The thing being built
A federated network
The protocol defines a system for trusted edge compute anchored on HSM boundary guarantees — hardware you and a stranger can both rely on, because the sealed enclosure is the boundary (the hardware root of trust). On that trust it runs AI inference and ML protocols: many nodes reasoning together, verifiably — why it works, and how it maps onto regulation.
The direction is one thing: compute moving from the cloud out to the edge. Past fog — all the way to mist. Compute condensing right next to the person using it.
The model
Three revenue lines, kept lean
That's the whole of it. No sprawling price sheet, no rent-seeking layer in the middle. The network earns in exactly three places — and it draws its cut thin.
5% protocol revenue on your own hardware
When your node earns — serving inference, ML or trusted compute to the network — the protocol takes 5%. The other 95% is yours. You run the box; the network draws a thin slice for the trust it makes possible.
The AI / ML products Peeramid Labs builds
Peeramid earns its keep the same way you do — by building genuinely useful AI / ML products on the open protocol and charging for those products. We eat our own cooking, on the same rails everyone shares. Our revenue is the work of our own hands — never a cut of what the community makes.
€20 / month for gateway access
The Peeramid Labs gateway service — the trust-and-coordination layer that connects your node to the network and mints your agent credentials — is a flat €20 / month per seat. Predictable, the same for everyone, and never a cut of what you earn. It keeps the rails lit; your hardware's earnings stay yours.
Three lines. A thin cut on what you earn, the products we build on what we build together, and a flat €20/mo gateway that keeps the rails lit.
A familiar shape
A gig economy for trusted compute
Strip the jargon and it's a shape you already know. Like ride-hailing, or any marketplace — except the thing being matched is trusted compute. The protocol matchmakes compute seekers on the demand side with trusted hardware operators on the supply side, takes the same small 5% for making the match and running the rails, and routes the rest to whoever did the work.
And the payout isn't a handshake. Each trusted hardware operator's share is backed by the applicable operator agreement with Peeramid Labs — real, enforceable, not vibes. Peeramid is the matchmaker and the first operator, not a landlord.
First builder, not the owner
Toward a vendor-independent federation
Someone had to build the first node. Peeramid Labs is the first-of builder and operator of the protocol — but we are not meant to own it forever. The point was never to become the new landlord.
As other companies build their own compatible federated systems, those systems can be joined together with MPC cryptography into one vendor-independent network — no single company owns the federation. It's the easy-to-build trust boundary that makes this composable; see the founder's write-up on how we build TEEs the easy way →.
This strengthens the whole ecosystem — more operators, more resilience, no single point of control. And it changes what our cut means.
Once the federation forms, the 5% becomes dynamic — Peeramid's share tracks the hardware we make versus what other vendors make, not a fixed rent on everyone.
Why the hardware holds its value
Compute is a scarce, income-producing asset
Advanced-packaging and memory supply is booked out for years, DRAM contract prices roughly doubled in a single quarter (Q1 2026), and hyperscaler capex is running ~$725B in 2026. And AI chips don't obsolete like consumer goods — they cascade from training to inference over a 5–7 year life and hold a value floor. Owning a node means holding the scarce asset that earns, instead of renting it and funding someone else's balance sheet. The full case, with sources → the economic rationale.
How to invest
The best way in is to become part of the community
There's no token, and no round to wait for. You join. And the gate is honest and singular.
The one strict requirement is getting the hardware itself. Own a node — and you're in.
It doesn't take much. And we try our best to make hardware available to all types of users — small makers, individuals, businesses — despite surging RAM and GPU prices. The point is that ownership stays within reach, not reserved for whoever can build a datacenter. Already own a pro GPU? Trade it in — we credit up to 90% of its original unit price toward a node. Start at the node preorder.