High-voltage transmission towers carrying lines across open country at dusk

Your energization date is a distribution.

Synthara prices the probability that power arrives — by when, at what cost, and with what confidence.

Send us one project Read the methodology

ERCOT & PJM live Cluster-level cascade simulation Calibration published quarterly

synthara · underwriting console cluster 2026-1 · ERCOT West

We have 400 MW under option at a greenfield site in ERCOT West, filed into cluster 2026-1. The developer is telling our investment committee Q3 2029. What are we actually holding?

@queue-graph Normalised 63 ISO and non-market utility filings into one schema. Cluster 2026-1 resolves to 84 active requests totalling 11.2 GW, with 9 already past their deposit deadline.

@attrition Scored withdrawal probability per cluster-mate against 11 years of archived queue snapshots — developer track record, site control, offtake status, deposit behaviour, study-milestone slippage.

@cascade Ran 10,000 correlated cluster futures. Network upgrade cost reallocated at each withdrawal; your share moves from $71M at P50 to $164M at P90 as neighbours exit.

@lead-time Priced main transformer, GSU and 345 kV breaker procurement as a binding constraint. Equipment — not the study queue — sets your P50 date on this site.

Synthara Score B− · energization P10 Aug 2029 / P50 Feb 2030 / P90 Nov 2031
Upgrade cost $71M P50 · $164M P90 · stated target Q3 2029 sits at the 11th percentile

The problem

Roughly two thousand gigawatts sit in interconnection queues. About one in ten will ever be built — and every project that walks away hands its upgrade cost to whoever is left standing.

2,060–2,600GW

Capacity in US interconnection queues — at or above the entire installed US grid

~10%

Of queued capacity that actually gets built. ISOs put ~20% under construction

2,100days

Average queue to commercial operation — up roughly 60% since 2017

40 / 24%

Conversion rates in ERCOT and comparable PJM projects

Under FERC Order 2023, requests arriving inside a 45-day window are studied as a cluster and the network upgrades are costed for the group. When a member withdraws, its share is reallocated across everyone remaining. Some withdraw because of the cost, which raises it again, which triggers more withdrawals. It is a cascade, it has a mathematical structure, and essentially nobody is simulating it.

We do not forecast the weather. We forecast a well-defined human decision.

A decade of queue history contains tens of thousands of projects that either energized, withdrew, or are still pending — each with observable features at every point in its life. That is a supervised learning problem with real labels, not a guess dressed as a model.

Request the methodology

Silhouetted transmission structure against a low sun

ingest · queue graph63 jurisdictions

14:02:11 ERCOT GIS report q2 parsed · 4,118 rows

14:02:19 PJM new services queue schema drift · remapped

14:02:26 MISO DPP cycle snapshot v412 archived

14:02:31 CAISO cluster 15 · 61 withdrawals captured

versioned since2015-03

Archive what the ISOs overwrite

Queue data is public but published as current state, never as history. Files get overwritten, statuses change silently, withdrawn projects vanish from the sheet. The longitudinal dataset cannot be assembled retroactively — only accumulated.

Steel transmission tower shot from below against dark sky

attrition · cluster 2026-184 requests

P(withdraw) > 0.631 of 84

top driverupgrade cost / MW

label depth11 yrs

Score the projects around you

For every request in every queue: the probability it withdraws and the distribution of when. Developer track record, upgrade cost relative to project economics, offtake, site control, deposit behaviour, study delays, procurement signals.

Low angle view of a transmission structure and conductors

cascade · monte carlo10,000 futures

your upgrade cost P10$44M

P50$71M

P90$164M

correlationon — exits raise exits

point estimate suppressed by policy

Simulate the cascade, not the average

Withdrawals are not independent events. A cost shock that pushes one project out raises the cost for everyone remaining. Modelling that correlation is what makes a single number not merely imprecise but structurally misleading.

What we believe

We publish probabilities rather than predictions, we publish the method, and we publish our own calibration record — including the calls we get wrong.

