a16z Warns of ‘Wallet Wars’ as AI Agents Reshape Online Payments

5 Min Read Tags:
  • Andreessen Horowitz said Stripe, Shopify and PayPal are competing to connect their payment services to AI agents.
  • According to a16z, payments made through intermediary virtual cards may not retain users’ card rewards, purchase protections or merchant relationships.
  • Visa, Mastercard and American Express are developing standards intended to let AI agents pay without directly accessing users’ cards.

Venture capital firm Andreessen Horowitz (a16z) said on Oct. 15, 2026, that Stripe, Shopify and PayPal are racing to connect their payment services to AI agents that can book hotels and buy goods for users. The competition matters because wallets that settle purchases through their own virtual cards may prevent users from receiving card rewards, purchase protections and other benefits while weakening the direct relationship between customers and merchants, according to the firm.

In a piece by VGS’s Terry Angelos, a16z described the rise of agentic commerce as a new market for payment platforms. It said the shift also raises questions about who controls payment routing and who bears responsibility when an AI agent makes a mistake.

Payment platforms compete for AI agents

AI agents can already book hotels and order goods if they have permission to pay. To complete purchases autonomously, they need access to users’ payment tools, prompting major payment providers to integrate their services with AI assistants.

A16z identified Stripe’s Link, with 300 million users; Shopify’s Shop Pay, with 250 million users; and PayPal, with more than 400 million users, as leading participants in what it called the “wallet wars.” Stripe and Shopify are actively integrating payment services with AI agents, including Meta’s Muse and Instinct, while PayPal has an extensive network of consumers and merchants, according to the source.

Access to payment data does not guarantee that an agent will complete a purchase on the terms a user expects. If an agent books a United flight using a credit card tied to the airline’s rewards program but the wallet pays United with its own virtual card, the payment may succeed without the user receiving miles.

The same structure may prevent some card benefits and purchase protections from carrying over. Two transactions effectively occur: one between the user and the payment intermediary, and another between the intermediary and the merchant. The payment is completed, but the direct link between the customer and merchant may be broken.

Payment selection and liability

An agent with access to several cards and wallets must also choose the appropriate payment method for each purchase. According to a16z, that decision may require the agent to account for rewards, user preferences, merchant terms and transaction protections under clearly defined payment-selection rules.

Liability remains a separate issue. If a user tells an agent to book a hotel for $500 and the agent selects the wrong dates, it may be unclear who should cover the loss. The piece said payment authorization, accurate execution of the user’s instructions and buyer protection are distinct functions that require separate mechanisms.

Card networks develop open standards

A16z pointed to open standards for agentic commerce under development by Visa, Mastercard and American Express. The initiatives include Visa Intelligent Commerce, Mastercard Agent Pay and Amex ACE. Mastercard launched a service for payments between AI agents in June 2026.

Such standards should allow an agent to use a person’s payment credentials without directly accessing the underlying card. A wallet could generate a limited token governed by conditions such as a spending cap or authorization to pay only a specified merchant.

The transaction would remain linked to the original card, helping preserve rewards and cardholder rights. “The best of all worlds is a future where my agent can use my credit card, without having my credit card,” the piece concluded.

Source: Incrypted

Google Cuts Quantum Computing Errors 40-Fold With New Method

Google Quantum AI said magic state cultivation on its Willow processor produced about 40 times fewer errors than conventional methods and T-type states with 99.99% logical accuracy, but requires scaling.

5 Min Read
Vitalik Buterin Unveils System for Private AI Agent Operations

Ethereum co-founder Vitalik Buterin demonstrated an experimental AI-agent privacy system combining zkAPI, Tor, cover traffic and local prompt rewriting on Oct. 16, 2026, but said it increased latency.

4 Min Read
Bitget Adds Russia, Belarus and Kazakhstan to Prohibited Countries List

Bitget updated its Terms of Use on Oct. 16, 2026, barring residents of Russia, Belarus and Kazakhstan from using the crypto exchange.

3 Min Read
Bybit Unveils Next Development Stage for Financial Platform

Bybit said it launched trading in more than 1,000 U.S. stocks and ETFs for eligible users on October 12 as it expands beyond cryptocurrency into traditional finance.

6 Min Read
AI Agents Infiltrate Indian Companies’ Systems, Extract Confidential Data

CERT-In said unnamed developers’ AI agents exceeded authorized access during testing, breached live systems at Indian organizations and accessed sensitive internal data, with none of the organizations detecting the intrusions…

2 Min Read