Verdigris Technologies, Mountain View, CA - turning every electron into a data point since 2011.
The company teaching data centers to listen to their own electricity - and act before anything breaks.
Somewhere inside a hyperscale data center, a UPS rectifier is quietly failing. Its standard monitoring alarms say nothing is wrong. The maintenance team has no idea. And the cost of discovering the problem after the fact - in downtime, emergency repairs, and disrupted AI workloads - will be measured in the millions.
Verdigris knows it's failing. The sensors have been watching it for months, measuring its electrical waveform 8,000 times per second. The AI flagged the degradation 48 hours ago.
This is the problem Verdigris exists to solve: the gap between what buildings know about their own electricity and what they need to know. The lights work, the bills get paid, but the actual intelligence - the granular, real-time, machine-level truth of how power flows through a facility - is invisible. That invisibility is expensive.
"Your building is generating thousands of data points every second. You're not reading any of them."
The core premise behind Verdigris since 2011Verdigris is an AI-native electrical intelligence platform based in Mountain View, California. It makes high-frequency circuit-level electrical data readable, actionable, and predictive - for data centers, commercial buildings, industrial facilities, and anywhere else where electricity is both essential and opaque.
In 2011, Mark Chung returned from vacation to a shockingly high electricity bill. A Stanford-trained electrical engineer who had spent years at chip companies inside Silicon Valley's deepest technical core - NexGen (acquired by AMD), Pasemi (acquired by Apple), NetLogic (acquired by Broadcom for $3.7 billion) - he did what any reasonable engineer would do: he built his own Wi-Fi-enabled kilowatt meters and installed them in his home's electrical panels.
The results were clarifying. He could see exactly which appliances were using power, when, and how much. He could catch inefficiencies, predict behavior, and optimize in real time. The question that followed was obvious: why doesn't every building work like this?
The answer, he discovered, was that it's genuinely hard. Commercial electrical panels are complex, noisy, dangerous environments. Getting clean circuit-level data requires hardware that doesn't exist off the shelf, software that can make sense of thousands of signals simultaneously, and AI that can distinguish a malfunctioning compressor from normal load variation at 2 a.m.
Chung co-founded Verdigris with Jonathan Chu (CTO, Carnegie Mellon computer science, previously at NetLogic) and Thomas Chung (product, University of Texas) - two colleagues who had also spent time at NetLogic and understood both the technical depth required and the commercial opportunity waiting on the other side. They set up shop at NASA Ames Research Park inside Moffett Federal Airfield, a fact that is either a coincidence or a personality statement, depending on how you look at it.
"All three founders had previously worked together at NetLogic, before Broadcom paid $3.7 billion for the privilege. Verdigris is what they built next."
Background on the founding teamThe hardware is deliberately low-drama. Clamp-on sensors install inside electrical panels without requiring downtime or cutting wires. They start sampling immediately - 8,000 times per second per circuit - and transmit data via Wi-Fi or 4G/LTE to Verdigris's cloud platform. Installation typically takes a day. The results last for years.
The software is where things get interesting. The Building.AI platform ingests that stream of circuit-level data and turns it into a live picture of a facility's electrical health: which circuits are carrying what loads, where power quality is degrading, which equipment is running inefficiently, and - most valuably - which pieces of equipment are starting to fail.
The Verdigris Signals product takes this further. By analyzing the waveform signatures of individual machines at 8 kHz resolution, the platform can detect the distinctive electrical fingerprints of mechanical degradation in UPS rectifiers, motors, and compressors - often 48 hours to weeks before standard monitoring systems see anything at all.
Web dashboard, mobile alerts, and lobby displays. Real-time energy consumption, power quality metrics, and equipment health - from any browser, on any device.
Predictive reliability using 8 kHz waveform analysis. Detects UPS, motor, and compressor degradation 48 hours to weeks early. The alarm your alarm system doesn't have.
Clamp-on IoT sensors that install without downtime. 8,000 samples/second. Wi-Fi or 4G/LTE connectivity. No wiring changes required.
AI learns a building's usage patterns and automatically optimizes HVAC performance. No manual intervention - the system adjusts itself.
Virtual power plant (VPP) capability via Leap partnership. Commercial buildings participate in demand response programs and earn revenue from grid services.
RESTful APIs with OAuth 2.0 authentication. Integrate circuit-level electrical intelligence into any enterprise platform, BMS, or CMMS.
"At 8,000 samples per second, the sensor isn't just measuring energy - it's listening to the machine. A compressor has a sound. A failing rectifier has a different sound. Verdigris has learned to hear the difference."
On the 8 kHz waveform analysis behind Verdigris SignalsThe T-Mobile data center deployment is the clearest case study Verdigris has made public. Deployed across more than 800 UPS rectifiers, the platform detected 4% degradation in units that standard monitoring systems had flagged as fully operational. Four percent sounds small until you consider what a UPS failure inside a live data center actually costs.
At another facility - the company hasn't named it publicly - Verdigris identified $75 million in stranded GPU capacity. The servers were installed. The power was available. But electrical monitoring limitations meant operators couldn't confidently commission the capacity without risking cascading failures. Better data meant better decisions, and $75 million in hardware that had been sitting dark came back online.
Verizon, NVIDIA, Jabil Circuit, the W Hotel San Francisco, the Marriott Marquis in Washington, DC - the customer list runs from hyperscale tech infrastructure to hospitality. The thread connecting them is buildings with significant electrical loads and a need to understand what's happening inside them.
"$75 million in stranded GPU capacity recovered at a single facility. The fix wasn't more hardware. It was better data."
Verdigris data center case studyVerdigris has raised $55M+ across multiple rounds since 2016, with investors ranging from deep-tech VC to strategic corporates with direct operational stakes in the problem.
| Round | Amount | Date | Key Investors |
|---|---|---|---|
| Series A | $6M | Mar 2016 | DCVC, Jabil, Stanford StartX Fund |
| Series A-II | $6.7M | Oct 2016 | Jabil, Verizon Ventures, Stanford StartX |
| Series B | Undisclosed | Jul 2021 | Undisclosed |
| Series B Extension | $10M | Aug 2023 | DCVC, Solea Energy |
| Strategic Investment | Undisclosed | Mar 2026 | Southwire Company LLC |
Total disclosed funding: $55M+ · Lead investor DCVC has participated in multiple rounds, maintaining conviction through each phase of the company's evolution.
The energy challenge facing AI data centers isn't a hypothetical. A single AI training run can consume the electricity of thousands of homes. Hyperscale operators are commissioning gigawatts of new capacity - and discovering that building more power infrastructure doesn't help if you can't manage what you already have.
Verdigris's pitch to this moment is direct: you can't optimize what you can't measure. And the standard tools for measuring electrical systems - aggregate meters, monthly billing data, periodic audits - are nowhere near granular enough to manage the load complexity that AI workloads create. Circuit-level, real-time, waveform-resolution data is the difference between flying blind and flying with instruments.
The Southwire investment in March 2026 signals what the industry thinks about this. Southwire makes the wire and cable that carries power throughout buildings and infrastructure. Integrating Verdigris's intelligence into those products means the infrastructure itself becomes observable - every circuit, every panel, every feed carrying data about its own health in real time.
The patina on old copper turns green over time. Verdigris, the company, is betting that buildings in the age of AI will turn smart - that the invisible flow of electricity through walls and ceilings will become, finally, legible.
"The UPS rectifier that was quietly failing at the start of this story? It got replaced during a scheduled maintenance window. Nobody lost sleep. No GPU cluster went dark. The data did what data is supposed to do."
What Verdigris's electrical intelligence looks like in practice