Data Centre Load Flexibility Is Becoming The New Currency For Quicker Grid Access

Grid planners have long treated large new loads as fixed demand that needs firm capacity in every hour of the year. That assumption is breaking. With new data centre connections taking five to seven years – or more – in key US markets, a load’s willingness to flex is starting to decide how quickly it connects.

The reason for this is that a little flexibility buys a lot of time. A study by Camus Energy, encoord and Princeton University modelled six candidate 500MW data centre sites in one PJM utility’s territory and found pairing a flexible connection with on-site capacity could bring them online three to five years sooner, with grid power still available in more than 99% of hours and on-site resources needed for only 40 to 70 hours per year. That changes the role of large loads: a data centre that can flex stops being a problem the grid must absorb, and becomes a resource grid operators can plan around.

At the end of 2025, Verdantix predicted that Google would become the US power market’s most sought-after source of flexibility, provided a new federal rule let utilities fast-track large loads willing to curtail at peak times (see Verdantix Market Insight: 10 Predictions For Energy Leaders In 2026 And Beyond). That rule has not arrived in the form we expected, as regulators remain wary of making flexibility mandatory rather than voluntary. For example, on October 9, FERC rejected PJM’s proposal to let large loads connect without new capacity if they accepted being curtailed first, citing concerns over discrimination and state jurisdiction.

Instead, flexibility is advancing through commercial contracts. By March 2026, Google had written 1GW of demand response into long-term contracts with five US utilities, curtailing machine-learning workloads when the grid is stressed, and it says this is part of how its new data centres connect more quickly.

In September, Google went further by co-founding the AI Energy Management Alliance with Nvidia and Emerald AI, which is pushing for flexible data centres to earn faster and larger grid connections. That flexibility doesn’t have to mean switching off servers. The alliance counts shifting workloads, discharging batteries and running on-site generation as ways to cut grid draw. It also has limits, with Google acknowledging that not every data centre can flex, and only at certain locations. Latency-sensitive inference and colocation tenants with strict uptime contracts have far less room to move.

It’s not just data centres – any large load stuck in a connection queue can offer the same bargain, whether that is an EV fleet depot that shifts its charging or an industrial site that can curtail processes at peak. In Great Britain, Ofgem is already considering connection terms for new demand that are non-firm (curtailable when the network is constrained) or phased (capacity stepped up over time), in exchange for earlier connection dates.

For businesses planning to add load, demonstrable flexibility is becoming a currency for earlier grid access, not just a lever on energy bills. For vendors, the alliance's emphasis on verifiable, performance-based flexibility shows where value will sit: in the metering, control and orchestration that prove how much load can move, how quickly and for how long.

Until regulators settle the rules, grid access will be won case by case, and the loads that can promise flexibility will move to the front of the queue.

For more on the value of flexibility, read Best Practices: Unlocking The Energy Flexibility Opportunity; for more on how grid regulation shapes the technology market, see Market Insight: Using Grid Regulations To Shape Go-To-Market Strategy.