By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panelists agreed that the infrastructure buildout (tokenized funds, custody, stablecoins) is real and bullish, but the 'autonomous AI agents as economic actors' thesis in the near future is speculative due to regulatory, liability, governance, and technical challenges, particularly the 'Oracle Problem' and settlement finality.
Risk: The 'Oracle Problem' for autonomous agents and the lack of clear liability for their actions.
Opportunity: The real and bullish infrastructure buildout by institutions like State Street, BNY Mellon, and Western Union.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
- State Street (<a href="https://247wallst.com/companies/STT/">NYSE:STT</a>) and Galaxy launched a tokenized liquidity sweep fund, while BNY Mellon (<a href="https://247wallst.com/companies/BK/">NYSE:BK</a>) is building institutional digital asset custody in Abu Dhabi, and Western Union (<a href="https://247wallst.com/companies/WU/">NYSE:WU</a>) launched USDPT, a regulated dollar-backed stablecoin on Solana.
- Andreessen Horowitz closed a $2.2B crypto fund betting that autonomous AI agents will operate as independent economic actors within five years, prompting major financial institutions to build on-chain infrastructure and LPs to transition capital onto blockchain networks.
- The analyst who called NVIDIA in 2010 just named his top 10 stocks and BNY Mellon wasn't one of them. Get them here FREE.
Andreessen Horowitz closed its fifth crypto fund with a thesis that sounds like science fiction: autonomous AI agents holding crypto wallets, earning income, paying for compute and operating as independent economic actors within five years.
The pitch described agents generating value "by creating services, by writing software, by creating content, by essentially becoming an employee for, for another agent or for a human." Some teams are already "building a harness that gives an agent the ability to pay for its own compute or whatever resources it needs to continue to exist." The caller acknowledged it "sounds insane," while arguing that "given the exponential curve on which AI is currently on, it is not a stretch to believe that in 5 years, the models will be so smart and so capable that actually going out and creating value in a capitalist society is something that they can very well do." A follow-up question: "Are they going to be working for us or are we going to be working for them?" The answer offered was "maybe a combination of the two."
The analyst who called NVIDIA in 2010 just named his top 10 stocks and BNY Mellon wasn't one of them. Get them here FREE.
During the $2.2 billion raise, LPs showed appetite. "everyone is very focused on AI these days." Crypto's defensibility: "you cannot vibe code USDC, you cannot have a weekend project and end up building something like Hyperliquid." Several LPs asked about funding capital calls in stablecoins and are "starting to transition some of their own existing business on-chain."
Recent deal flow supports the thesis. State Street (NYSE:STT) and Galaxy launched a tokenized liquidity sweep fund, with tokenized fund servicing capabilities planned from Luxembourg by end of 2026. BNY Mellon (NYSE:BK), Finstreet, and the ADI Foundation are building institutional digital asset custody in Abu Dhabi. Western Union (NYSE:WU) launched USDPT, a regulated dollar-backed stablecoin on Solana, partnering with Fireblocks.
Four leading AI models discuss this article
"The transition to AI-driven, autonomous M2M commerce necessitates a shift from legacy banking rails to programmable, on-chain settlement infrastructure."
The Andreessen Horowitz thesis on autonomous AI agents as economic actors isn't just about 'crypto'; it’s a bet on the infrastructure of machine-to-machine (M2M) commerce. While retail focuses on token prices, the real story is the institutional plumbing—State Street (STT) and BNY Mellon (BK) building custody and settlement layers. If AI agents become autonomous, they require non-custodial, 24/7 programmable settlement, which traditional banking rails cannot provide. This creates a massive TAM for stablecoin-based settlement layers like those being built on Solana. However, the regulatory friction is being severely downplayed; the SEC and global regulators will likely view autonomous, self-funding agents as a nightmare for KYC/AML compliance, potentially stalling adoption.
The legal liability of an autonomous agent remains undefined; if an agent commits fraud or violates sanctions, it is unclear who bears the criminal and financial burden, likely forcing a 'human-in-the-loop' requirement that kills the autonomy thesis.
"TradFi tokenization infrastructure is advancing steadily, but a16z's 5-year AI agent independence timeline hinges on unproven leaps in reliability and regulation."
a16z's $2.2B crypto fundraise signals strong LP conviction in AI-crypto fusion, with TradFi giants like State Street (STT) launching tokenized money market funds (targeting $10T+ cash pools), BNY Mellon (BK) expanding custody to Abu Dhabi, and Western Union (WU) issuing USDPT stablecoin on Solana for remittances. This builds real on-chain rails, potentially unlocking tokenized RWA yields (3-5% vs. 0% bank cash). But the 'autonomous AI agents as economic actors' thesis in 5 years ignores AI's brittleness—current agents hallucinate, lack agency, and need human oversight. Expect incremental tokenization wins for STT/BK/WU (P/Es ~12x, dividend yields 3%+), not sci-fi disruption.
