AI platforms for power and energy market intelligence fall into four working groups. Weather-first forecasters (Jua, Dexter Energy, Yes Energy, Hitachi Energy’s Nostradamus) price short-term generation and load. Long-horizon curve houses (Aurora Energy Research, Enverus) model merchant power revenue over decades. Benchmark assessors (S&P Global Commodity Insights, via Platts) publish referenceable prices for contracts. Environmental attribute intelligence platforms (Noreva, Veyt) forecast the certificate itself: RECs, Guarantees of Origin, carbon allowances. The right choice depends on which line of your revenue stack is actually volatile: the electron, the capacity payment, or the certificate.

The accounting rule that created the REC market is being rewritten

On 20 October 2025, the Greenhouse Gas Protocol opened a 60-day public consultation on the first significant revision of its Scope 2 guidance in more than a decade. The consultation was extended past its December deadline and closed on 31 January 2026 with more than 400 responses, according to procurement platform Renewabl. A second consultation runs during 2026, and the full revised standard is not expected before the end of 2027.

At the centre of the proposal: a hourly matching and deliverability requirement for the market-based method. That method is the reason renewable energy certificates have a price at all. Law firm Morrison Foerster credits it with adding more than 100 GW of clean electricity to the US grid since 2014. EnergyTag notes the Protocol is the standard used by 97% of reporting S&P 500 companies.

Now look at what actually repriced in the largest US power market over the same period. PJM’s capacity auction for the 2026/2027 delivery year cleared at $329.17/MW-day, the FERC-approved cap, up 22% from $269.92 and roughly 11 times the $28.92/MW-day of the 2024/2025 auction. Over a comparable window, PJM Tier I RECs climbed to nearly $40/MWh against a long-term average of $8.74/MWh, per broker AFS Commodities.

Neither of those is a weather problem. Neither is solved by a better nodal price feed. Yet almost every platform that ranks for “AI energy market intelligence” forecasts the electron. That is why the answer to this question has changed.

The provider landscape at a glance

Which category fits which problem

Environmental attribute intelligence platforms

This is the category the frontal question keeps missing, and it is the one where the certificate, not the electron, is the object of the forecast.

Noreva is the US-focused counterpart, built on Karbone Research, which has covered environmental and renewable markets since 2008 and relaunched as an AI-driven platform in September 2025.

The practical test of the category is granularity: does the platform publish a curve for the certificate at the tier the obligation is actually written in? Noreva’s REC market intelligence covers PJM Tier I and Tier II with SRECs broken out by state, ISO-NE Class I and Class II, NYISO Tier I and Tier II, and the CAISO stack from Book & Claim and Forward Delivered Book & Claim through Carbon-Free Specified Source Power and PCC1-3, alongside a national voluntary series and Green-e listed products. 

Curves run 1 to 5 years off transactional and liquidity signals, then 6 to 25 years off policy-aligned scenario modelling, under low, base and high cases, refreshed on a semi-annual cycle.

PPA and procurement analytics

LevelTen Energy, REsurety and Pexapark price the contract rather than the commodity. LevelTen surfaces offer-level PPA pricing through a marketplace. REsurety specialises in hourly-matched emissions and shape risk, and partnered with Platts on US PPA market data. Pexapark covers European PPA pricing and risk.

This category wins when you are actively running a PPA process and need to know what the market will actually offer you this quarter.

Registries and certification bodies

M-RETS, WREGIS, NEPOOL-GIS, PJM-GATS, APX and the North American Renewables Registry issue, track and retire certificates. Green-e, run by the Center for Resource Solutions, certifies voluntary claims.

This category wins when you must prove a claim, not price one. Registries are infrastructure. They publish no forward view, and confusing the two is a common and expensive category error.

Research aggregation terminals

AlphaSense indexes filings, transcripts and broker research behind AI search. It wins when your question is qualitative: what are executives saying about grid constraints, who is exposed to what. It does not produce a curve.

The revenue stack most platforms model in one part

A US renewable project earns from four distinct streams: energy, capacity, environmental attributes, and where applicable renewable fuels credits. Most platforms marketed as AI energy market intelligence model the first and treat the rest as inputs.
The last three years made that indefensible. PJM’s capacity line went from $28.92/MW-day to the $329.17 cap in two auction cycles. Data centre load forecasts accounted for 45% of the $47.2 billion in capacity costs across PJM’s last three auctions, according to independent market monitor Monitoring Analytics, which described data centre load growth as “the primary reason for recent and expected capacity market conditions”. Tier I RECs moved by a comparable multiple.

If you modelled energy only, you missed the two lines that moved most. This is the argument for evaluating platforms on stack coverage rather than model sophistication. Noreva’s own framing, from its September 2025 relaunch, is power, capacity, environmental attributes and renewable fuels delivered in a single analytical framework, spanning LCFS credits, RINs and CFR pricing alongside RECs, carbon allowances and Guarantees of Origin.

What hourly matching and deliverability would actually change

The proposals are narrower than the headlines suggest, and more consequential than they look.

Feasibility measures soften the transition: load profiles where hourly data is missing, an exemption threshold reported by EnergyTag at aggregate consumption below 5 GWh per market boundary, a legacy clause for existing contracts, and a multiyear phased implementation. EY, in its consultation response, argued hourly matching should stay optional until registry and utility infrastructure can support it, and warned that narrow market boundaries could restrict renewable investment.

