ERCOT's Demand Outlook Cut: The AI Power Boom's Expectation Gap

The market has fully priced in a massive AI power boom. Analysts are throwing around numbers that suggest a coming surge in demand, but the reality on the ground is starting to show cracks. The expectation gap is widening between the bullish whispers and the tangible pressures building in the nation's largest grid.

On paper, the story is explosive. BloombergNEF recently raised its forecast for U.S. data center power demand to 106 gigawatts by 2035, a 36% jump from its prior estimate. That's a clear signal that the market consensus is looking out a decade and seeing a transformative load growth. The sheer scale of planned projects mirrors this optimism. The Electric Reliability Council of Texas (ERCOT), the grid operator for most of the state, now faces an interconnection queue ballooning to over 164 gigawatts. For next year, the forecasted queue is even larger, at 266 gigawatts, with data centers dominating the list. This isn't just a backlog; it's a pipeline of demand that appears to be priced into the long-term outlook.

Yet, industry experts are sounding a warning about the quality of these plans. There's growing skepticism that developers are flooding grids with inflated, 'phantom' project plans that may never materialize. This creates a dangerous disconnect. Utilities are forced to plan for new generation and transmission based on these inflated numbers, with the costs eventually passed to ratepayers. As one analyst noted, the industry needs to prevent "double, triple and quadruple counting" of projects. The sheer size of the ERCOT queue, if fully built, would require a massive expansion of the grid that simply cannot happen overnight.

The bottom line is a classic expectation gap. The market is betting on a smooth, inevitable ramp-up of AI power demand. The grid reality, however, is one of physical constraints, planning delays, and potential overbuilding driven by speculative or opportunistic proposals. The priced-in boom is facing a reality check from the other side of the meter.

ERCOT's Reality Check: A Guide to the Guidance Reset

The expectation gap is now triggering a formal guidance reset. As the sheer volume of speculative projects chokes the system, ERCOT and utilities are taking concrete steps to separate the wheat from the chaff, testing the sustainability of the bullish demand narrative head-on.

The scale of the disconnect is stark. While the market consensus looks out a decade to a transformative load, the official government baseline is far more cautious. The Energy Information Administration, which tracks demand for the federal government, generally only publishes detailed projections out two or three years. This creates a vacuum that is being filled by aggressive, long-term estimates like BloombergNEF's forecast for U.S. data center power demand to hit 106 gigawatts by 2035. That's a clear signal of what is priced in for the distant future, but it contrasts sharply with the immediate planning horizon of regulators and utilities.

To manage this uncertainty, ERCOT itself is projecting a dramatic acceleration in near-term demand. The grid operator forecasts that future demand could potentially jump from 85,000 MW to 145,000 MW within five years. This is a massive, five-year ramp-up that would require unprecedented investment.

Yet, industry experts are already warning that the interconnection queue is a graveyard of 'phantom' project plans that may never materialize. This is the core of the expectation gap: the official guidance reset is happening on the ground, not in the long-term whispers.

Utilities are responding with blunt tools to weed out the speculative from the viable. AEP Ohio recently cut its list of pending projects by nearly 30%, citing a need to remove opportunistic developers. More broadly, utilities are introducing financial barriers to discourage "queue squatters." One notable example is ComEd's plan to impose 'queue fees' of $500,000 per 100MW. The goal is to force developers to demonstrate serious financial commitment, thereby clearing out zombie projects that clog the system and distort demand forecasts.

The bottom line is a market in search of a new baseline. The priced-in AI boom is facing a reality check from the other side of the meter, where physical constraints and speculative overbuilding collide. ERCOT's projection of a 70% demand surge in five years sets a high bar, but the utility actions-like ComEd's hefty fees-are a clear signal that the industry is no longer willing to treat every queue entry as a guaranteed future load. This guidance reset is a necessary step to align long-term expectations with the grid's actual capacity and the financial reality of project execution.

The Tech Giant's Playbook: Mitigating Risk and Shifting Costs

Major tech companies are no longer content to let the expectation gap play out on utility balance sheets. They are actively mitigating risk by taking financial and operational responsibility for their AI infrastructure build-out, a shift that could fundamentally alter the investment thesis for utilities.

Microsoft's new "Community-First AI Infrastructure" plan is a textbook example of this strategy. The company is explicitly committing to pay higher utility rates to cover the electricity costs of its data centers. This is a direct move to absorb cost pressure that would otherwise be passed through to ratepayers via utility rate cases. By doing so, Microsoft is shifting a major financial burden away from the grid operator and its customers, effectively insulating the utility from a key source of future rate hikes.

More broadly, the plan signals a recognition that the AI infrastructure build-out will be costly. The company's pledge to add to the tax base for your local hospitals, schools, parks, and libraries and to not ask for local property tax subsidies is a clear attempt to manage community relations and regulatory risk. This approach directly addresses a past criticism where tech companies received subsidies that didn't always deliver promised jobs. By promising jobs and tax revenue without asking for breaks, Microsoft is trying to build goodwill and reduce local opposition that could delay projects.

The bottom line is that this playbook reduces the perceived need for massive, risky utility capital expenditure. If tech giants like Microsoft are willing to pay the premium for power and fund community benefits themselves, utilities may not need to shoulder the same level of upfront investment to secure new load. This could lead to a more stable, less volatile utility earnings profile, as the expectation gap narrows between speculative demand and the actual, paid-for capacity. For investors, it's a shift from betting on utility-led grid expansion to betting on tech companies managing their own infrastructure costs.

Catalysts and Risks: The Path to a Guidance Reset

The path forward hinges on a few key signals that will either validate the bullish demand thesis or confirm the looming guidance reset. The expectation gap is now being tested by real-world data, and the catalysts for a correction are becoming clearer.

First, watch for a decline in new interconnection applications or a slowdown in queue growth. The current data center boom is being fueled by speculative activity, with developers often submitting the same project to multiple utilities to find the lowest rates. This leads to phantom data centres that clog the system and distort forecasts. If the speculative wave is cooling, we should see a tangible slowdown in the number of new filings. A sustained drop would be a major red flag, indicating that the initial surge of "build now, think later" enthusiasm is fading.

Second, monitor utility earnings and guidance for any adjustments. As utilities implement new financial barriers like ComEd's $500,000 per 100MW queue fees, they will be forced to reassess the viability of projects in their queues. This filtering process is already underway, with utilities like AEP Ohio cutting their pending project lists by nearly 30%. The next step is for these companies to formally adjust their long-term demand forecasts and capital expenditure plans. Any guidance that shows a downward revision to projected data center load growth would be a direct market signal that the expectation gap is closing.

The primary risk, however, is a more fundamental guidance reset. The bullish narrative assumes a smooth, aggressive build-out of AI infrastructure. The counter-risk is that AI adoption or data center efficiency gains could outpace these plans. Two of the nation's largest independent power producers have already warned investors that the utility industry's demand projections for just three major markets exceed many credible projections for data center demand growth for the entire country. They argue that interconnect queues may be overstated anywhere from three to five times what will actually materialize. If this proves true, the entire investment thesis for new power generation built to serve this demand could be called into question, leading to stranded assets and a painful reassessment of utility earnings.

The bottom line is that the market is waiting for these catalysts to resolve the uncertainty. Until we see a slowdown in speculative filings and hear from utilities themselves that they are scaling back their growth assumptions, the expectation gap will persist. The risk is that when the guidance reset finally happens, it will be sharper than the market is currently pricing in.