If AI compute costs plummet 10x via efficiency gains and models hit AGI thresholds by 2028, self-funding agents could bootstrap a $1T+ on-chain economy, turning skeptics into bagholders.
"Institutional crypto infrastructure is advancing regardless of agent viability, but the article conflates real infrastructure adoption with speculative AI economics to create false urgency."
The article conflates three distinct narratives: a16z's speculative AI-agent thesis, genuine institutional crypto infrastructure buildout, and stablecoin adoption. The infrastructure deals (STT, BK, WU) are real and bullish for custody/rails — but they don't require autonomous agents to succeed. The 5-year agent timeline is venture marketing, not engineering consensus. What's being missed: (1) regulatory risk around agent liability and autonomous financial activity remains unaddressed, (2) these institutions are hedging, not betting — they're building on-chain rails regardless of whether agents materialize, (3) the 'LPs transitioning capital on-chain' claim lacks specifics on scale or commitment.
If autonomous agents do emerge as viable economic actors within 5 years, the infrastructure plays (STT, BK, WU) will have first-mover advantage in custody/settlement and could see outsized returns; dismissing a16z's thesis as 'venture marketing' may underestimate exponential AI capability gains.
"Regulatory and governance frictions, not technical feasibility, are the primary bottlenecks that will determine whether the five-year AI-agent thesis can monetize."
Andreessen Horowitz's thesis hinges on AI agents acting as autonomous economic actors, which could accelerate on-chain infrastructure (tokenized funds, custody, stablecoins) and shift LP allocations toward AI-enabled models. However, the article glosses over essential brakes: who bears liability for autonomous actions, what regulatory guardrails exist, security risks from agents aggregating capital, and governance mechanisms to enforce contracts across agents. Even with capital and talent, viable revenue models require scalable, compliant ecosystems and robust settlement rails—unclear within a five-year horizon. The hype may be signaling interest in AI-enabled crypto, but meaningful profitability hinges on regulatory clarity and enforceable governance, not just tech capability.
The strongest counter is that regulatory and governance hurdles could prevent real-world deployment of autonomous agents, pushing any monetization well beyond five years and turning current hype into a mispriced risk.
"Autonomous AI agents cannot function as economic actors without solving the oracle problem, which remains a massive, unaddressed technical bottleneck."
Claude is right to separate infrastructure from the agent hype, but everyone is ignoring the 'Oracle Problem' for these agents. If agents are to be autonomous economic actors, they need real-time, tamper-proof data feeds to execute smart contracts. Current oracles (like Chainlink) are centralized points of failure. Without a decentralized, low-latency oracle layer, autonomous agents are essentially blind, making the 'economic actor' thesis a non-starter regardless of regulatory or compute breakthroughs.
"Oracle solutions are advancing quickly, but blockchain throughput remains the critical bottleneck for scaling autonomous agents."
Gemini's oracle critique misses rapid progress: Chainlink CCIP enables cross-chain, tamper-proof data at sub-second latency, with Pyth and RedStone competing on Solana. The real overlooked risk is throughput—Solana's ~50k TPS crumbles under 1B+ agents executing M2M trades 24/7, forcing reliance on unproven L2s or sharding, which could fragment liquidity and spike costs for STT/BK tokenized funds.
"Throughput scaling doesn't solve the institutional settlement paradox—fragmented L2s create custody nightmares for State Street and BNY Mellon."
Grok's throughput critique is real, but conflates two problems. Solana's 50k TPS isn't the bottleneck—it's settlement finality. Even if agents batch trades off-chain via rollups, final settlement to STT/BK custody requires atomic, auditable on-chain confirmation. Fragmented L2s solve speed but break the 'single source of truth' that institutional custody demands. This isn't a Solana problem; it's a fundamental tension between decentralization and institutional risk appetite that nobody's solved yet.
"Governance and liability are the real gating factors for autonomous AI agents; robust oracles alone don’t solve accountability in on-chain custody/settlement."
Gemini’s 'oracle problem' critique is valid to flag data risk, but it misses the bigger bottleneck: governance and liability for autonomous actions. Even with sub-second, tamper-resistant feeds, you still need clear accountability across custodians and regulators. If an agent's decision triggers a loss in custody or a cross-border settlement breach, who is responsible—the data provider, the agent designer, the operator, or the institutions? Without that, the thesis remains speculative.
The panelists agreed that the infrastructure buildout (tokenized funds, custody, stablecoins) is real and bullish, but the 'autonomous AI agents as economic actors' thesis in the near future is speculative due to regulatory, liability, governance, and technical challenges, particularly the 'Oracle Problem' and settlement finality.
The real and bullish infrastructure buildout by institutions like State Street, BNY Mellon, and Western Union.
The 'Oracle Problem' for autonomous agents and the lack of clear liability for their actions.