Blog

    Expert perspectives on energy management, AI, sustainability, and building operations

    GPU rack power trace at 120 kW with a single white dot showing what 1 Hz monitoring captured and red markers showing the sub-cycle peaks that fell between samples — the failure mode the article describes
    Featured

    When training jobs die and the GPU logs don't explain it

    Multi-day GPU training jobs sometimes collapse without a useful error in the logs. Most of the time the failure started in the electrical domain, not in the GPUs themselves, at timescales the observability stack never sampled.

    Apr 29, 20267 min read
    Read Article
    Comparison of panel-level aggregated monitoring versus circuit-level individual breaker monitoring
    Technology

    Circuit-Level vs Panel-Level Monitoring in Data Centers

    Panel-level monitoring shows total draw per panel. Circuit-level shows each breaker. The difference matters when you need to find which rectifier is degrading.

    Apr 17, 20267 min read
    Data center reliability engineering evolution from redundancy to intelligence
    Data Centers

    The Three Eras of Data Center Reliability Engineering

    Data center reliability evolved through three eras: redundancy, monitoring, and intelligence. Most facilities are still in era 2. Era 3 changes the economics.

    Apr 10, 20267 min read
    Automated fault detection and diagnostics architecture for data centers
    Industry Insights

    AFDD in the Data Center: Beyond Hardware Circuit Protection

    The $1.86B AFDD market focuses on hardware arc-fault interrupters. Data centers need the analytics layer: continuous fault detection at the circuit level.

    Apr 7, 20266 min read
    Power quality comparison between legacy data center loads and AI-era GPU compute
    Technology

    Power Quality in the AI Data Center: What Changed and What Needs to Change

    Power quality standards were written for motor loads, not GPU compute. The load profile changed. The measurement approach needs to change with it.

    Mar 31, 202610 min read
    Continuous electrical monitoring at 8 kHz versus 1 Hz polling
    Technical

    What 'Continuous' Actually Means in Electrical Monitoring

    Most monitoring systems poll once per second and call it continuous. Real continuous monitoring captures 8,000 samples per second. Here is what you miss at 1 Hz.

    Mar 24, 20268 min read
    Electrical signature analysis applied to data center infrastructure
    Technical

    Electrical Signature Analysis Comes to the Data Center

    Electrical signature analysis has protected industrial motors for decades. Verdigris applies the same physics to data center infrastructure at 8 kHz, continuously.

    Mar 17, 202611 min read
    Why Event-Triggered Waveform Capture Misses Degradation — Verdigris article
    Technical

    Why Event-Triggered Waveform Capture Misses Degradation

    Event-triggered waveform capture catches disturbances after they happen. Continuous 8 kHz capture catches degradation before it causes them.

    Mar 10, 20268 min read
    Silicon-to-Utility Power Intelligence: The AI Factory's Buffered Power Delivery Network
    Data Centers

    Bringing intelligence to the cable layer

    A Southwire + Verdigris perspective on power delivery in the AI era. Why the cable layer is the next frontier for observability and how we're building toward an intelligent interconnect.

    Mar 4, 20264 min read
    Comparison of three predictive maintenance approaches (schedule-based, condition-based, and physics-based early warning grounded in waveform physics) for AI data center infrastructure
    Data Centers

    Three Approaches to Predictive Maintenance in AI Data Centers

    Predictive maintenance covers three distinct methods: schedule-based, condition-based, and physics-based early warning grounded in waveform physics. Most vendors offer the first two. Here is how they compare.

    Mar 3, 20269 min read

    Three ways to start.

    Get a free rectifier health audit

    We deploy continuous waveform monitoring and flag the specific units at risk in your facility over 4 weeks. Free assessment, no hardware commitment.

    Request your free audit

    Detect degradation weeks before failure

    Continuous waveform analysis flags equipment degradation that threshold alarms structurally cannot, with runway to act before a breaker opens.

    Explore Signals

    How one operator found stranded capacity

    Circuit-level visibility revealed over 1 MW of recoverable capacity at a single facility, with roughly $100M projected across the portfolio.

    Read the case study