Grid Watch

PG&E crossing one million rooftop solar interconnections is an operational milestone, not a capacity milestone, and the distinction matters. PG&E's own language — 'a one-way grid to an interactive system' — is accurate engineering description. A million distributed resources on the grid means a million new decision points for frequency response, voltage regulation, and export-flow management. That's complexity, not just capacity. The question Grid Watch always asks is: do the electrons exist when the load demands them? Rooftop solar does not provide firm capacity at evening peak, and California's duck curve is not hypothetical.

The NOAA degree-day snapshot for the week of 2026-06-03 to 2026-06-09 shows zero cooling degree-days across all ten monitored metros and 1,444 total heating degree-days, with Seattle carrying 151.7 HDD over seven days. That cross-metro CDD of zero tells you summer load hasn't arrived yet in the monitored system. The grid is not being tested this week. The stress test comes when CDD loads hit simultaneously with the solar intermittency problem at dusk — and that season is weeks away, not months.

The AEI corpus item is worth flagging: AI infrastructure buildout could push U.S. data-center power demand past $1 trillion in capex, with the binding constraint identified as power — not capital. Meta's $115M workforce academy for data-center construction (corpus: Construction Dive) and Google's Virginia community energy investments point to the same bottleneck: interconnection queues and grid capacity, not financing. We have not seen the interconnection queue data in today's corpus, but the pattern is consistent with what Grid Watch tracks: large loads seeking grid access faster than the system can accommodate them.

The U.S. floating LNG terminal story has a grid angle that gets missed: LNG liquefaction is an energy-intensive industrial load. The nine existing U.S. terminals are already drawing significant power from regional grids. A floating unit, depending on siting, could draw from offshore generation (unlikely at scale) or from onshore grid interconnection. The policy assumes electrons that do not yet exist in sufficient firm capacity. Here is what the grid can actually deliver: more solar, more wind, more batteries — but the firm dispatchable capacity that liquefaction requires runs on gas or stays on the interconnection queue.

The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver — and what a prolonged Hormuz disruption does to the inputs that keep it running. Let's start with the degree-day picture: the NOAA 7-day snapshot for June 2–8 shows cross-metro totals of 1,420 HDD and exactly zero CDD. San Francisco leads at 150.9 HDD over the 7-day window. This is a late-spring heating signature, not a summer cooling peak — load is currently manageable. New York registered zero CDD. We are in the pre-stress window before summer peak demand, and that is the only saving grace in an otherwise tightening picture.

The AI-driven power demand surge is the structural backdrop. The U.S. News headline citing EIA projections of record-high power use in 2026 and 2027 driven by AI data center load is not a surprise to anyone managing reserve margins — it's the constraint that was already stressing interconnection queues before today's geopolitical shock. The Inside Climate News piece on rural Alabama's opposition to a hyperscale data center on the historic Selma-to-Montgomery corridor illustrates the permitting and community-opposition friction that slows the capacity additions needed to serve that load.

Now layer in Hormuz. U.S. natural gas grid exposure to a Hormuz closure is indirect but real: Henry Hub spot sits at $3.07/MMBtu (week of June 1), and Lower-48 NG storage is at 2,578 Bcf with a healthy +95 Bcf WoW injection. The domestic gas buffer is solid for now. But the longer-tail risk is LNG export competition — the Alaska LNG tax break bill advancing to the House floor (adn.com) signals that U.S. policymakers see LNG export as a strategic tool in exactly this kind of geopolitical moment. Every incremental LNG export tightens the domestic supply available to gas-fired generation, especially if Hormuz disruption accelerates the global LNG scramble. Grid operators need to watch the Henry Hub-to-TTF spread as a leading indicator of domestic gas diversion pressure.

The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver — and what today's data says about the gap. The EIA is on record projecting U.S. power use to smash record highs in 2026 and 2027 as AI data-center load surges, a consensus-rated signal per independent review. Musk's xAI is already facing class-action litigation over data center 'nuisance,' which is what happens when load growth outruns grid integration planning. These are not abstract forecasts; they are queue pressure expressed as legal filings.

