NODELIFT.
Developers ↗

EXISTING HARDWARE. UNWRITTEN POSSIBILITIES.

More possibility. From what you already own.

AI asks for more. Your existing devices may still have more to give.

We are building a way to bring their memory and computing power into useful collaboration, with each owner in control.

Experimental pilot. Useful contribution depends on the device, connection and workload.

NodeLift three-dimensional cyan emblem
A LONGER LIFE FOR POSSIBILITY

01 / THE WORLD AROUND THE IDEA

Demand is accelerating. Potential is everywhere.

The next chapter of AI cannot be understood only by counting new machines. It also asks what we can do with the resources already in our hands.

01 $581.7B

2025 / GLOBAL CORPORATE AI INVESTMENT

Capital is moving into AI at extraordinary scale.

Historical investment across AI, not NodeLift revenue or addressable market.

Stanford AI Index 2026 ↗
02 5.8B

2026 REPORT / UNIQUE MOBILE SUBSCRIBERS

Connectivity reaches billions of people.

Subscribers are people, not qualified devices or available compute nodes.

GSMA Mobile Economy 2026 ↗
03 485 → 950

TWh / 2025 ESTIMATE → 2030 PROJECTION

The electricity question is getting bigger.

All data centres. The 2030 value is the IEA central projection, not an observed result.

IEA Key Questions on Energy and AI ↗
04 62 Mt

2022 / GLOBAL E-WASTE GENERATED

Useful life deserves more attention.

22.3% was documented as formally collected and recycled. This includes all e-waste, not just reusable devices.

ITU / UNITAR Global E-waste Monitor 2024 ↗

Market context, checked 17 November 2026. These are different populations and periods. They are not summed into a NodeLift market-size, energy-saving or available-RAM claim.

Sources and methodology ↗

02 / REIMAGINE THE RESOURCES AROUND YOU

One shared direction. Many ways to contribute.

A personal workspace, an office or a project site. The ambition is to make existing hardware part of the answer, while keeping participation understandable.

HIVE / REGIONAL HUBS / PEER DEVICES

ILLUSTRATIVE TARGET
HIVE Global coordination
AM Americas
Local peers
Remote peers
EU Europe
Local peers
Remote peers
AP Asia-Pacific
Local peers
Remote peers
ME Middle East / Africa
Local peers
Remote peers
Local connection Remote connection

Motion illustrates relationships, not live traffic.

Target architecture, not a map of deployed hubs. Regional federation and partitioning one model across arbitrary devices are not yet qualified.

FROM CONNECTED DEVICES TO USEFUL WORK

What comes back to you.

PILOT WORKFLOW

You bring a task. Device owners decide which resources can help. Hive connects the two within an authorized cluster.

  1. Your request

    Choose a supported model and submit a task from your authorized workspace.

  2. Hive coordinates

    It assigns work to an available, compatible worker within the allowed cluster and operating limits.

  3. The node executes

    The selected worker runs the complete model. Its memory belongs to that worker.

  4. You get the result

    The response returns to your workspace, with the available run and resource records.

The resources come from people and teams.

Create your account, enroll devices by QR code and choose when they may contribute. Availability and consent come before assignment.

Open your workspace →

FOR THE PERSON OR TEAM REQUESTING WORK

Access to your authorized devices. An answer you can use.

The pilot returns model responses through Hive. It does not grant access to someone else’s hardware or promise a faster run.

FOR THE DEVICE CONTRIBUTOR

Control, visibility and a record of the work.

You keep operating limits and can pause participation. Hive shows available activity and energy readings, separating measured, estimated and unavailable data.

A broader exchange of shared resources is the next question. Usage credits, contributor payments and service guarantees are not active in the pilot.

Animation explains the direction of work and results. It does not show live traffic. Regional federation and sharing one model across arbitrary devices remain research goals.

Your devices. Your choice.

Explore a personal cluster of computers and phones. Each device needs its own compatible runtime, available budget and permission to contribute.

