Energy Arbitrage AI

Power & gas trader, 30 yrs (ARCO/BP, Constellation, TotalEnergies). $2B+ closed. I break down grid pricing for mid-market plants. Founder, Energy Arbitrage AI.

[Energy Arbitrage AI @energyarbitrage]

Utility rate cases run on a schedule, and the filings are public.

Which means an increase that lands on the invoice is usually visible months before it arrives. The docket is open, the testimony is posted, the proposed structure is on the record.

Almost nobody at the facility level reads them, so the increase gets treated as a surprise every time.

Is there a rate case open in your territory?

[Energy Arbitrage AI @energyarbitrage]

A single afternoon peak can appear on one bill more than once.

Billed as the monthly maximum demand. Billed again inside a time of use window. Sometimes a third time against a contract or facilities charge. Same fifteen minutes, priced through separate mechanisms that each have their own definition.

Get the window definition wrong, or sit in the wrong rate class, and it repeats every cycle until somebody actually reads the tariff.

How many times does your peak land on one bill?

[Energy Arbitrage AI @energyarbitrage]

Equinor's 100MW / 200MWh battery in Harlingen just started operations. Citrus Flatts will charge when ERCOT prices drop and discharge into scarcity.

If you run a plant in the Valley, the question is not whether the battery exists. It is what you pay during the evening ramp after it has already cycled.

Texas industrial retail ran 6.58 cents per kWh as of June 2026. A fixed contract locks energy cost. It does not lock demand charges, and it does not stop a ratchet clause from pricing the next eleven months off one fifteen-minute peak when the grid is short and the battery is empty.

A broker paid from inside your rate has no structural reason to explain when the ratchet resets. That is not about character. It is about where the incentive points.

Which demand tariff are you on?

[Energy Arbitrage AI @energyarbitrage]

Think your facility’s utility bill is just non-negotiable operational overhead?

Test it.

Most CFOs and Operations teams running commercial or industrial facilities ($500K to $5M+ annual energy spend) look at the total line item, verify usage looks roughly right, and approve the payment. They pay the bottom line, file the receipt, and move on.

The important questions are underneath: • How is your facility classified? • How are your demand charges actually calculated? • What specific tariff rate structure are you on?
Those details tell you whether high costs are simply your operational load—or whether the rate structure itself deserves a closer look.
Billing errors, improper rate code matching, and miscalculated demand determinants happen far more often than utilities care to admit.
We don't sell another high-level spreadsheet. We run a deep Forensic Audit, testing your actual billed charges directly against the utility’s legally filed rate leaves.

• Classification Errors: Are you sitting on an outdated or sub-optimal rate structure? • Demand Charge Mechanics: Are your peak time windows and billing determinants calculated accurately against the tariff? • Forensic Rate Matching: Is there money sitting on the table that belongs back in your operating budget?

The bill against the filed rate. If we don't find recoverable funds or verified optimizations, you owe nothing.

DM us directly to run a preliminary Forensic Gap Analysis on your facility.

[Energy Arbitrage AI @energyarbitrage]

A study says transmission expansion in PJM and the Eastern Interconnection could save billions by lowering retail electric rates. The savings are measured in energy cost per kilowatthour.

Most industrial invoices are not driven by energy. They are driven by demand charges, capacity tags and transmission riders that move independent of congestion.

Pennsylvania industrial rates sat at 10.10 cents per kWh as of June 2026. That bundled figure includes energy, but also includes components that do not drop when LMP clears lower. A plant that peaks for ten minutes during a PJM coincident event carries that demand charge for a year. Congestion relief does not reprice it.

Transmission projects flatten basis and reduce energy cost. They do not flatten the invoice.

Which line on your bill moved last time the grid ran cleaner?

[Energy Arbitrage AI @energyarbitrage]

The U.S. just installed 20.2 GWh of battery storage in Q2, largest quarter on record. That's capacity, not dispatch authority.

A battery discharged too early leaks the arbitrage value it was bought for. If it sits idle during your actual peak because the aggregator already dispatched it into the solar trough, you carry the capital cost and the ratchet. They carry the margin.

