21Shares Says September Market Begins Convergent Pricing
21Shares stated that September marked the first time the market priced cryptocurrency and artificial intelligence based on convergence rather than competition. The reason is that AI agents require wallets and settlement rails, which cannot rely on bank accounts.
The company's configuration report on September 2 described both as complementary technologies. AI needs callable funds, memory, and computing power, while cryptocurrency requires something to perform complex operations on behalf of users. The report cited that by 2030, agents could mediate $30 trillion to $50 trillion in consumer business. If 10% of this goes through stablecoin rails, the scale would be several times larger than today's stablecoin card transactions.
The report also noted that agents would handle finance, commerce, and security in a way similar to how broadband and smartphones simplified the internet. Losses in Ethereum due to user errors alone exceeded $1 billion, and automation was described as a removable cost. Agents cannot hold bank accounts, which was emphasized as a primary requirement for settlement.
In September, prices moved strongly. Grayscale's AI cryptocurrency sector rose 54% that month, while the broad cryptocurrency market increased by 24%. NEAR rose 183%, Venice rose 70%, World rose 47%, and Bittensor rose 37%. The sector's size remains around $15 billion, making it the smallest among its six major sectors.
BlackRock's digital asset research in September characterized AI as machine-native intelligence and digital assets as machine-native currency. The circulation of stablecoins exceeded $300 billion in September, with adjusted trading volume expected to surpass $11 trillion by 2025. Coinbase's x402, Google's AP2, and Visa's TAP were listed as more suitable protocols for all-weather, small, high-frequency transactions. Over the past 12 months until April, agents settled approximately 176 million payments on public chains, most of which were between 1 and 10 cents.
Buyers view agent payments as new demand for configuration funds, while sellers previously considered AI and cryptocurrency as mutually exclusive positions. The event was driven by excess returns in September and the settlement gap where agents cannot open bank accounts, with funds flowing from pure model trading to wallets, stablecoins, and verifiable computation. Beneficiaries include stablecoin rails and agent payment protocols, while traditional bank rails that rely on daily batch clearing and cannot accept small machine payments are under pressure.
Source: Public Information
ABAB AI Insight
21Shares is involved in cryptocurrency exchange-traded products, where pricing power originally lay in the transaction fees of Bitcoin and Ethereum. Changing the configuration language to convergence in September effectively tells existing holders that AI budgets can remain in the cryptocurrency basket without having to choose one over the other. Grayscale separately listed the AI sector that month, with a 54% increase compared to 24%, giving this narrative a tradable month.
The capital path is linked to agent payments, not model equity. In September, BlackRock characterized stablecoins as machine-speed currency, while Coinbase, Google, and Visa pushed x402, AP2, and TAP, respectively. Stripe acquired OpenRouter in August, buying a model routing and payment interface. If McKinsey's projected $30 trillion to $50 trillion in agent business sees 10% enter stablecoins, the incremental funds would not go into bank accounts but into programmable balances.
The analogy is that in 2017, broadband companies did not understand mobile payments and settlements, which were taken over by QR codes, and in 2020, stablecoins consumed cross-border retail before central bank digital currencies. The current situation is not about applications winning but about rail bidding. The $15 billion sector indicates that prices have already moved, but the scale has not yet left the cryptocurrency internal.
This represents a replacement of the settlement layer in technological substitution. Bank accounts require entities, business hours, and minimum amounts, while agents require no entities, all-day service, and penny-level payments. September's pricing placed AI capital expenditures and cryptocurrency rails into the same position, causing bank batch clearing to lose its default entry point for machine business.
ABAB News · Cognitive Laws
Agents cannot open bank accounts, so rails shift on-chain.
Convergent pricing occurs after excess returns.
The entry point for machine business is penny settlements, not accounts.