03 / WHY THE PROBLEM IS STILL OPEN

The opportunity is real. The coordination is hard.

Distributed inference already has important successes. What remains difficult is useful, repeatable performance across everyday devices and changing networks.

  1. 01

    Memory has a location.

    Adding RAM figures does not make a model fit. Work and model data need a valid placement plan.

  2. 02

    Every transfer has a cost.

    Network waits, synchronization and discarded work can consume the time a second device was meant to save.

  3. 03

    Hardware speaks different languages.

    CPU, GPU, operating-system budgets and runtime formats need compatible execution paths.

  4. 04

    Peak speed is only a moment.

    Heat, battery, background limits and owner availability determine sustained contribution.

  5. 05

    Trust is part of the system.

    Data boundaries, consent, model rights and correct results must survive every handoff.

STANDING ON REAL PRIOR WORK

The next benchmark must be earned.

Apple demonstrates model partitioning and MLX scaling over Thunderbolt RDMA. llama.cpp RPC provides remote compute backends; exo connects supported peers for distributed inference; DistAI prototypes Android/PC layer pipelines. The open challenge is useful, repeatable performance across mixed everyday devices and changing networks.

Our evidence and open questions ↗

04 / THE BIGGER PICTURE

Six forces shaping what comes next.

PESTEL is our strategic reading of the opportunity: the external pressure, and the design question it creates. Open a factor to explore it.

P Political Who controls the infrastructure?

THE PRESSURE

AI capacity and digital sovereignty are strategic concerns.

OUR DESIGN QUESTION

Make location, authority and operating policy explicit. A planned region is not a deployed facility.

European Commission / AI Continent ↗
E Economic What is the cost of useful work?

THE PRESSURE

Investment is growing, but purchased capacity is only one part of total cost.

OUR DESIGN QUESTION

Compare useful completed work, transfer overhead and power before claiming a saving.

Stanford AI Index 2026 ↗
S Social Who gets to participate?

THE PRESSURE

Connectivity is widespread; access to suitable computing resources remains uneven.

OUR DESIGN QUESTION

Let owners understand, limit and stop participation. Connectivity alone does not qualify a device.

GSMA Mobile Economy 2026 ↗
T Technological Can coordination beat its overhead?

THE PRESSURE

Distributed runtimes work, but their results depend on hardware and interconnects.

OUR DESIGN QUESTION

Qualify each model, execution path and connection. Keep a strong single-device baseline.

Apple MLX research ↗
E Environmental Does reuse improve the whole balance?

THE PRESSURE

E-waste and electricity demand make hardware life an urgent design question.

OUR DESIGN QUESTION

Measure energy per useful task. An older device is not automatically the lower-impact option.

ITU / UNITAR Global E-waste Monitor ↗
L Legal What may a node process?

THE PRESSURE

Personal data, AI responsibilities and model licences create distinct obligations.

OUR DESIGN QUESTION

Assess workload and jurisdiction, define access boundaries and verify rights. Architecture alone is not compliance certification.

European Commission / AI Act ↗ GDPR / EUR-Lex ↗

Strategic interpretation, not legal advice, compliance certification or a claim that every proposed response is implemented.

05 / THE PERSON BEHIND THE PROJECT

Fabrizio Degni
FABRIZIO DEGNI / NODELIFT

Who I am. Why I am building this.

Fabrizio Degni

Creator of NodeLift. Independent builder.

I build tools that make complex digital systems easier to inspect and understand. I created and maintain PolicyWatcher, a public-interest project that connects changes in digital policies to their source evidence.

NodeLift begins with a personal conviction: useful hardware deserves a longer story. I want to explore how the resources we already own can widen access to AI, with evidence, understandable limits and people in control.

Built with respect for the researchers and open-source communities who made this exploration possible.

Acknowledgements and prior work ↗

THE NEXT CHAPTER STARTS WITH WHAT YOU HAVE

Give possibility a place to begin.

Join the invitation-only pilot in Hive. Explore the evidence on Developers. Help ask better questions about the hardware we already own.