The NOAA 7-day degree-day window ending June 8 shows 1,420 HDD cross-metro and precisely 0 CDD — San Francisco leading heating demand at 150.9 HDD over the 7-day window, New York posting 0 CDD. The grid is not yet in summer peak-load territory. That is the narrow good news. The structural bad news is that AI load does not follow a seasonal curve the way residential cooling does. Data center demand is flat and relentless, 8,760 hours a year, and it is landing on a transmission and interconnection queue that was already backlogged before the current AI buildout cycle.

The Jamaica islandwide blackout — triggered, per a preliminary JPS report to the Office of Utilities Regulation, by issues that caused at least three prior grid collapses — is a useful mirror. Small-island grid fragility is not the U.S. situation, but the failure mode is instructive: cascading causes, deferred maintenance, demand that exceeds redundancy. The U.S. has a renewable share of 5.94% of generation as of March 2026 — a number that should make anyone pause who assumed grid decarbonization was moving fast enough to absorb AI load growth cleanly. Henry Hub at $3.07/MMBtu and NG storage at 2,578 Bcf keep the gas-fired backup cushion intact for now, but gas at $3 is also a signal that the market does not yet believe AI demand has tightened the power sector. That could reprice fast if a hot July materializes.

The Trump administration's emergency order keeping the Orlando Utilities Commission coal plant running — originally scheduled for retirement in 2025 as part of OUC's net-zero-by-2050 plan — is a grid reliability intervention that deserves scrutiny on both the policy and the engineering merits. The administration's implicit argument is that retiring the plant creates a reliability gap. That argument is not inherently wrong, but it requires a specific, quantifiable answer: what is the reserve margin in Florida's peninsular grid without this plant, and what firm capacity is the replacement solar build actually delivering?

From the NOAA degree-day snapshot for the week ending June 6, the cross-metro 7-day total was 1,461 HDD and zero CDD. That is a winter/shoulder-season load profile — not the Florida summer peak that would stress the grid. The real test of whether the OUC retirement schedule was grid-safe is the July-August peak load window. If the emergency order was issued in June on the basis of summer reliability concerns, the administration should be publishing the reliability study. Without it, this is a policy decision dressed as an engineering necessity.

The policy assumes electrons that do not yet exist — specifically, the solar-plus-storage capacity that OUC had planned to deploy as the replacement. Whether that capacity is actually permitted, contracted, and deliverable before the next summer peak is the operative question. Henry Hub at $3.07/MMBtu and L48 NG storage at 2,578 Bcf (week ending May 29, +95 Bcf WoW) suggest natural gas backup is plentiful and cheap, which makes the coal retention order's economics even harder to defend on pure grid terms unless the coal plant provides specific ancillary services — voltage support, inertia — that gas peakers in that footprint do not.

Two grid stories arrived this week that should not be read in isolation. The Texas grid operator flagged reliability risks after data centers and crypto mining sites failed voltage tests, per Reuters. This is not an abstract warning—voltage compliance is a foundational grid interconnection requirement, and failures mean these loads are drawing power in ways that stress local distribution infrastructure. The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver: Texas is entering summer peak season with large, fast-ramping industrial loads that were provisionally interconnected without adequate voltage discipline. That is a recipe for localized reliability events, not a statewide blackout, but a serious operational problem regardless.

The AI electricity demand angle reinforces this. The UN warning cited by Live Science—that AI could consume up to 3% of global electricity—is arriving in the same news cycle as the Texas voltage failures. These are related phenomena: hyperscale AI infrastructure is among the large industrial loads straining distribution-level voltage management. The NVIDIA-Doosan collaboration on 'AI factory infrastructure' announced this week is further evidence that the load buildout is not theoretical.

On the thermal picture: the NOAA 7-day degree-day pull (window May 30–June 5) shows Boston leading with 151.7 HDD over seven days, and the cross-metro total of 1,468 HDD with zero CDD tells us we are still in a late-spring heating pattern across the Northeast, not a cooling-load surge. Summer peak stress has not arrived yet. That is the one grid-favorable data point this week—the demand spike that will test Texas and the broader ERCOT system is still weeks away. But the voltage compliance failures documented by Reuters mean ERCOT is heading into that peak season with known structural vulnerabilities already on the books.

