The grid is entering a more complicated phase.

For the past decade, rising demand has quietly helped stabilize costs by spreading fixed infrastructure across more load. Now demand is accelerating again, but the system it’s landing on looks very different. Queues are clogged, retirements are ahead of replacement, and reliability is being managed in real time.

Two new reports this week point in the same direction from opposite ends of the timeline. One looks backward and says load hasn’t been the problem. The other looks forward and says that only holds if we get the next phase right.

A new analysis from the Institute for Energy Research challenges one of the most persistent narratives in energy policy: that data centers are pushing electricity prices higher.

Across all 50 states, there is effectively no relationship between data center concentration and electricity prices. States with the most data center activity have average prices that are nearly identical to the national average. Even more interesting, states with the fastest demand growth have seen lower price increases over time.

That pattern shows up clearly over the past decade. States with high electricity demand growth saw prices rise about 20%, while low-growth states saw increases closer to 40%. The underlying dynamic is simple. The grid is capital-intensive. When more energy flows across the same infrastructure, the cost per unit falls.

Why it matters - For the grid: Load growth has historically improved cost efficiency, not degraded it.
For policy: The case for restricting data center development on cost grounds looks weaker than advertised.

GridTake - IER is describing how the system has behaved when it still had room to absorb growth. The key question now is whether that condition still exists. Load is not inherently inflationary. It becomes inflationary when it forces poorly timed or poorly located investment.

The Brattle Group’s latest report flips the conversation from how much we need to build to how much we are failing to use.

On average, the U.S. power system operates at roughly half of its potential capacity across time. That gap creates an opportunity. If load growth is steered toward underutilized parts of the system and away from peak stress periods, the need for new infrastructure drops significantly.

Brattle models a scenario where demand rises 20–30% over the next five years. If that load is layered onto the system inefficiently, rates rise modestly. If it is integrated with a focus on utilization, rates fall. The difference is meaningful. Their modeling shows a swing of several percentage points on retail electricity prices, translating into over $100 billion in potential savings over a decade.

Why it matters - For the grid: There is substantial “hidden capacity” in when and where electricity is used.
For markets: Flexibility is becoming as important as generation.

GridTake - This is not a counterargument to building. It is a sequencing argument. Utilization can delay investment, reduce costs, and buy time. But it only works within physical limits. The risk is not that we optimize too much. It is that optimization becomes a substitute for expansion rather than a bridge to it.

At CERAWeek, Microsoft and Nvidia outlined a joint initiative to use AI to accelerate nuclear plant development.

The effort focuses on streamlining permitting, standardizing plant design, and deploying digital tools like “digital twins” to simulate projects before construction begins. Early results from partner companies suggest that AI-assisted permitting could dramatically reduce timelines and costs by identifying inconsistencies, reusing prior regulatory work, and aligning applications with existing approvals.

The broader goal is to move nuclear away from bespoke, one-off engineering and toward repeatable, scalable deployment.

Why it matters - For the grid: Nuclear remains one of the few firm resources capable of meeting sustained load growth.
For the industry: The bottleneck is shifting from technology to process.

GridTake - This is an attempt to turn nuclear into something closer to manufacturing. It may speed up parts of the lifecycle, but it does not eliminate the harder constraints. Supply chains, skilled labor, and capital discipline still govern how fast reactors get built. AI can compress timelines, but it cannot yet pour concrete.

The Department of Energy has extended emergency orders requiring coal units in Indiana to continue operating past their planned retirement dates.

The plants were initially kept online during winter reliability concerns, and federal officials now argue the conditions that justified those orders still exist. During peak winter demand and periods of low renewable output, the units provided measurable support to the Midcontinent system.

Critics argue the extension is unnecessary and expensive. Estimates suggest the continued operation of these units is costing ratepayers hundreds of thousands of dollars per day. Federal regulators have now allowed those costs to be spread across a broader regional base.

Why it matters - For the grid: Retirements are colliding with real-world reliability constraints.
For policy: Emergency authority is becoming a recurring tool, not a rare exception.

GridTake - Indiana is not an anomaly. It is an early signal. The system is being forced to reconcile planning assumptions with operational reality. Instead of adjusting the model, policymakers are extending legacy assets to close the gap. That works in the short term. Over time, it raises a more uncomfortable question about how much capacity is actually being replaced versus assumed.

A major U.S.–Japan energy agreement is driving a new wave of large-scale generation projects, including nearly 10 GW of natural gas capacity in Texas and Pennsylvania and 3 GW of small modular nuclear reactors in the Southeast.

The projects are tied directly to expected demand from data centers and advanced manufacturing. They also rely on coordinated federal support, long-term offtake agreements, and access to infrastructure like transmission and fuel supply.

The scale is notable. Individual gas hubs are being designed to support multiple gigawatts of continuous load, effectively functioning as dedicated power systems for large industrial customers.

Why it matters - For the grid: Capacity is being built in anticipation of demand, not just in response to it.
For markets: Load risk is becoming a central factor in investment decisions.

GridTake - This is the other side of the utilization story. If demand materializes as expected, these projects look essential. If it does not, they become overbuilt assets in a system that is already struggling with cost allocation. The bet is not just on energy. It is on the accuracy of demand forecasts.

Upgrade to Grid Brief Premium to get extra deep dives into energy issues all over the world.

Utilization Changes the Cost Curve

The Brattle analysis shows a clear split. Under a status quo approach, rising demand modestly increases electricity rates. Under a utilization-focused approach, the same demand reduces them.

The difference comes down to how the system is used. A grid that is built for peak demand but operated inefficiently carries excess cost. A grid that absorbs new load into underused capacity can defer investment and reduce pressure on rates.

The implication is straightforward but easy to miss. The cost of electricity over the next decade will be shaped as much by operational decisions as by physical buildout.

We rely on word of mouth to grow. If you're enjoying this, don't forget to forward Grid Brief to your friends and ask them to subscribe!