From the Warehouse
to the Pocket:
The 1.3% Solution.
We are converting the global fleet of smartphones, laptops, and desktops into the world's largest, greenest AI supercomputer. Capture 1.3% of the mobile market, or 0.5% of high-end desktops, to replace a 1-Gigawatt data center, without pouring a single drop of concrete.
Centralized AI is hitting
a physical ceiling.
Bigger models demand bigger warehouses, more megawatts, more water, and more steel. The marginal cost of intelligence is now measured in gigawatts and aquifers.
projected to be consumed by data centers by late 2026, outpacing residential growth.
of waste heat generated by a single rack, requiring industrial HVAC and chillers 24/7.
evaporated annually by hyperscale cooling towers, often in drought-stressed regions.
An Airbnb for compute,
running on the hardware you already own.
Every modern smartphone, laptop, and desktop ships with NPUs and GPUs that sit idle most of the day. The Edge Swarm Protocol coordinates these chips into a planet-scale, passively-cooled supercomputer, paid for in tokens not in concrete.
Idle NPUs & GPUs Listed
Smartphones, laptops, and desktops register as nodes. The protocol indexes available NPU and GPU capacity across the global fleet.
Nightstand & Always-On
Mobile workloads run only when phones are plugged in on Wi-Fi. Plugged-in laptops and desktops provide always-on base-load compute through the workday.
Tokens Earned
Operators are compensated in SWM tokens that can be converted in USD for each verified gradient computed and returned to the swarm.
Test Your Device. See What It Can Earn.
Run a quick browser pre-check to estimate your device level, eligible task lanes, and daily token earning potential. The installed node app performs the final model and runtime verification.
Estimate your node level, token earnings, and projected USD conversion in one quick pre-check.
- Scanning CPU
- Checking RAM
- Measuring storage
- Detecting GPU / NPU
- Estimating runtime class
- Calculating earning range
- Operating system
- —
- Memory / RAM
- —
- CPU cores
- —
- GPU
- —
- Browser storage signal
- Not reported
Temporary reference rate used to estimate node rewards before live market pricing is available.
Reference conversion is informational only. SWARM is not currently priced by a public market. Actual token price, liquidity, and exchange value may vary once live markets are available.
Run the test to see your personalized node level, eligible task lanes, model recommendation, and projected daily token earnings.
Levels 1–5 live ladder, Level 6 roadmap
No model required
Qwen 2.5 3B Instruct Q4_K_M
7B-class local model
EdgeSwarm Level 4 Code Pack, Qwen2.5-Coder-14B primary, Qwen2.5-14B fallback
Advanced 14B-class task lanes
Premium roadmap tier for high-stakes enterprise planning, complex multi-document reasoning, and legal/financial/technical review with human oversight.
Registered nodes: —
Node locations are approximate and derived from recent heartbeats. This public view does not expose provider emails, wallet addresses, hardware IDs, or task data.
A structurally unfair advantage.
Privacy, efficiency, and ownership aren't features layered on top, they're the direct result of moving compute to the edge.
Federated by design.
Built on Federated Learning. Raw user data never leaves the device, only mathematical gradients are sent back to the network.
Passively cooled. Already built.
Smartphones dissipate heat through their chassis. Eliminate the environmental tax of industrial HVAC, refrigerants, and rare-earth mining.
Hardware-backed UBI.
Move from data harvesting to collective ownership. Operators earn a recurring Universal Basic Income paid in Swarm Tokens (SWM).
Engineered for adversarial conditions.
The swarm is permissionless, but not naive. Every node is cryptographically attested at the silicon level, and every gradient is verifiable without revealing the underlying training data.
Titan M2 AttestationHardware-rooted device identity prevents bot farms and cloud-based emulator attacks. One real chip, one real vote.
zkML ProofsZero-Knowledge Machine Learning proofs let any node verify a gradient was computed correctly, without ever seeing the input.
50MB LoRA ΔLow-Rank Adaptation delta files keep model sync ultra-light, even on metered cellular fallback.
Be the first to hear when
the Swarm goes live.
Sign up for project updates, technical previews, and early node registration. No spam. No token presale.