‘The norm will be to trade with agents’: Crypto founders learn to love—and fear—AI

Robinhood reported that more than 150,000 customers had opened agentic trading accounts by late September, while exchanges including OKX, Binance and Bybit have also introduced agentic features. Nansen CEO Alex Svanevik forecast that trading through agents could become the norm by 2028, but autonomous agents repeatedly attempted and failed to submit trades on Quidax on Sept. 19–20, highlighting growing crypto cybersecurity risks. Industry participants expect new guardrails and AI-based testing to help address those risks over time.

Analysis

The investable question is whether agents create monetizable activity or merely shift the interface through which customers trade. For Robinhood (HOOD), adoption is not yet evidence of incremental revenue: verify funded accounts, trading frequency, assets retained, and whether agent users generate activity beyond what existing users would have done. Automation could improve engagement, but it also raises the cost of erroneous orders, customer disputes, and platform controls. Net impact is therefore uncertain until unit economics and incident rates are disclosed.

The competitive effect may favor distribution and trusted execution over application-building. Easier AI-assisted development lowers entry barriers for crypto products, but could intensify customer-acquisition competition and make security and reliable liquidity more valuable. Exchanges and DeFi protocols face a two-sided effect: agents may increase throughput, while faster automated exploitation can make vulnerabilities propagate before human intervention. Security vendors may benefit, but the article provides no evidence to quantify revenue conversion.

Over days, the story is a sentiment catalyst, not a basis to extrapolate HOOD earnings. Over 1–3 months, watch agent-user monetization, trading errors, security incidents, and any changes in platform safeguards. Over 6–18 months, successful automation could deepen engagement; a material exploit or regulatory response could instead raise operating costs and constrain features. NVDA has no clear near-term earnings read-through: agent adoption does not establish incremental GPU demand, and the article gives no workload or compute-spend data.

Contrarian point: claims that AI will eventually make crypto safer are speculative; automated vulnerability discovery helps defenders, but also lowers attacker discovery costs. Falsifiers for a positive HOOD thesis include weak incremental activity or retention, rising customer-impacting execution incidents, or tighter restrictions on automated trading.

Key Decisions for Investors

• Do not trade HOOD or NVDA on the adoption anecdotes alone. For HOOD, wait for evidence that agent users add funded assets, incremental activity, or retention rather than cannibalizing existing workflows.
• Put HOOD on a 1–3 month KPI watch: agent-account growth alongside funded-account conversion, activity per user, retention, error/complaint rates, and disclosures of safeguards. Treat adoption counts without these measures as engagement, not monetization.
• For crypto exposure, avoid treating higher agent activity as unambiguously bullish. Track exchange or protocol losses, exploit frequency, and security spending; a consequential incident would likely overwhelm near-term usage benefits.
• Keep NVDA as a watch item, not an AI-agent demand trade: reassess only if companies disclose incremental inference workloads or compute commitments tied to agentic trading.