but without supplying hardware. Instead, it taps into batteries and storage systems that are already installed, bundling them together with proprietary software to form a decentralised energy resource. This “bring your own battery” model could be critical in Europe and other markets where residential solar and storage adoption has outpaced utility-scale upgrades.

Coordinating solar, wind, and batteries through software layers

Other startups are going upstream to address another critical challenge: the lack of coordination between disparate renewable sources. Wind, solar, and battery systems often operate in silos, meaning they idle unnecessarily or miss opportunities to stabilise the grid. Companies like Texture, Uplight, and Camus are developing integration platforms that synchronise energy production and storage across the grid. These platforms can manage supply and demand in real time, helping reduce waste and boost the value of clean energy assets.

“The goal is to make every asset contribute more and idle less,” said a spokesperson for Uplight. “That means you need software that can respond to volatility minute by minute, not quarterly planning cycles.”

Big players are moving in: Nvidia and Google

Tech giants are also getting involved. Nvidia has partnered with EPRI, a power industry R&D organisation, to develop AI models tailored specifically for the grid. These models are designed to improve forecasting, load balancing, and outage detection. Google is working directly with PJM, one of the largest regional grid operators in the US. The collaboration focuses on using AI to process a massive backlog of connection requests for new power sources.

These include solar farms, wind installations, and emerging storage technologies that want to feed power into the grid but face long approval timelines. By reducing administrative friction, the PJM-Google partnership could unlock thousands of megawatts of new capacity, without building a single new plant.

Why 2026 could be a turning point

Utilities are notoriously slow to adopt new technologies, largely due to regulatory scrutiny and risk aversion. The grid’s primary function is reliability, and anything that threatens uptime, however innovative, faces an uphill battle. Infrastructure investments are also expensive and politically sensitive, often drawing resistance from ratepayers and public service commissions.

That’s where software’s low cost and flexibility come into play. “If you can prove reliability, utilities will listen,” said one grid software CEO. “And if it saves money, regulators will listen too.” As data centre expansion, electrified transportation, and AI demand continue to accelerate, software may be the only scalable tool that can keep pace with change. In some states, permitting a new power plant can take years. Deploying software can happen in weeks.

Key Figures:

• 13%: Increase in US electricity prices in 2025
• 3x: Projected growth in data centre energy use over the next decade
• $0: Infrastructure cost for some virtual power plant deployments using existing batteries.

Whether or not utilities adopt these tools in earnest, the direction is clear: the future of the grid isn’t just steel and wires. It’s code.