The energy landscape is undergoing a seismic shift, not driven by climate idealism, but by the brutal arithmetic of data center demand and the failure of intermittent resources to meet it. In a sprawling 98-page analysis titled "Fighting Words," Michael Cembalest of JPMorgan Chase dismantles the prevailing narrative that solar and storage can seamlessly replace traditional baseload power, arguing instead that we are witnessing a "misallocation of capital" on a historic scale. For the busy professional tracking infrastructure, this report is a stark warning: the grid is hitting a wall, and the solutions currently being funded may not be the ones that keep the lights on.
The Data Center Reality Check
The article centers on a single, undeniable driver of the current crisis: the insatiable appetite of data centers for reliable electricity. Cembalest highlights that the PJM Interconnection region, a massive grid covering parts of Virginia, Pennsylvania, Maryland, and Ohio, now hosts 67 gigawatts of existing and planned data center capacity. This concentration is creating a perfect storm for pricing. "The PJM region (data center alley: VA, PA, MD, OH) has 67 GW of existing and planned data center capacity, the largest cluster in the U.S.," the report notes, pointing to capacity payments acting as "insurance premiums" that have spiked to levels that would have cleared at $530 per megawatt per day without a cap.
This surge in demand is not just about total volume; it is about timing. The analysis reveals a disturbing trend where load growth is occurring at night, precisely when solar generation is zero. "Nighttime loads have increased in Virginia and ERCOT since 2023," Cembalest writes, noting that these facilities are driving evening demand while electric vehicles remain a negligible factor. This dynamic exposes a critical flaw in current grid planning: the rush to build solar and storage assumes a load profile that simply no longer exists.
"Nighttime load can be viewed as positive as it represents consistent load that allows utilities to monetize capital deployed assets and does not put additional strain during peak hours; but it's another sign of rising data center demand."
The author argues that this shift could force a re-evaluation of the entire deregulation experiment. "Last point: some utilities within PJM are questioning whether re-regulation would be the better option (Exelon, First Energy, PPL, and PSEG); I agree with them," Cembalest concludes. This is a significant admission from a major financial institution, suggesting that the market mechanisms designed to foster competition may be ill-equipped to handle the sheer scale and reliability requirements of the AI economy.
The Economics of Intermittency
Perhaps the most damning section of the report concerns the financial viability of renewable-heavy grids. The authors scrutinize the specialized electricity rates being proposed in seven states to cover the extra costs data centers impose. The conclusion is grim: even these higher rates are insufficient to fund the generation needed. "Some of the specialized rates imposed by various utilities would be able to pay for a combined cycle gas plant, but none of them are high enough to pay for a solar plus battery storage facility," the analysis states.
The cost disparity is widening rapidly. Solar power purchase agreements, once a steal at under $30 per megawatt-hour in 2019, have ballooned to over $60 per megawatt-hour. "It is increasingly looking like the rush to build solar and storage in the United States may go down as one of the largest misallocations of capital in our nation's history," Cembalest warns. This argument is bolstered by a direct comparison with the cost of new natural gas generation, which remains the only dispatchable option that current rate structures can support.
Critics might argue that this view ignores the long-term potential for renewable cost reductions or the necessity of decarbonization mandates. However, the report counters that the immediate economic reality is one of scarcity. "On this kind of topic, stick to peer-reviewed pieces in publications like Joule. EMBER's article is more a reflection of the world the authors want to exist rather than the world as it really is," Cembalest asserts, dismissing optimistic forecasts from think tanks that claim "baseload" solar is competitive. The math simply doesn't add up: for solar and storage to beat gas, gas prices would need to triple while solar capital costs plummet by 70%—a scenario the report deems highly unlikely in the current tariff environment.
The Capacity Illusion
The article also tackles the dangerous misconception that building more megawatts equals a more reliable grid. While the U.S. is installing new capacity at record rates, the quality of that capacity is deteriorating. "Even though the U.S. is building more new capacity annually than ever before, it's not equating to reliable, firm capacity because new capacity builds are increasingly made up of intermittent wind and solar resources," the report explains.
This trend is exacerbated by the retirement of thermal assets and a shortage of skilled labor and critical equipment like transformers. "The U.S. is now at pre-2005 levels of firm capacity on the grid at a time when electricity demand is projected to have the largest increases in over a decade due to data center and AI growth and electrification efforts," Cembalest notes. This creates a precarious situation where the grid is expanding on paper but shrinking in terms of actual, dependable power.
The demand for gas turbines has surged as a result, with global production capacity expected to be maxed out through 2030. "Three companies each have 20 to 25 percent of the global turbine market share: GE Vernova, Siemens, and Mitsubishi, and each is planning to expand production," the article details. Yet, even with these expansions, the backlog suggests that supply cannot keep pace with the sudden, massive demand from data centers.
"Whether the same pace can be maintained today is another question, particularly given shortages of skilled energy labor, shortages of transformers, breakers and other equipment and tariffs on grid equipment which do not benefit from the kind of exclusions granted to semiconductors and computers."
The report also casts a skeptical eye on Small Modular Reactors (SMRs), often touted as the silver bullet for clean, firm power. While investor enthusiasm is high, the economic reality is stark. "A 2025 paper from the former chair of the US Nuclear Regulatory Commission reviewed four SMR types and estimated SMR levelized costs of $200 to $400 per MWh," Cembalest writes, noting that these costs are far above what is considered affordable. The burden of proof, the author argues, remains firmly on the nuclear industry to demonstrate viability.
The Bottom Line
The strongest element of this analysis is its refusal to sugarcoat the mismatch between current renewable deployment strategies and the physical realities of data center demand. By grounding the argument in hard numbers from PJM and ERCOT, the report effectively dismantles the idea that solar and storage alone can solve the coming energy crunch. However, the piece's heavy reliance on current cost curves may understate the speed of technological breakthroughs in storage or the political will to force nuclear deployment despite high upfront costs.
The ultimate takeaway is clear: the era of assuming that "green" automatically means "cheap" and "reliable" is over. As the administration and grid operators grapple with these constraints, the focus is shifting decisively toward dispatchable power, regardless of the fuel source. The next few years will test whether the market can pivot fast enough to avoid a reliability crisis before the lights go out.