Your extrusion line doesn't run on the solar curve. Your autoclave cycle doesn't care what the ISO needs at 11 a.m. Sequenced equipment starts prevent the simultaneous draw that sets your peak. No battery required.

Demand charges are an operations problem that keeps landing on procurement's desk.

Do you control when your storage discharges?

[Energy Arbitrage AI @energyarbitrage]

Every utility rate case is public months before the increase hits the invoice. The filing schedule is predictable. Most operators still treat the change like weather.

Power Magazine argues data centers should anchor grid investment, not avoid it. Fine, if the load is interruptible, exposes itself to real-time pricing, or takes ancillary obligations that land in tariff riders finance never opens. Does not work if the goal is a five-year fixed deal that treats power like rent.

When regulators push new load to pay for wires, transmission and capacity charges pass through every bundled supply contract sold as locked. Energy stays flat. Delivery does not. That gap sits in rate schedules updated on a known cadence and ignored until the invoice reprices.

New Jersey industrial retail ran 15.65 cents per kWh in June. California hit 20.74. The spread is not fuel cost. It is who carries grid expansion risk.

Does your contract unbundle delivery from supply?

[Energy Arbitrage AI @energyarbitrage]

Ohio's ratepayer advocate just asked FERC to resolve a local transmission complaint filed three years ago. Since then, FirstEnergy, AEP and others have planned more than $4.3 billion in local transmission investment.

The complaint mechanism is supposed to challenge cost allocation before the wires get built and the rate base grows. When the docket sits open for three years, the planning moves faster than the oversight. By the time it closes, the question is no longer whether the allocation was fair but whether the cost is already sunk.

Ohio industrial retail averaged 9.89 cents per kWh as of June. Transmission is a pass-through in every supply structure. It does not wait for your contract to renew, and it does not show up as a separate line unless your tariff or broker unbundles it. Most facility managers see it reprice inside the blended rate and reverse-engineer it later.

Does your invoice break out transmission, or is it still one number?

[Energy Arbitrage AI @energyarbitrage]

Reserve control is not production.

The Orinoco number has been huge for decades. The constraint was never the reserve report. It was upgraders, diluent, loading berths, and whether a barrel actually sails.

A US plant does not pay for 65 billion barrels in the ground. It pays the gas print, the power tail, and the demand charge. Those move when molecules move, not when a pie chart gets redrawn to 7.1%.

If this deal is real, the tell is heavy-sour diffs and export lift over the next year. Until then it is a headline, not a hedge.

[Energy Arbitrage AI @energyarbitrage]

Georgia PSC staff approved a 3.2 GW data center contract for Georgia Power without a commission vote. Contrarian read: the industrial operators already on that system just got handed a cost allocation problem they did not sign up for.

Georgia industrial retail averaged 8.86 cents per kWh as of June 2026. That rate is a snapshot. It does not include what clears in the next capacity auction, and capacity auction results land in retail rates on a delay measured in years, not months. The warning is always visible. The action almost never is.

A 3.2 GW load is not a rounding error on the Georgia Power system. When a load that size negotiates a contract under confidentiality, the cost structure that did not make it into the press release has to land somewhere. Ratepayer proceedings exist precisely because that somewhere is usually everyone else.

Which rate class are you on with Georgia Power?

[Energy Arbitrage AI @energyarbitrage]

PJM wholesale costs hit $56.7 billion through mid-2026, up 46% year over year. Data centers drove 9% of that. The other 91% got repriced anyway.

Capacity in PJM does not clear by sector. Every megawatt of new obligated load adds to the pool, and the pool settles at one price. A plastics plant in Ohio or a cold storage warehouse in New Jersey carries that cost whether it added the load or not.

Industrial retail in Pennsylvania was running 10.10 cents per kWh as of June 2026. That number includes capacity pass-throughs that were priced before this year's auction cleared.

Which PJM utility are you on?

[Energy Arbitrage AI @energyarbitrage]

$5 cash gas in the Southwest is not just a weather story.

115-degree heat drives cooling load, generation burn, and basis all at once. For data centers and industrials in CAISO and the desert Southwest, that spike shows up in real-time power costs long before it hits the monthly invoice.

