On August 3, 2026 Texas Governor Greg Abbott directed the Public Utility Commission of Texas and grid operator ERCOT to conduct a “comprehensive verification and audit” of every data-center project in the interconnection queue before any further project advances, and said non-compliant projects should be denied grid access.

Requests to connect large loads to the Texas grid have grown from about 48 gigawatts in 2023 to more than 474 gigawatts, roughly 90% of them data centers; ERCOT’s all-time peak demand is under 90 gigawatts. The audit requires disclosure of each project’s ultimate owner, any taxpayer incentives claimed, water use, and on-site generation plans.

ERCOT missed its August 7 deadline for the “Batch Zero” interconnection study as a result and has said it aims to finish the audit by December 2026, in time to feed its next long-term load forecast. Texas will audit up to about 300 projects.

A Reuters review published September 1, 2026 found that large-load requests across the Midwest, Mid-Atlantic and South exceed 700 gigawatts — more than ten times industry estimates of current US data-center consumption — and documented what happens when scrutiny arrives: Exelon cut its “high-probability” data-center pipeline by about 40% to 11 gigawatts after imposing stricter collateral requirements (investor presentation, July 30, 2026), and AEP Ohio’s pipeline fell by more than half after Ohio introduced connection-study fees of up to $100,000.

Pennsylvania Governor Josh Shapiro signed a comparable executive order on August 18, 2026; of more than 100 data centers proposed in that state, only about 20 have applied for permits and most have neither a power source nor a customer. Oncor, the largest Texas transmission utility, is working through a pipeline of nearly 300 gigawatts; CenterPoint says about 14 gigawatts on its Houston system is likely eligible for Batch Zero. PJM’s market monitor attributes a $29.4 billion increase in capacity costs over roughly the past four auctions to existing and forecast data-center demand.

Opportunity

The obvious reading is “AI power demand is a bubble,” which is a debate rather than a mechanism. The mechanism is narrower: interconnection-queue numbers are the raw input to three things that carry real money — utility capital plans and the rate-base growth stories built on them, the demand curves that set prices in capacity auctions, and the order books of turbine makers and gas-pipeline developers.

Texas is the first major hub to force the raw input to be audited on a deadline, and the two precedents where scrutiny was applied cut the number by 40-50% before any project was actually cancelled.

Hypothesis: ERCOT’s December 2026 long-term load forecast could be the first official downward revision of AI power demand in a major market, and the companies most exposed are not the hyperscalers but whoever’s valuation leans hardest on the un-audited number — Texas wires utilities, independent power producers priced on data-center contracts and capacity prices, and equipment makers whose backlogs include speculative developers.

The same audit cuts the other way for projects that clear it: a hyperscaler-backed site with verified ownership and a grid connection becomes scarcer, not cheaper, and sites that already hold interconnection agreements gain value while new ones are frozen.

How it could play out

Texas publishes audit results around December 2026 and a large share of the 474 gigawatts is reclassified as duplicative, unfunded or anonymous. ERCOT’s long-term forecast is revised down; utilities that had guided to pipelines in the hundreds of gigawatts restate them in tens. Other states copy the disclosure-and-deposit model (Pennsylvania and Ohio already have), and PJM’s next capacity auction clears against a smaller forecast. Meanwhile the surviving, verified projects in Texas hold a scarcer asset — an approved grid connection — and developers with frozen requests turn to on-site generation to avoid the queue entirely.

Questions worth asking

• The deciding question: what fraction of ERCOT’s 474 gigawatts survives an ownership-disclosure and deposit test? Exelon lost 40% and AEP Ohio more than 50% to similar rules — is that the base case for Texas, or is the Texas queue worse because it grew fastest?
• Which listed companies’ guidance and valuations embed the un-audited numbers — Oncor’s ~300 gigawatt pipeline inside Sempra, CenterPoint’s Houston plans, and the data-center premium in Vistra, NRG and Constellation?
• How much of the gas-turbine backlog at GE Vernova and Siemens Energy is contracted by speculative developers rather than hyperscalers or utilities, and what are the cancellation terms?
• If collateral rules deflate forecast load in PJM, does the next capacity auction clear materially lower — and how much of the independent-power-producer rally rests on capacity prices staying high?
• Who already holds approved interconnection agreements in Texas — converted crypto-mining sites, existing power plants with spare capacity — and does a freeze on new approvals reprice those assets upward?

