Nyquis's $6.3M Bet That Relays Can't See the Fault That Starts the Fire

A branch leaning on a conductor draws less than one amp. The relay protecting that line is set to trip at hundreds.

On a windstorm night in southern New Mexico, a substation sensor built by Nyquis logged more than 50 high-impedance fault events on Otero County Electric Cooperative's lines, a pattern the co-op's existing protection could not see. It narrowed the events to four locations. Crews repaired three, and the signals there went quiet. The fourth kept alerting.

Nyquis closed a $6.3 million seed round on 5 November 2026 to put that box on more circuits, with Gigascale Capital and South Park Commons co-leading. The pitch is narrow and testable: distribution lines run close to blind, and a continuous waveform record at the substation is the cheapest way to open that window.

Our read is less tidy. Nyquis has the better vantage point on the physics and the worse position in procurement. Itron, which already sits inside the utility's metering estate, has the reverse. Procurement rewards the vendor whose software is already wired into the control room, and Itron is deploying fault analytics at a scale Nyquis cannot match this year.

The incumbent sells distribution, not detection

Itron's fault-prevention product does not arrive as a new box on a pole. It runs on endpoints the utility may already own. With PG&E, the company is deploying roughly 3,000 grid-edge devices in high fire-risk territory as a first phase, about 10,000 through 2027, and hundreds of thousands by 2030. Part of the work is funded through PG&E's Electric Program Investment Charge research budget.

The technical claim is narrower than the marketing. Itron captures waveforms at the endpoint, moves messages peer to peer and runs edge analytics to locate a fault, classify it and rank it for crew dispatch. Classification is the part that matters. A control room on a red-flag night does not need one more alarm; it needs an ordered list it can defend to a regulator after the fact.

“These fault indicators provide critical real-time inputs to drive faster crew dispatch upon detection, along with directed location information that reduces patrol time for crews.”— Don Reeves, senior vice president of Outcomes, Itron

Itron's advantage is distribution in the plainest sense. The endpoints are installed, the analytics already feed outage and distribution management systems, and a utility can trial the product without opening a new vendor file. That is a boring moat and a durable one.

Gridware shows a third path. The pole-mounted vendor raised $26.4 million in January 2025 and $55 million in November 2025, and reports tens of thousands of devices installed, according to Kurrant. Pole sensors localise a fault precisely because they sit near it. They also require hardware along the whole feeder, which turns a sensing decision into a capex programme.

One rack can watch a whole circuit

The Nyquis Node does the opposite. One unit per circuit bolts into an existing substation rack, taps the voltage and current transformers already there and records time-synchronised waveforms at about 10 kHz. It processes on the device, reports in under 800 milliseconds and installs in about 15 minutes without a shutdown or a change to protection settings. The models were trained on more than 200,000 simulated faults and are tuned to catch events below one amp.

That design has one obvious economic virtue. A substation sits upstream of every transformer on the circuit, so one rack covers what might otherwise take hundreds of pole devices. Nyquis says it now monitors about 800 miles of line. Dealroom counts two customers, both in New Mexico: Otero County Electric Cooperative and Socorro Electric Cooperative. Otero alone serves more than 16,000 homes and businesses across four counties.

South Park Commons, which co-led the round, says the New Mexico deployments surfaced four potential ignition events in a single four-month period. That number is the commercial argument in miniature. A utility that can point to four near-misses has a budget line; a utility that can only describe a capability does not.

The same design carries a real limit. A pole sensor measures at the point of contact. A substation device measures the circuit's aggregate behaviour and then infers where along the line the event occurred. Inference is cheaper to deploy and easier to get wrong on a long, noisy feeder with several branches of similar impedance.

What the Otero result does and does not prove

The cleanest part of the pilot is the control loop. Crews fixed three sites and the alerts stopped; the fourth stayed live. That is a falsifiable test, and it is more than most grid-sensor pilots publish. The caveat is scale: one co-op, one windstorm, a territory of 16,000 customers and a result reported through the company and its trade press.

Three sensing models now compete for the same wildfire line item, and they differ mainly in where the measurement happens.

Assessment based on company releases and Kurrant reporting, November 2026

The binding constraint is not sensitivity. It is alert triage. A detector that flags sub-ampere events on a quiet rural feeder can afford to be generous. The same threshold on a heavily loaded urban circuit, with capacitive coupling, switching transients and third-party construction, produces a different false-positive bill. Nyquis has not published a large investor-owned deployment where that threshold is tested in public.

PG&E's numbers explain why the budget exists anyway. Its 2026–2028 plan carries $18.88 billion in costs and about 1,100 miles of undergrounding, a programme measured in years. Monitoring buys time on the overhead miles that will stay energised through several more fire seasons.

The second product is where the story turns. Utilities plan connections on fixed worst-case limits. Google's Tyler Norris told the BNEF Houston summit this month that grid utilisation sits between 40% and 60%, and that non-firm service is the real whitespace. Aligned Data Centers and Portland General Electric already built the template: a 31 MW/62 MWh battery for a 72 MW data centre in Hillsboro, Oregon, which let the developer skip a fully firmed network upgrade.

As we wrote in November, the interconnection queue is a waiting room with no operator. Nyquis wants the same live waveform data to price the headroom on an existing feeder. That is a bigger market and a harder sale. A fault alert has a liability budget behind it; a capacity estimate has to beat an engineering study and a spreadsheet.

The test after the pilot

The round size says less than the customer list. $6.3 million is enough to build a second-generation sensor and train a larger fault set. It is not enough to carry a sales cycle through a large investor-owned utility, which is the only scale that would make substation-level sensing a category rather than a co-op story.

So watch three things. Whether a second and third utility signs after Otero. Whether the sub-one-amp threshold holds on a feeder with real industrial noise. Whether Itron's endpoint density makes a dedicated per-circuit box redundant before the market agrees it was necessary.

Anyone can record a waveform. The hard part is a procurement officer deciding that a 15-minute install is worth a new vendor file.