The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver — and what the past 7 days of degree-day data tell us about near-term load. The NOAA NCEI pull for the window May 29 through June 4 shows Boston leading with 152.3 HDD over 7 days, with cross-metro totals of 1,471 HDD and zero CDD. Zero cooling degree-days across ten major metros. This is the shoulder-season lull — the brief window before summer cooling load hits. Grid operators should not be complacent: the CDD clock is about to start, and the Montana wildfire outlook from Inside Climate News flags heat and wind events arriving in the West that will drive cooling load in that region faster than the Northeast degree-day picture suggests.

SPC Watch 281 covers severe thunderstorm risk across Connecticut, Nassau, and Suffolk counties as of June 7. Watch 279 covers Indiana and Ohio corridors. These are not headline grid emergencies, but simultaneous convective events across multiple ISOs during a demand ramp can stress transmission paths that are already operating close to limits. The Iraq DW story — Baghdad facing summer blackout risk despite solar potential — is an instructive mirror: the U.S. grid is better resourced, but the lesson is that heat-driven demand spikes without adequate dispatchable backup collapse reliability. The Bishkek outage caused by a single 110kV line failure taking out 70% of the capital is a stark reminder that single points of failure in transmission infrastructure are not a third-world problem.

On domestic generation mix: EIA reports renewable share of U.S. generation at 5.94% for March 2026. That figure reflects winter-spring measurement. As summer CDD load builds, that percentage will need to climb through increased solar and wind dispatch — but the binding constraint remains dispatchable capacity and interconnection. Henry Hub at $3.07/MMBtu with NG storage at 2,578 Bcf (up 95 Bcf week-over-week as of May 29) means gas peakers have cheap fuel and full tanks. That is your reliability buffer for the summer. The question is how long $3.07 gas holds if LNG export demand accelerates as Hormuz disruption forces Atlantic Basin buyers to bid harder for U.S. LNG cargoes.

Two signals dominate the grid picture today, and they pull in opposite directions on the reliability question. First, the administration's $425 million commitment to extend 12 coal plants and fund two new ones injects capacity onto the balance sheet — but the policy assumes that 'extended' coal plants deliver megawatts on demand, and that assumption requires scrutiny. Aging coal units face forced outage rates that have been climbing for years; a modernization grant does not automatically translate to a reliable dispatchable megawatt. The DOE's framing of 'energy dominance' is not an engineering specification.

Second, Fervo Energy's transmission constraint disclosure is the more structurally important grid signal. Jefferies analyst Julien Dumoulin-Smith cited management's own language: transmission constraints are a risk factor, with behind-the-meter deployment floated as a workaround. This is the interconnection queue problem made visible in a single geothermal developer's earnings language. The West is not short of generation ambition — it is short of transmission to move electrons from where they can be produced to where they are needed. The policy assumes electrons that do not yet exist on the wire.

On the demand side, the NOAA degree-day snapshot for the seven-day window ending June 3 shows 1,471 total HDD across ten metro stations, with Boston leading at 152 HDD — a heating load signal in early June that indicates unseasonably cool Northeast conditions. Cross-metro CDD are zero. This suppresses summer cooling load in the near term, providing modest breathing room on reserve margins. But the SPC severe thunderstorm watch active across multiple Iowa counties introduces near-term reliability risk in the Midwest: storm-driven outages can spike load-serving requirements rapidly, and a generation mix still reliant on aging thermal units is not the most resilient response asset. Watch MISO reserve margins over the next 48 hours.

New York's legislature just did something no state has done before: it passed a one-year moratorium on data-center permits (Inside Climate News). If Governor Hochul signs it, New York becomes the first state to explicitly freeze power-hungry load additions through regulatory fiat rather than interconnection queue management. This is not a climate story. This is a capacity story. Data centers are among the fastest-growing load classes on any grid they touch, and the northeast grid has been under sustained strain from insufficient new generation capacity to match load growth. The legislature is reaching for a blunt instrument because the normal tools—interconnection queues, capacity market signals, demand response procurement—have not moved fast enough.

The NOAA degree-day data is currently providing some breathing room. Across our ten monitored metros for the week of May 27–June 2, cross-metro CDDs totaled zero. Heaviest heating demand was Seattle at 151.4 HDD over seven days; New York recorded zero CDD. With 1,468 HDD and 0 CDD across all ten metros, the summer cooling load that would stress northeastern grids hasn't materialized yet. That won't last past late June. The moratorium, if signed, will slow load additions on paper. But the grid doesn't run on permits—it runs on electrons. The policy assumes data centers that won't get built will not draw power. What it cannot do is add generation capacity or transmission, which remains the binding constraint.