Where to look

• Sempra (Oncor) — the largest Texas wires utility, carrying a ~300 gigawatt data-center pipeline through the audit
• CenterPoint Energy — Houston wires, about 14 gigawatts eligible for the paused Batch Zero process
• Vistra, NRG Energy, Constellation Energy — independent power producers whose premium rests on data-center contracts and capacity prices
• GE Vernova, Siemens Energy — gas-turbine backlogs whose customer mix (hyperscaler vs. speculative developer) matters
• Exelon, AEP — utilities that already cut their pipelines and may be through the reset
• IREN, Cipher Mining, Core Scientific — holders of existing interconnected Texas sites that become scarcer if new approvals stay frozen
• Caterpillar, Bloom Energy — on-site generation for developers routing around the queue

Thesis check

The strength is that this rests on primary documents — the Governor’s order, ERCOT’s own statements, utility investor presentations — with a dated catalyst in ERCOT’s December 2026 audit and two real precedents (Exelon, AEP Ohio) where scrutiny cut pipelines roughly in half.

The weakness is that even a halved queue may still exceed what can be built: PJM’s own position is that verified load is still arriving faster than generation, so deflating the forecast may not change the binding scarcity, and the direction of the trade — losers from a smaller forecast versus winners from scarcer approvals — is unresolved until the audit reports.

Timing

Audit results targeted for December 2026

Sources

Reuters via Yahoo Finance, Sep 1 2026 · Holland & Knight, Aug 2026 · Texas Tribune, Aug 14 2026 · Gibson Dunn, Aug 2026 · Houston Public Media, Aug 3 2026

Indonesia, which with Malaysia produces roughly 90% of the world’s palm oil, implemented its B50 biodiesel mandate on July 1, 2026, raising the palm-based share of diesel from 40% to 50%. The plan had been shelved in January 2026 and was revived in March 2026 specifically to reduce fuel imports after the US-Israel war with Iran disrupted oil supply.

Industry estimates put the additional domestic palm oil absorbed by B50 at about 2.2 million tonnes a year (taking biodiesel use to roughly 13.6 million tonnes) and the potential reduction in exports at about 3 million tonnes a year if production does not rise. The mandate is subsidised from a levy on palm oil exports; the plantation fund collected 17.4 trillion rupiah in January-May 2026, 64% of its full-year target.

On the supply side, Indonesia’s weather agency expected 71.6% of the country to receive low rainfall in August 2026 and more than 77% in September; NASA satellite imagery on September 1, 2026 showed peatland fires darkening skies across drought-hit Indonesia.

The USDA has cut its forecast for Indonesian palm oil production in 2026/27 from 48 million to 47.2 million tonnes on drought risk, and Indonesian industry bodies have warned output could fall by 1-2 million tonnes versus 2025, with fertiliser about 30% more expensive because of the Middle East war. Purdue University’s agricultural economists note that El Niño yield damage to oil palm builds over a 6-24 month lag, so the consequences of the June-September 2026 stress window will not fully appear until 2027.

Malaysian benchmark futures reached RM4,977 per tonne on August 20, 2026, the highest since December 2024; analysts quoted by Jakarta Globe see a path to $1,500 per tonne. Against that, Malaysian palm oil stocks at the end of July 2026 were at a five-month high, and rice prices at Southeast Asian export hubs rose about 15% in a month on the same drought.

