NEAR Protocol

Founded 2017 by Illia Polosukhin (one of the eight co-authors of "Attention Is All You Need," the 2017 transformers paper) and Alex Skidanov (ex-OpenAI, MemSQL). NEAR began life in 2017 as Near.ai (an AI company), pivoted to a sharded L1 in 2018 to bootstrap funding, and in 2024-2026 has steered all the way back to its AI origin thesis.

Across decentralized-AI projects, NEAR has the single strongest founder AI credential by a wide margin, plus a verifiable substrate that already powers Venice's March 18, 2026 TEE/E2EE inference launch.

The L1 vs. NEAR AI Distinction

This matters and is the first thing to get right:

• NEAR Protocol (L1): sharded Nightshade-consensus blockchain, claimed 1M TPS benchmark (Q4 2025), <600ms finality. Two production-grade overlays — NEAR Intents (cross-chain execution abstraction, $13B cumulative volume, $77M/day average) and Near.com (consumer super-app launched Feb 2026, spanning 35+ chains).
• NEAR AI (near.ai): Polosukhin's separate-but-tightly-coupled AI initiative. The substantive AI stack including NEAR AI Cloud (TEE-based confidential GPU inference, Intel TDX + NVIDIA H100/H200), Private Chat, IronClaw (Rust agent runtime in encrypted enclaves, NEARCON 2026), Confidential GPU Marketplace, AI Agent Market, Confidential Intents, and as of May 2026 private USDC stablecoin payments for agents.

The L1 monetizes via gas fees + Intents fees; NEAR AI is the demand engine that creates Intents/agent activity and routes value back to $NEAR.

Founder Credibility (the load-bearing part)

Illia Polosukhin is one of eight co-authors of "Attention Is All You Need" (Vaswani et al., 2017) — the foundational transformers paper. Spent years at Google Brain on NLP before founding Near.ai. This is the single strongest AI credential of any crypto founder. Most "AI/ML" claims in crypto are stretched or non-existent; Polosukhin's is the inverse — he understated his crypto-AI credentials for years while building L1 infrastructure.

Alex Skidanov worked at OpenAI before ChatGPT's 2022 launch and architected NEAR's sharding. ML-engineering + distributed-systems crossover credentials.

This is verifiable, not narrative-painted.

AI Substance — Real, at the Infrastructure Layer

What NEAR AI is: the AWS of confidential agent compute. TEE-attested inference, agent runtime, payment rails, model storage, agent marketplace.

What NEAR AI is NOT: a frontier-model training lab. The Nov 2024 promise of a 1.4T-parameter frontier model has not delivered and appears quietly deprecated. NEAR is not in the model-training race.

Live shipped products with named customers:

• Venice.AI routes its TEE/E2EE inference through NEAR AI Cloud + Phala (see Venice §"AI Substance")
• Brave Nightly (privacy browser) using NEAR AI Cloud
• OpenMind robotics
• Phala as deeper-stack partner

The Phala + NEAR AI Cloud + Venice three-party stack is the cleanest TEE-private-inference architecture in crypto as of May 2026.

$NEAR Token — Real Value Capture

The token-AI link is mechanically real, not aspirational:

• Base gas: 70% burned
• Intents fees: 100% buy $NEAR on the open market (live since Feb 2025)
• Inflation cut from 5% → 2.5% (Oct 30, 2025)
• BitMEX research frames a 56x annualized fee multiple (cheaper than ETH on this metric)
• $177M/day Intents volume threshold would flip $NEAR net deflationary; current $77M/day puts the network ~half-way to that flip

As agent activity drives Intents volume, the reflexive buyback compounds. This is one of the cleanest "real revenue → token value" loops in crypto.

Scale and Traction (May 2026)

• 46M MAU (May 2025)
• 2,500+ monthly active devs (+40% YoY Q4 2024)
• $13B cumulative Intents volume
• >100M end-users reached via customer platforms (Brave, OpenMind, Phala)
• TVL compressed -50% to ~$160M (the bear-case datapoint)
• Visible "AI agent activity" mostly demos/launches; production volume not yet measurable

Recent State (last 6 months)

• Oct 30, 2025: inflation cut 5% → 2.5%
• Late 2025: NEAR AI Cloud live + 1M TPS benchmark
• Feb 2026: Near.com consumer super-app launch + NEARCON 2026 with IronClaw + Confidential GPU Marketplace + AI Agent Market announcements
• March 2026: Confidential Intents + Venice TEE integration via NEAR AI Cloud + Phala
• May 2026: private USDC payments for agents — direct fit for the "stablecoin micropayments for agents" category

Critiques

• Intents isn't uniquely AI-native: comparable to LI.FI/Across with an AI overlay. The cross-chain routing layer commoditizes.
• Wrong-layer-capture risk: substrate L1s historically commoditize; value accrues to apps (Venice) more than the layer beneath. If true, NEAR is the second-best Venice trade.
• 1.4T frontier model promise unfulfilled — credibility tax for stretching the AI-lab framing
• Token down vs. cycle highs despite shipping — market not pricing in the AI thesis yet
• Agent-economy adoption risk: Gartner-pilot-graveyard problem; could stay demos forever

Competitive Slot

NEAR is a substrate L1, not a peer to most decentralized-AI projects. Correct relationships:

• Complement to Venice — partner via TEE stack, not competitor
• Complement to Phala — NEAR wraps Phala's TEE primitive into a full stack with tokenomics + agent runtime + payments
• Different layer from Bittensor — NEAR does not produce or coordinate models
• Competitor to Solana for "agent commerce L1" mindshare — Solana has Pay.sh distribution; NEAR has stronger privacy primitives and Polosukhin's AI brand

Real-AI-Company Viability

NEAR is the upstream infrastructure pick on the same Venice trade — strongest AI credentials in crypto, real shipped product (not vapor). The honest bear case: L1 substrates commoditize and value accrues to apps.