MIT Technology Review's coverage of virtual power plants for data centers is the more operationally interesting signal. If large data center loads can be dispatched as demand-response resources—essentially converted from grid burdens to grid assets—that changes the capacity math. The Sabine Pass LNG expansion (Construction Dive) is also grid-adjacent: more LNG throughput requires sustained gas-fired generation to backstop intermittent renewables, especially with renewable share of U.S. generation sitting at just 5.94% as of March 2026 (EIA). The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver: not enough new generation to absorb unconstrained AI-driven data center load, and a moratorium that delays the problem without solving it.

The NOAA degree-day window for May 26 through June 1 shows zero cooling-degree-days across all ten monitored metros, with cross-metro HDD totaling 1,465 — Seattle alone contributing 151.6 HDD over seven days. The grid is in that narrow seasonal corridor between winter heating demand and summer cooling load. For system operators, this is the maintenance and interconnection window, not a reliability crisis. But the absence of cooling load right now makes the New York climate law story more politically legible, not less urgent technically.

Governor Hochul's retreat from the CLCPA is operationally significant for the New York grid. The law had set aggressive targets for phasing out natural gas in building heating and power generation. Walking back those commitments means Con Edison and NYISO planning assumptions now carry more uncertainty — utilities have been queuing interconnection requests for offshore wind and battery storage predicated on policy certainty that just moved. The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver: New York's current generation mix still leans on natural gas for dispatchable capacity, and the offshore wind projects in the interconnection queue cannot replace that firm capacity on a linear timeline even if the policy had held.

The EIA reports that U.S. underground working natural gas storage capacity in the Lower 48 increased slightly in 2025, with growth concentrated in the South Central and Mountain regions. Lower-48 NG storage sits at 2,483 Bcf as of May 22, with a weekly injection of plus 92 Bcf — a healthy injection pace as summer refill season begins. Henry Hub spot is $3.07/MMBtu as of June 1. For grid operators, cheap and available gas is a reliability buffer. It is also the reason Hochul's political calculus is what it is.

The grid does not care about geopolitics until it has to. And it may have to care sooner than the policy calendar expects. The NOAA 7-day degree-day window (May 25–31) shows zero cooling-degree-days across all ten measured metros — cross-metro CDD total: 0. That means the summer demand surge has not yet arrived. Seattle led the heating load at 151.5 HDD over seven days; cross-metro HDD totaled 1,432. We are still in the shoulder season. The relevant question is not what the grid is doing today but what it will be doing in eight weeks when the El Niño signal — flagged at 80% probability by the UN's World Meteorological Organization for June-through-August development — starts compressing cooling load onto a system that has not finished its capacity additions.

The Massachusetts vehicle-to-everything (V2E) demonstration is the quietly consequential story. Utility Dive reports that light-duty EVs enrolled in the state's virtual power plant can earn roughly $3,000 per summer, with school buses earning up to $12,000 — bidirectional capacity that is real, contracted, and dispatchable. This is not vaporware. But it is also not yet at scale. The interconnection queue for distributed resources remains the binding constraint. A school bus fleet discharging to the grid during a peak demand event is operationally compelling; a thousand school bus fleets requires DERMS software, utility tariff reform, and interconnection agreements that exist nowhere near fast enough.

Californically, EIA reports that natural gas spot prices in California hit record lows in the first five months of 2026 — PG&E Citygate and SoCal Border Average both at historic floors, with SoCal Citygate near-record. Above-average inventories and decreasing in-state demand are the cited drivers. That is a reliability backstop story in the near term: cheap gas means gas-fired peakers will run. But it also suppresses the revenue signal that would otherwise incentivize new storage and transmission build. The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver: adequate for now, but the El Niño summer is the stress test that matters.

The NOAA 7-day degree-day pull through May 30 tells a clear story for the near term: cross-metro HDD totals of 1,463 with 0 CDD across the 10 sampled stations, Seattle leading heating demand at 150.8 HDD over the week. That is a late-spring thermal picture, not a summer stress scenario — peak cooling load has not materialized yet. But the grid does not get credit for calm weather it has not yet faced. The summer reserve margin question is coming, and the supply-side inputs are now less certain than they were 90 days ago.