Opportunity

The obvious reading — palm oil up on El Niño — treats this as a weather trade, which is why futures already sit at a 20-month high. That reading is incomplete for three reasons. First, the demand side is not weather, it is policy locked to an oil shock: B50 exists to substitute imported diesel, so higher crude makes the mandate stickier rather than weaker, and Indonesia has effectively converted about 3 million tonnes of annual export supply into domestic fuel.

Second, the supply damage is lagged, so the production numbers that will actually tighten the market belong to 2027 and are not yet in any monthly data. Third, palm is the swing vegetable oil; a shortfall transmits into soybean, sunflower and rapeseed oil prices and into biofuel feedstock costs in the US and Europe.

Hypothesis: a structural export withdrawal arriving at the same time as a lagged yield hit could tighten global vegetable oils into 2027 by more than a 20-month-high price implies, with the winners being producers that get the price without the mandate obligation (Malaysian planters, soybean crushers) and the losers being consumer companies and importing countries for which palm is a top-three input cost. The single biggest risk is political: at some price Indonesia’s subsidy arithmetic breaks and the government either raises the export levy (bearish for exports, bullish for price) or trims the mandate (bearish for price).

How it could play out

The dry season peaks in September 2026 and monthly Malaysian and Indonesian production data through early 2027 start to show the yield loss. Indonesian exports fall as B50 absorbs supply, importers in India, China and Europe bid for Malaysian cargoes, and the price spread to soybean oil closes or inverts, pulling soybean and sunflower oil up with it. Consumer-goods companies flag vegetable-oil cost inflation in 2027 guidance. Indonesia faces a choice between a higher export levy and a mandate cut; either decision becomes the catalyst.

Questions worth asking

• The deciding question: at what crude-palm-oil price does Indonesia’s biodiesel subsidy fund run dry, and what does the government do first — raise the export levy or cut the blend? The answer sets the direction of the whole idea.
• Where in the monthly data will 2026 drought damage first appear, and are listed planters’ 2027 production guidance figures still built on trend yields?
• Which substitute oil is cheapest relative to palm right now — US soybean oil under the biofuel volume mandates, Argentine and Brazilian soy crush, or Black Sea sunflower — and who owns that capacity?
• Which consumer companies have palm oil as a top-three input and hedge the least — Indonesian noodle makers, Indian fast-moving consumer goods, global staples?
• Do Malaysian planters, which receive the price without an Indonesian-style mandate, trade at a discount that doesn’t reflect that asymmetry?

Where to look

• Kuala Lumpur Kepong, IOI Corporation, SD Guthrie — Malaysian planters that receive the higher price without carrying Indonesia’s domestic mandate
• Wilmar International, Golden Agri-Resources — Singapore-listed integrated palm processors with Indonesian and Malaysian supply
• Astra Agro Lestari — Indonesian planter exposed to both the mandate and the drought
• Bunge, Archer-Daniels-Midland — soybean crush and vegetable-oil trading that benefits from substitution
• CME soybean oil and Bursa Malaysia crude palm oil futures — the direct commodity expression
• Unilever, Nestlé, Indofood — palm oil as a major input cost on the losing side

Thesis check

The strength is that the demand change is a dated government mandate with arithmetic export consequences, the supply cut has already been made by the USDA, and the weather signal is a very strong El Niño that forecasters put above 90% probability — three independent forces pushing one commodity the same way.

The weakness is that seasonal forecasts are unreliable about regional rainfall — a decent monsoon in Sumatra and Kalimantan would leave only the mandate, which the market has already seen — Malaysian stocks were at a five-month high at the end of July 2026, and Indonesia can change the blend rate by decree at any time.

Timing

Dry season peaks Sept 2026; yield impact 2027

Sources

Hydrocarbon Processing, Jun 2026 · Palm Oil Magazine, Aug 21 2026 · Jakarta Globe, Aug 2026 · UkrAgroConsult (USDA cut), Aug 2026 · Purdue Center for Commercial Agriculture, 2026 · NASA Earth Observatory, Sep 1 2026