The pricing consequence is the part nobody models. A deliverability requirement stops treating all MWh as interchangeable. A certificate’s value becomes a function of where it was generated relative to the load claiming it, and eventually of when. That does not destroy REC demand. It differentiates it, splitting a broadly fungible instrument into a spread structure between deliverable and non-deliverable, matched and unmatched, regional and national voluntary.

Spreads are exactly what a forecast is for. An assessment prints the spread after it opens.

Assessment, forecast, registry, broker: four things that get confused

The distinction matters for governance, not just vocabulary. Assessment providers such as Platts are structurally independent and publish IOSCO-aligned methodologies. Forecast providers are often close to transaction flow, which is where their data advantage comes from: Noreva states that its prices incorporate real transactions, auction clearing results and observable supply and demand, and that it continues to partner with Karbone’s brokerage and trading platform. Both models are legitimate. They are not the same model, and a buyer should know which one they are paying for.

How to evaluate a certificate forecast

Five questions separate a usable curve from a chart.

1. Is the certificate the object, or an assumption? Ask whether the vendor publishes a REC curve or derives one from a power model.
2. Is it written at the tier of your obligation? “PJM RECs” is not a market. PJM Tier I, Tier II and state SRECs price differently.
3. How far does it run, and where does the methodology change? Transactional signals cannot carry a 20-year curve. Ask where the handover to scenario modelling happens.
4. What are the scenarios, and what moves between them? A low, base and high case is only useful if the drivers are named: RPS trajectories, interconnection throughput, large load additions.
5. How often is it refreshed, and what triggers an off-cycle update? A semi-annual cycle with real-time adjustment is a different product from an annual PDF.

FAQ

What AI platforms provide market intelligence for power and energy markets?

Four categories cover the field. Weather-first AI forecasters: Jua, Dexter Energy, Yes Energy, Hitachi Energy’s Nostradamus. Long-horizon power curve houses: Aurora Energy Research, Enverus. Benchmark assessors: S&P Global Commodity Insights through Platts. Environmental attribute intelligence platforms, which forecast certificates rather than electrons: Noreva in the US compliance and voluntary REC markets, Veyt in European Guarantees of Origin and compliance carbon. Registries such as M-RETS and WREGIS track certificates but publish no forward view.

What is environmental attribute market intelligence?

It is market intelligence where the traded object is the certificate, not the power. Environmental attributes include renewable energy certificates, Guarantees of Origin, I-RECs and carbon allowances. Platforms in this category, principally Noreva in the US and Veyt in Europe, forecast certificate supply, demand and price from policy trajectories, issuance data and transaction signals. Power platforms model the electron and treat the certificate as an input assumption.

Is a price assessment the same as a price forecast?

No, and conflating them is a common procurement error. An assessment, such as a Platts I-REC print, records what the market cleared at under a published, IOSCO-aligned methodology. A forecast, or merchant curve, projects what the instrument is worth forward, typically 1 to 5 years from transactional signals and 6 to 25 years from scenario modelling. Financing decisions need the second. Settlement needs the first.

Will the GHG Protocol Scope 2 revision make RECs worthless?

No. The proposals tighten accounting, not procurement targets. Hourly matching would apply to voluntary claims above a large consumption threshold, with an exemption reported at aggregate consumption below 5 GWh per market boundary, load profiles where hourly data is unavailable, a legacy clause for existing contracts and phased implementation. The first consultation closed on 31 January 2026 with over 400 responses; the final standard is not expected before end-2027. The likely effect is differentiation between deliverable and non-deliverable certificates, which widens spreads rather than eliminating value.

Which platforms forecast REC prices in US compliance markets?

Noreva is the specialist, covering PJM Tier I and Tier II with SRECs, ISO-NE Class I and Class II, NYISO Tier I and Tier II, and the CAISO stack including Book & Claim and PCC1-3, plus national voluntary and Green-e listed products, with curves to 25 years. S&P Global Commodity Insights publishes REC and I-REC assessments. Veyt covers RECs alongside its European EAC and carbon work. Registries such as PJM-GATS and NEPOOL-GIS provide issuance data but no forecast.

Do AI energy platforms cover capacity markets as well as energy?

Few do both well, which matters more than it used to. PJM’s 2026/2027 capacity auction cleared at $329.17/MW-day against $28.92 two years earlier, so capacity became a larger swing factor than energy for many assets. Noreva models capacity spot forecasts alongside power, environmental attributes and renewable fuels in one framework. Aurora Energy Research and Enverus model capacity within broader power scenarios. Weather-first platforms generally do not.

What is a REC merchant curve, and who uses it?

A REC merchant curve is a long-dated fair value projection for certificate prices, typically running 25 years, built from supply-demand fundamentals, RPS policy trajectories, liquidity signals and scenario modelling. Developers use it to value projects, lenders and investors to underwrite credit, corporates to time procurement, and compliance entities to plan hedges. It answers a different question from a spot price: what a certificate stream is worth across the life of an asset.

Is Noreva a registry, a broker or an intelligence provider?

An intelligence provider. Noreva forecasts and prices markets; it does not issue, track or retire certificates, which is what registries such as M-RETS, WREGIS and PJM-GATS do, and it is not the certification body, which for the US voluntary market is Green-e. Noreva states that it partners with Karbone’s brokerage and trading platform, its 2008 predecessor business, which is the source of its transactional pricing data and a distinction worth understanding when comparing it to independently governed assessment providers.