Everyone else is building a search engine for megawatts.

Siting platforms answer where could I build. Synthara answers whether it energizes, when, and at what cost — for the capital committee, and for the capital behind it. A tool gets replaced. A unit of measurement gets written into contracts.

Who is behind this

pathway search300 MW · Q3 2029

surplus interconnection rightsP50 Q1 2029

behind-the-meter gen + storageP50 Q3 2029

curtailable-load fast trackP50 Q2 2030

queue position acquisitionP50 Q4 2031

ranked on risk-adjusted cost and date

Search the whole option space

Surplus interconnection rights at underutilised or retiring plants, behind-the-meter configurations, co-location, phased energization, curtailable-load fast tracks. This is a portfolio optimisation problem and it is currently being solved in spreadsheets.

Transmission towers and wind turbines across open farmland

the synthara scoremonitored

stated target11th percentile

gradeB− · outlook stable

One number two counterparties can agree on

A published, methodologically transparent rating of energization risk for a named project at a named site — designed from day one to be quotable in a credit memo. The value is not accuracy alone. It is a common measurement.

Transmission tower silhouetted against a dark sky

portfolio surveillance2 alerts

Permian Basin 220 MWB+ → A−

Loudoun County 480 MWBB → B−

Hardin County 150 MWB stable

trigger: 3 cluster-mates withdrew · cost reallocated

trigger: GSU lead time +14 wks · P50 slipped

Watch the whole book, continuously

Energization risk is not a diligence snapshot. Scores are re-run as clusters move, studies land, tariffs change and equipment lead times shift, so a fund learns its exposure moved before the quarterly report does.

What you get

Four things a consultant's opinion cannot give you.

A versioned archive

Queue snapshots captured and preserved since 2015, across sixty-plus jurisdictions that change format without notice. Public data, privately archived — the asset that cannot be bought after the fact.

Correlated cascade simulation

Cluster cost reallocation modelled with the correlation structure intact, because withdrawals trigger withdrawals. The distribution of your upgrade cost, not a number that will move.

A published track record

Calibration and Brier score updated quarterly, misses included. Of the projects we assigned a 70% withdrawal probability, we report what fraction actually withdrew.

Equipment as a hard constraint

An executed interconnection agreement without a transformer, a GSU, switchgear or a turbine is not power. Lead times are modelled first-class, because for many sites they are now binding.

Founder

Sarvagya Singh

Founder · Synthara

Unpriced risks eventually get priced by somebody.

Synthara exists because of a pattern that everyone in this market has lived through and almost nobody has written down. A study result lands at an acceptable number. Capital gets committed against it. Then, over the following year, companies you have never met make decisions you had no way to anticipate, and the number you underwrote is gone.

The industry files that under bad luck. It is not bad luck. It is a measurable, simulable cascade with a known mechanism, and treating it as noise is the single most expensive habit in infrastructure right now. Synthara is being built to measure it in public, publish the method, and be graded on the record.

“I watched a project's upgrade cost triple in eleven months because four companies I had never met made decisions I had no way to anticipate. Everyone treated that as bad luck.”

founder@synthara.cloud LinkedIn

Trust & data

A ratings business that hides its record has no business.

Every input we use is public-source and provenance-tracked back to the filing it came from — ISO queue reports, study results, tariff dockets, interconnection agreements. What clients send us stays isolated to their tenant and never enters a shared model. Our rating framework is published, our methodology committee is separate from any advisory engagement, and named public calls go through legal review before they go out.

Public-source provenance Versioned snapshots Tenant isolation RBAC SOC 2 Type II — in progress No client data in training Rating / advisory separation Published methodology Quarterly calibration

Coverage · ERCOT, PJM live
Next · MISO, CAISO, SPP
Then · UK & Ireland, Nordics, Japan, India
Archive depth · jurisdictions × years, published

Become a design partner Rating framework

Get started

Send us one project. We'll return a distribution, not a date.

Send us one project Request the methodology