The Hormuz disruption is not just a crude story — it is a fuel-oil and distillate story with grid implications. Peaker plants and dual-fuel generators that fall back on distillate when gas is tight are exposed to a Middle East-disrupted refined product market. Henry Hub at $3.10/MMBtu with 2,483 Bcf in storage provides near-term gas supply comfort, but if LNG export pull intensifies — which Barrel Report's Hormuz rerouting thesis implies — domestic gas prices could firm into peak summer demand. The interconnection queue for new generation is not going to bail anyone out on a 90-day timeline.

The DOE rebate guidance stripping electrification from the $8.8 billion program is operationally significant for load forecasting. Less heat-pump adoption means slower transition of heating load to the grid, which sounds like relief — but it also means continued dependence on fuel oil and propane for residential heating, precisely the fuels whose supply chains run through the refining infrastructure most exposed to Hormuz disruption. The policy assumes the grid's problems are on the demand side. The supply-side risk is in the fuel mix.

The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver. The NOAA 7-day degree-day snapshot (2026-05-23 to 2026-05-29) shows zero cooling degree-days across all ten monitored metros, with the heaviest heating demand in Seattle at 151.9 HDD. Cross-metro total: 1,439 HDD, 0 CDD. That is a late-spring profile with no heat-driven demand emergency on the current 7-day window. The grid is not under acute load stress today. But the wildfire season warning from Inside Climate News changes the forward picture significantly for the Western Interconnection.

The January 2025 Los Angeles fires destroyed 16,000+ structures and killed 31 people. The grid implication that goes underreported: wildfire events in the West do not just destroy demand-side infrastructure — they destroy transmission corridors, force de-energization of high-voltage lines (the primary tool utilities use to prevent ignition), and create reliability emergencies precisely when cooling loads are highest. If 2026 fire season forecasts are worse than 2025's 'dodged bullet,' Western grid operators face a simultaneous pressure of de-energization events reducing deliverable capacity and heat-driven demand spikes increasing load. That is a reserve margin problem, not a generation capacity problem.

On the macro grid side: U.S. renewable share at 5.94% (EIA, March 2026) is the binding constraint Transition Monitor identifies. From a grid operations standpoint, the relevant question is not just total renewable share but dispatchable capacity backing it. With NG storage at a healthy 2,483 Bcf and Henry Hub at $3.10/MMBtu, the gas backup stack is well-positioned for the near term. The interconnection queue backlog — not directly cited in today's corpus but the structural backdrop for every renewable deployment number — remains the operational bottleneck that neither the RFF report nor any policy target has a mechanism to clear quickly.

The Axios framing—'the gold rush beneath the AI boom'—is colorful, but it obscures the operational question: where, exactly, does the load land, and is there capacity to serve it? U.S. renewable share of generation stood at just 5.94% as of March 2026 per EIA, a figure that should give pause to anyone assuming clean electrons will absorb AI datacenters on current timelines. The NOAA 7-day degree-day window through May 29 shows zero cooling-degree-days across all ten monitored metros—cross-metro total 0 CDD—meaning we are not yet in the summer peak demand season. Seattle led the heating side at 151.9 HDD over seven days; the cross-metro HDD total was 1,439. The pre-peak lull is exactly when interconnection queues pile up and developers lock in capacity; the question is whether the projects in queue are real megawatts or paper promises.

The spent nuclear fuel story from OilPrice flags a geopolitically significant angle: if domestic uranium reprocessing becomes viable, it meaningfully changes the baseload math for grid operators who have been staring down the retirement of dispatchable generation without adequate replacement. But the headline possibility remains far from committed capacity. Ukraine's strikes on Russian energy infrastructure—the Saratov oil refinery fire confirmed by multiple outlets—do not directly threaten U.S. grid operations, but they reinforce the lesson that centralized energy infrastructure is a single-point-of-failure problem. The AI demand surge is pushing U.S. grid operators toward exactly that kind of concentration risk: massive datacenters co-located with generation, creating load pockets the bulk transmission system was not designed to serve. The policy assumes electrons that do not yet exist. Here is what the grid can actually deliver: 5.94% renewable share, a nuclear fleet under geopolitical uranium pressure, and a summer peak season that hasn't even started.