Introduction to Nesa
The Layer-1 for trusted AI on-chain.
Nesa means “miracle”—a tribute to the golden age of AI we are entering. As Arthur C. Clarke famously wrote, “Any sufficiently advanced technology is indistinguishable from magic.” Trusted AI should feel the same: powerful, invisible, and reliable by design.
But today's AI infrastructure is none of those things. Centralized providers dominate the landscape, offering AI inference as black-box APIs with no privacy, no verifiability, and no user control. These systems are fragile, opaque, and exclusionary—especially for enterprises with sensitive data or compliance-critical use cases.
What Is Nesa?
Nesa is a privacy-preserving, verifiable, decentralized AI execution layer.
It allows users to run powerful AI models—language, vision, and beyond—across a globally distributed network of compute nodes, without trusting any single party. The models remain confidential, the data remains private, and the results are verifiable.
Built as a lightweight Layer-1, Nesa supports on-chain verification of off-chain inference using our innovative cryptographic primitives (see Major Innovations):
Equivariant Encryption (EE)
Homomorphic Secret Sharing over Encrypted Embeddings (HSS-EE)
Zero-knowledge proof systems for model correctness (zkDPS), published in KDD 2024 — The Fourth International Workshop on Smart Data for Blockchain and Distributed Ledger (SDBD'24)
Model-agnostic hybrid sharding (BSNS) for scalable distributed execution, published in KDD 2024 — The Fourth International Workshop on Smart Data for Blockchain and Distributed Ledger (SDBD'24)
MetaInf, a meta-learned inference scheduler that dynamically selects the best execution strategy per task and hardware profile, published in Conference on Language Modeling (COLM) 2025 (Main Conference)
The result is a fully decentralized AI network where computation is secure, trustless, scalable, and adaptively optimized—across diverse models and hardware.
How It Works
Nesa enables users to run AI models in a way that is private, verifiable, and decentralized—without trusting any single server or company.
The figure below illustrates how a query travels through the Nesa network:

Submit a Query A user or dApp submits an encrypted AI query through Nesa’s network interface.
Model is Split and Distributed Nesa breaks the AI model into smaller parts and assigns them to a group of nodes across the network—each node only sees a small piece.
Private Inference is Performed The network processes the query using encrypted data. No node sees the full input or the full model.
The Result is Verified and Returned The network generates a cryptographic proof that the result is correct—then returns the answer to the user.
What Makes Nesa Different
Private by default: No one can see your data, not even the nodes doing the computation.
Verifiable: Every result is backed by cryptographic proof.
Scalable: The workload is spread across a global network—no single node is a bottleneck.
Open: Anyone can contribute models, run nodes, or use the system.
This makes Nesa the first AI platform that is trustless, auditable, and built to scale securely, from chatbots to critical enterprise applications.
Nesa’s Infrastructure and Marketplace
Nesa introduces the first decentralized query marketplace for AI:
A public model store that supports both open-source and proprietary models
Support for LLMs, vision-language models, and other modalities
Token incentives for developers, model reviewers, validators, and queriers
Governance and access policies built into the protocol, not enforced off-chain

The architecture also allows nodes with limited compute to participate. Model sharding and cryptographic coordination reduce the dependency on top-tier GPUs, democratizing AI infrastructure beyond data centers.
What's Next?
This section covers several more core discussions:
Why Nesa Represents the Future of the AI Economy The case for a decentralized AI system that aligns economic incentives and access.
Why Centralized AI Infrastructure Is Broken An analysis of the limitations and risks of today’s closed AI platforms.
Why Existing DeAI Approaches Are Inadequate A critical look at why previous attempts at decentralized AI failed—and how Nesa improves upon them.
In addition, later sections will cover Nesa’s architecture and design innovations in depth, including cryptographic design, execution workflow, node structure, tokenomics, and model integration pipelines.
Nesa is not just a new AI infrastructure—it’s a new AI paradigm. Verifiable, trustless, and built for a world where computation must be both private and public, scalable and secure.
Last updated