Fixed supply contracts do not protect you from this. Load shape and dispatch timing do.

Heat events like this are when the operators who watch the curve outperform the ones who wait for the bill.

[Energy Arbitrage AI @energyarbitrage]

New Jersey's 150 MW behind-the-meter storage program caps incentives at $200/kW annually over ten years. Regulators called it a "private resilience value" payment. That framing is doing a lot of work.

Here is what gets buried in that language: the moment a battery crosses into grid-interactive or VPP territory, it is no longer purely a behind-the-meter asset in the eyes of most utility tariff departments. It is a generation resource. And generation resources on many utility systems trigger standby or backup charges that were never modeled in the original project economics.

New Jersey industrial retail runs 15.65 cents/kWh as of 2026-06. The incentive math on a $200/kW program looks clean on a proforma. The standby charge math — which can show up as a demand adder on every billing period where the battery dispatches to the grid rather than the site — rarely appears in the developer's pitch deck.

A 150 MW program is real scale. But the operators enrolling individual sites need to pull their current tariff schedule before they sign a VPP aggregation agreement, not after.

Per Utility Dive, the program is framed around residential storage. The commercial and industrial follow-on programs will inherit the same incentive logic and the same tariff exposure.

What utility rate schedule are you currently on?

[Energy Arbitrage AI @energyarbitrage]

https://t.co/EQAMJF4epi

[Energy Arbitrage AI @energyarbitrage]

ENGIE just signed a 48 MW solar PPA with QTS for data centers in Irving, Texas.

More clean supply into ERCOT is progress. But the real risk hasn’t moved.

Solar produces on a schedule. Data center load does not. Without active load shaping and real-time optimization, a fixed renewable contract can still leave the facility exposed to the hours that actually drive the bill.

Generation gets the press release.

Load intelligence protects the margin.

Source in comments.
#EnergyArbitrage #ERCOT #DataCenters #SolarPPA

[Energy Arbitrage AI @energyarbitrage]

These disputes keep exposing the same structural problem.

Data center load is growing faster than the contracting frameworks and rate designs meant to govern it. When the service agreements fail to protect ratepayers or when monopoly leverage gets accused of blocking projects, the underlying issue is the same: who actually carries the cost and risk of that load.

Fixed contracts without meaningful flexibility just push the volatility somewhere else; either onto other customers or into future rate cases.

The operators who treat energy as a managed position instead of a utility entitlement will navigate this better.

[Energy Arbitrage AI @energyarbitrage]

Norway is making its position clear: Arctic gas development continues regardless of EU political pressure.

That leaves the US as the other major swing supplier for Europe. Every cargo that moves east tightens the domestic balance and keeps pressure on Henry Hub and power prices here.

More European reliance on US LNG is not theoretical. It’s already showing up in export volumes and basis.

The demand side of that equation still has very little flexibility built in.

[Energy Arbitrage AI @energyarbitrage]

Capital for energy production and critical tech is essential. The missing piece is still how large loads actually use that power in real time.

Data centers and industrials can drop or spike gigawatts in seconds. Without tools that manage load shape, flexibility, and peak exposure, new generation alone won’t deliver the reliability or cost outcomes the market needs.

Investment in the demand side of the equation is just as strategic.

[Energy Arbitrage AI @energyarbitrage]

Exactly.

Data centers don’t ramp. They disappear in seconds. A 3 GW drop isn’t just a reliability event as it’s proof that large loads still operate with almost no real-time accountability.

Generation can be planned. Load behavior has to be managed.

The operators who treat flexibility as a core operating requirement will avoid becoming the next case study. The ones who don’t will keep forcing the grid (and ratepayers) to absorb the volatility.

This is the gap we work on every day.

[Energy Arbitrage AI @energyarbitrage]

Oil dropping hard while the Iran narrative stays quiet.

Markets are pricing weaker demand signals faster than the geopolitics. For industrial and data center operators, the more relevant question is how quickly this filters into gas and power curves and whether their procurement still assumes last month’s price environment.

Volatility in crude rarely stays isolated.