For compute buyers
Governed compute capacity, priced at or below cloud on-demand.
HUMENS aggregates available compute capacity — idle CPU/GPU inside enterprises of any kind (a bank, a retailer, a back-office estate, a university) plus energy-backed sites — and makes it sellable to a serious buyer through a zero-trust governance layer: every node is attested, isolation is enforced fail-closed, every core-second is metered with evidence. You submit containerized jobs through an API; you never install anything.
Capacity
How much, and where
Why now
The idle-compute opportunity Industry estimate
The world is full of already-powered, already-paid-for compute that sits idle — because until now there was no way to safely sell time on a machine you don't own. These are third-party industry estimates of the addressable pool, not HUMENS-owned capacity.
Figures above are industry estimates of the total addressable idle pool — the market HUMENS unlocks. HUMENS's own live governed capacity is pilot-scale (shown above), not these numbers.
Supply footprint
A worldwide aggregation network Illustrative concentrations
That idle pool is everywhere (the lit landmass below). HUMENS aggregates it into governed capacity worldwide — across non-embargoed jurisdictions. The markers (US, Philippines, India) are illustrative concentrations, not the limit of coverage. Markers are region-level only; HUMENS never exposes an individual supplier's identity or a machine's location.
Pricing
FMV-anchored, market-tracking, floor-guarded
HUMENS does not deep-discount to win — it tracks the market. Your rate is anchored to the fair market value (FMV) of each lane (a rolling blend of major cloud on-demand rates), then quoted at or below that anchor and floor-guarded so it never drops below a sustainable level. The exact discount is a quote-time lever, negotiable by volume and term.
The reference anchor
Public cloud general-purpose on-demand rates we track (illustrative, sampled 2026-07-04). Your governed HUMENS rate is quoted at or below these, floor-guarded.
| Lane | Cloud on-demand anchor |
|---|---|
| CPU (per vCPU-hr) | ~$0.049 |
| GPU · basic (T4 / L4 class) | ~$1.00 |
| GPU · pro (A100 class) | ~$3.50 |
| GPU · elite (H100/H200 class) | ~$8.00 |
Anchor = rolling AWS/Azure on-demand blend, refreshed periodically. Not a HUMENS price; it's the market reference your governed rate is measured against.
How your rate is set
- At or below cloud on-demand — capped at the FMV anchor; you never pay above market.
- Floor-guarded — a hard floor keeps supply sustainable, so capacity doesn't evaporate.
- Market-tracking — the anchor refreshes with cloud list rates; pricing follows the market, not a stale sheet.
- Interruptible / spot-tier — aggregated idle capacity is priced as preemptible, not reserved.
- Volume-negotiable — the exact opening discount and committed-use terms are set in your evaluation.
The energy edge (behind-the-meter solar) is what lets HUMENS hold a competitive rate above the floor — power + governance is the value; price is the closer.
Trust
Why you can run on machines you don't own
Zero-trust supply
- Nothing is trusted by default. Every node enrolls with a signed token issued on contract, attests its hardware, and is admitted only after review.
- CHO-attested (CHOSE loop): Checked · Halted · Oversight · Secured · Evaluated — a human operator governs dispatch and can ground-stop the fleet.
- Isolation, enforced fail-closed: untrusted workloads run only on a node whose isolation boundary has passed independent acceptance — until a node is verified, HUMENS refuses the job rather than run it unsafely. The proven live path today runs first-party / trusted workloads.
Evidence, not promises
- Metered per workload: core-seconds, GPU-seconds, and wall-time are recorded per job with a result hash.
- Auditable settlement: billing and supplier payout trace to the same governed evidence (HUMENS Canonical Schema, HCS).
- Kill switch: a CHO HALT stops all new dispatch immediately; jobs stay queued until dispatch resumes.
Today HUMENS runs first-party / trusted workloads live. Open third-party isolation is enforced fail-closed and pending independent node acceptance — until a node's isolation boundary is verified, the platform refuses the job rather than run it unsafely.
How you integrate
You submit jobs. You don't install anything.
The node agent is a supplier tool. As a buyer, you send work through the marketplace job API with a HUMENS-issued buyer token — think AWS Batch or a Kubernetes Job. HUMENS schedules each job onto a governed, dispatch-eligible node, runs it under enforced isolation, meters it, and returns the result.
- Submit a containerized job (
POST /jobs) with an image, command, lane, and limits. - HUMENS schedules + governs — the job is leased to an approved node and runs under enforced isolation (fail-closed).
- Poll (
GET /jobs/{id}) for state and pull the result by reference when complete. - Billed on metered evidence — you pay for measured core/GPU/wall-time, nothing else.
Illustrative request
POST /jobs Authorization: Bearer <buyer-token> Idempotency-Key: <your-unique-key> { "image": "your-registry/workload:tag", "command": ["python", "run.py"], "lane": "vap", "gpu_required": 0, "wall_seconds_limit": 3600, "tenant_id": "your-tenant" } → { "job_id": "job_…", "state": "QUEUED" }
Shape is illustrative; the governed integration + SDK is set up during onboarding.
Energy edge
Behind the meter Roadmap
The structural cost edge is power. Beyond aggregating already-powered enterprise capacity (CaaS-VAP, no premium), HUMENS is building owned behind-the-meter energy sites (CaaS-VPP) for firm 24/7 renewable compute — solar for daytime generation and modular hydrogen for firm overnight power (an alternative to battery-only storage). The first solar proof site has its permit under review; the hydrogen approach is on the roadmap.
Evaluating governed capacity for your workloads?
Bring a workload profile; leave with a governed rate and an integration path.
Have idle compute to sell instead? Onboard your capacity →