The agentic commerce hype cycle has hit its peak, and with it, a chaotic spread of market size projections. A new analysis from DEV.to cuts through the noise, revealing that while major analysts agree the sector is growing rapidly, their numbers are often incomparable. McKinsey projects a $3–5 trillion global market by 2030, while Morgan Stanley pegs US e-commerce impact at $190–385 billion. The discrepancy isn't just disagreement; it's a difference in scope, methodology, and what counts as an 'agent-initiated' transaction.

Why The Numbers Don't Add Up

Analysts are sizing different markets. McKinsey’s trillion-dollar figure covers global consumer and merchant goods but explicitly excludes services and most B2B transactions. In contrast, Bain & Company and Morgan Stanley focus narrowly on US e-commerce sales, with Bain predicting $300–500 billion and Morgan Stanley $190–385 billion. Juniper Research offers a different perspective entirely, forecasting a 43,240% growth in global transaction value from $8 billion in 2026 to $3.5 trillion by 2031. These figures cannot be quoted interchangeably without context, as adding them together would result in massive double-counting.

Adoption Is Happening Now

While the trillion-dollar forecasts are future projections, current adoption signals in 2026 show the infrastructure is already live. Forrester reports that 11% of US online transactions over $75 already involve an AI agent. Shopify data indicates AI-driven traffic to its stores tripled year-over-year in Q2 2026. Furthermore, Morgan Stanley notes that 23% of US adults have already tried an agentic shopping assistant. On the payment rail side, Coinbase’s x402 protocol processed 165 million agent transactions by mid-2026, with Visa, Mastercard, and Stripe joining the foundation, signaling that the payment layer is maturing alongside consumer adoption.

Implications For Builders

For developers and merchants, the practical takeaway is not to obsess over the final trillion-dollar number, but to prepare for the operational reality of agent-mediated transactions. The article emphasizes that being discoverable is critical; if a catalog isn't machine-readable, it is invisible to the existing agent traffic. Builders must also design for authorization and recovery, as millions of daily agent transactions require scoped mandates, expiring grants, and durable commitment records. The underlying protocol stack, including x402 for payments, AP2 for mandates, and ACP for checkout, is becoming the standard infrastructure layer for this new commerce model.

Key Takeaways

  • McKinsey's $3–5T figure is global goods only; it excludes services and most B2B, making it the widest scope.
  • Bain ($300–500B) and Morgan Stanley ($190–385B) are the most comparable US-specific forecasts.
  • Juniper's 43,240% growth rate is a directional signal from a small 2026 base, not a precise prediction.
  • Current adoption is real: 11% of US transactions over $75 use AI agents (Forrester), and 165M x402 transactions occurred by mid-2026.
  • Do not add forecasts from different sources; they measure different market baskets.

The Bottom Line

Stop quoting trillion-dollar figures without defining the scope. The real opportunity for builders in 2026 is not guessing the final market size, but building robust authorization and recovery layers for the millions of agent transactions already happening today.