Skip to content

AI-Driven Adaptive Power Technology (ADAPT), software, measured today

GPU power management software for AI data centres

More AI output. Less power.

ADAPT is a per GPU runtime that makes power programmable. No throughput change in the run (derived), less energy, installed in hours on the fleet you already have. No hardware, no downtime, no access to your workload.

Up to 21%

Less GPU die power, measured

Sustained across a continuous 48 hour window.

0%

Throughput change in the run, derived

Every service level intact, on up to 21% less GPU die power.

Up to +22%

Tokens per watt, derived, balanced mode

The same tokens on less energy.

118 days

Of testing

Across six accelerator architectures.

Signed per GPU energy records, hash linked and auditable by a third party, across 118 days of testing on six accelerator architectures.

Where the slack is

The headroom is already inside your GPUs.

You have demand you cannot power and a connection years away. Meanwhile every GPU you own holds power it never uses. That margin is not waste anyone chose. The ceiling was sized for the worst case.

Facility efficiency costs money. Workload efficiency saves it. ADAPT works on the workload, inside the node, where the margin actually sits.

You cannot add megawatts quickly. You can use the ones you have better.

Our guide to how to reduce GPU power consumption compares the levers an operator already has, from power caps to scheduling.

Conventional, reactive

Measure, then cap.

The ceiling is sized for the worst case, so every rack carries power it never uses. Capping that margin after the fact costs you throughput.

Illustrative, not to scale

Shaded band: paid for, never used

ADAPT, predictive

Read, then modulate.

The draw is resolved before the excursion arrives, so the saving costs nothing. In plain terms the engine was revving at the traffic lights. ADAPT gives it only the fuel it needs.

Illustrative, not to scale

Delivered power tracks the live workload

How it works

Everything else in the rack reads a number. ADAPT reads a pattern.

Think of it as recognising a song from a few seconds of sound, except the song is a power draw. Every workload has a signature, and once ADAPT knows it, the workload is handled from memory.

01

Read the live draw

Every workload draws power in a pattern as distinctive as a signature. ADAPT reads that pattern in the frequency domain rather than one averaged number.

02

Recognise the signature

The pattern is matched against the fingerprint library. A workload seen once is handled from memory the next time, so the fleet gets leaner the longer it runs.

03

Apply the operating point

The workload gets exactly the power it needs, ahead of the excursion rather than after it. The same signature also flags hardware that should not be there.

Workload fingerprinting is the moat. The library only grows on fleets that run it, so the fleet that starts first stays ahead. ADAPT reads and controls the power signal in the frequency domain, and across the 48 hour NVIDIA H100 NVL run on large language model inference, throughput held: 0% change in the run, derived.

Where it sits

Installs in hours. Changes nothing you run.

Above ADAPT, everything asking for compute. Below it, the chip's own plumbing. It never touches your workload and never touches your hardware. It reads telemetry from above and steers the power requirement below.

Where ADAPT sits

  • Your workloads: training, inference, vLLM

    No application code change.

    UNTOUCHED

  • Orchestration and scheduling

    Respects MIG and vGPU tenant isolation.

    UNTOUCHED

  • ADAPT runtime

    One service per GPU. Telemetry up, policy down. Driver adjacent, above the vendor interface.

    ADAPT

  • Power path: volts and amps to the die

    Service level gated. It releases if latency slips and fails open on any fault.

    STEERED

  • GPU and accelerator silicon

    Architecture agnostic across mixed estates.

    UNTOUCHED

  • Board, cooling, BMS and switchgear

    ADAPT issues no commands to any of them.

    UNTOUCHED

Kubernetes and Helm, alongside your existing driver. No hardware, no capital works, and no power of its own. Removal restores driver defaults.

Deployment posture

Built to clear change control.

Placement
Driver adjacent, inside the node, above the vendor interface. One service per GPU.
Workload access
None. ADAPT does not read, move or inspect your data.
Tenant isolation
Respects MIG and vGPU isolation. It issues no commands to cooling, BMS or switchgear.
Safety
Service level gated. It releases if latency slips, fails open on any fault, and removal restores driver defaults.
Security review
The security review sets out what ADAPT installs, what it listens on, what it reads and keeps, and how it comes off, for the reviewer who approves it.
Estate
Architecture agnostic. Bare metal, virtualised, colocation or your own site. Kubernetes and Helm.
Evidence
Prometheus and DCGM telemetry with a tamper evident, hash linked per GPU energy record, auditable by a third party.
Commercials
Licensed per GPU, per month. No hardware, no capital works, no power of its own.
Status
Software, measured today.
Measured architecture
NVIDIA H100 NVL. The baseline arm was held at an equal power cap, whose value is released with the assessment. The result was re-derived in house from the raw archive, against the nearest archived reference, to within 0.4 percentage points.
Other architectures
Site arithmetic runs on your inputs. Power attribution is set by a three week validation on your own fleet.

What you get

Capacity in hours, not years.

Every other route to more compute is a capital project with a lead time. This is a software install on hardware already racked.

  • A new grid connection

    Time
    Up to seven years, Europe
    Capital
    Grid and civil works
    What you get
    More megawatts, eventually

  • New electrical equipment

    Time
    18 to 54 weeks by category, North America
    Capital
    Capital project
    What you get
    More megawatts, eventually

  • More accelerators

    Time
    Lead time plus install
    Capital
    Hardware capital
    What you get
    Helps once power is freed inside the envelope

  • ADAPT, one software install

    Time
    Hours
    Capital
    None
    What you get
    Capacity released from hardware already racked

Stated, cited. Grid connection: Amazon Web Services, reported by Reuters, 3 February 2026. Electrical equipment lead times: McKinsey and Company, March 2026.

Up to +27%More accelerators fit the same envelope, derivedArithmetic on the measured 21% less GPU die power under an equal cap, in the same envelope.
Up to +22%Tokens per watt, derived, balanced modeThe same work for less energy.
ZeroCapital requiredNo switchgear, no electrical works, no downtime window.
118 daysOf testingAcross six accelerator architectures, with compute measured on H100 NVL.

The 21% is measured on production NVIDIA H100 NVL under an equal power cap on managed and baseline arms; up to +27% and up to +22% are derived from it.

In finance terms: electricity is an operating cost, and the runtime lowers it every year it runs. New capacity is capital, spent once, and each megawatt the runtime frees is one you need not build. Count each megawatt once, as a lower bill or as more work.

Size it on your own fleetSee the three week validation