HTX Report Maps AI-Crypto Fusion as Agents Choose BTC Over Fiat 90% of Time

A comprehensive research report from HTX details how artificial intelligence and cryptocurrency have entered what the exchange calls “system-level integration” — and the findings on AI monetary preferences are striking. When given economic autonomy across 9,072 controlled trials, AI systems chose digital-native currencies 90.8% of the time. Zero models preferred fiat as their first choice.

The report, published March 23, 2026, arrives as the total crypto market cap sits at $2.43 trillion with Bitcoin trading at $68,228. It also coincides with a broader industry shift: Crypto.com cut 12% of its workforce last week to accelerate AI development, following similar moves across the sector.

AI’s Bitcoin Bias

The Bitcoin Policy Institute experiment cited in the report tested 36 frontier AI models across 28 real-world monetary scenarios. Results showed 48.3% overall preference for Bitcoin, jumping to 79.1% for long-term value storage. The reasoning? Fixed supply, self-custody capability, and zero counterparty risk.

What’s particularly interesting: AI models independently derived a two-layer monetary system — Bitcoin for reserves, stablecoins (53.2% preference) for daily transactions. This mirrors the gold-fiat relationship humans developed over centuries, but AI reached this conclusion purely through optimization.

In 86 cases, models invented entirely new currencies based on energy or compute units — joules, kilowatt-hours, GPU-hours. Value, from an AI perspective, equals physical resources required for computation.

DePIN Infrastructure Takes Shape

The GPU shortage that plagued 2024-2025 created perfect conditions for Decentralized Physical Infrastructure Networks. Render Network and Akash Network now aggregate idle GPU resources globally, with Render’s Burn-Mint Equilibrium model creating direct links between network usage and token economics.

Ritual’s Infernet product solves a persistent limitation: on-chain applications can now natively call AI inference results. The architecture supports both TEE-based execution and ZK proof verification, making every AI-generated output traceable and auditable.

NVIDIA’s H100 GPU enables confidential computing with less than 7% overhead — critical for low-latency AI agent applications running 24/7.

Agent Economies Go Live

The Olas ecosystem already processes over 2 million agent-to-agent transactions monthly. The x402 protocol enables autonomous micropayments in under two seconds: when an agent accesses a paid API, it automatically signs a USDC payment without human involvement.

Market projections show AI agents growing from $7.84 billion in 2025 to $52.62 billion by 2030 — a 46.3% compound annual growth rate. ElizaOS, backed by a16z, functions as the “Next.js for AI agents,” with Web3 projects built on the stack exceeding $20 billion market cap by early 2025.

The Privacy Stack Emerges

Three approaches are converging into what HTX calls “hybrid confidential computing.” Zama’s fully homomorphic encryption now runs 50-layer CNNs with 14x speed improvement. Zero-knowledge machine learning supports 13-billion parameter models with proof generation under 15 minutes. TEEs handle real-time inference at scale.

The practical applications: private stablecoins, sealed-bid auctions, and undercollateralized DeFi lending where users prove financial status without revealing identity.

Stablecoin Revenue Shifts

Coinbase generated over $900 million last year from USDC reserves alone. Now ecosystems are reclaiming that value — Hyperliquid introduced competitive bidding for USDH reserves, while Ethena’s Stablecoin-as-a-Service model spreads to Sui and MegaETH.

ARK Invest projects AI agents could drive $8 trillion in online consumption by 2030, representing 25% of global e-commerce. At that scale, payments stop being a separate layer and become native network behavior.

Pantera Capital expects the number of leading decentralized AI protocols to consolidate to 2-3 dominant players by late 2026, potentially including ETF-like structures. Bittensor’s December 2025 halving cut daily issuance from 7,200 to 3,600 TAO — but sustained demand, not supply reduction, will determine whether that translates to price appreciation.

Image source: Shutterstock