iProDecisions Research
Financial Infrastructure · Agentic Commerce · Published February 2026
The Autonomous Agent
Economy Will Redefine
Financial Infrastructure
Stripe's 2025 annual letter signals a structural inflection point: AI agents are poised to become the primary conductors of internet transactions. This report analyses the infrastructure imperative, quantifies the scale gap, and maps the strategic landscape for the decade ahead.
1B TPS
Stripe's Projected Requirement
Exact Collison quote: "1M or even 1B TPS"
$1.9T
Stripe 2025 Payment Volume
+34% YoY - ~1.6% of global GDP
$400B
2025 Stablecoin Volume
Doubled YoY - 60% B2B - Bridge 4x growth
4,768x
Infrastructure Gap to Close
ICP theoretical max vs. Stripe's 1B TPS target
Contents
Section 1 - Executive Summary
Section 2 - The ACSM Framework
Section 3 - Data Exhibits
Section 4 - Stakeholder Implications
Section 5 - Scenario Analysis
Section 6 - Strategic Thesis
About iProDecisions
Section 01
Executive Summary
Three Structural Conclusions
✓
All primary figures independently verified from Stripe's 2025 Annual Letter (Feb 25, 2026): $1.9T TPV (+34% YoY) · "1M or even 1B TPS" (exact Collison quote) · $400B stablecoin volume (doubled YoY, 60% B2B) · Bridge volume 4x · 12-hour settlement delay + 35x fee spike from memecoin congestion · ICP 1,196 TPS avg / peak 25,621 TPS (theoretical max 209,708 TPS) · Solana 1,140 TPS avg / peak 5,289 TPS (theoretical max 65,000 TPS) via Chainspect Feb 2026 · Stripe-OpenAI Agentic Commerce Protocol (ACP) confirmed · Stripe Tempo blockchain (with Paradigm) confirmed · Global fintech $321B-$653B (2025-2030) via Mordor Intelligence. Strategic frameworks and TAM layer proportions are iProDecisions analytical models, clearly labeled.
Situation
Stripe processed $1.9 trillion in payment volume in 2025 - yet the Collison brothers warn this is only the opening act. AI agents are transitioning from experimental tools to primary economic actors on the internet.
Stripe's Five Levels of Agent Capability
L1
Automation - Filling out web forms and standard online tasks
L2
Descriptive Search - Finding results based on situational descriptions -- Agents are here now
L3
Context Preservation - Remembering user preferences and requirements
L4
Delegation - Performing tasks on behalf of users including commerce
L5
Anticipation - Suggesting solutions without explicit prompts
Source: Stripe Annual Letter 2025 - Patrick and John Collison
Complication
ICP and Solana - the two fastest blockchains - average ~1,196 and ~1,140 TPS respectively. Their theoretical maximums are 209,708 and 65,000 TPS. Stripe needs 1,000,000 to 1,000,000,000 TPS. Even at theoretical maximum, ICP closes only 0.02% of the gap. A memecoin frenzy in 2025 delayed one Bridge settlement by 12 hours and spiked fees 35x. That is today. Under agent load, that becomes the baseline.
Network
Avg TPS
Theoretical Max
Gap to 1B TPS
Solana
1,140
65,000
15,385x
Stripe Target
1,000,000,000
Not yet built
Baseline
Source: Chainspect (Feb 2026) · Stripe Annual Letter 2025
Resolution
Stripe is explicit: "meeting AI demands will likely require a horizontal architecture of multiple, interacting chains rather than single high-TPS networks." This validates the ACSM framework entirely. The Agentic Commerce Protocol (ACP, built with OpenAI), Stripe Tempo (with Paradigm), and Bridge acquisition are the opening moves of this re-architecture.
I
Throughput is no longer a performance optimization - it is a prerequisite for market existence
When millions of agents transact continuously - paying per inference, per API call, per millisecond of compute - payment infrastructure becomes the physical substrate of machine-driven markets. Institutions designing for peak human checkout flows are architecturally miscalibrated for the coming decade. The Stripe Collisons are explicit: "agents will most likely soon be responsible for most internet transactions."
Infrastructure decisions in 2025-2027 will be path-dependent for 10-15 years
II
Money velocity in digital ecosystems will compound superlinearly as agent density grows
Each AI agent compresses the decision-to-settlement cycle toward zero. As agent populations scale, money circulates with exponentially higher frequency through digital value chains. This has no historical precedent in physical economies. Traditional monetary velocity models, credit pricing frameworks, and systemic risk models require fundamental revision to accommodate continuous machine-initiated settlement flows.
Traditional monetary velocity models require fundamental revision
III
The central competitive question has permanently shifted from scalability to re-architecture
This is not an engineering challenge with an incremental solution. It is a platform shift analogous to the move from mainframes to internet-native applications. The question is whether the global financial stack can be rebuilt from first principles for autonomous commerce at planetary scale - and who captures the infrastructure layer when it is. First-mover incumbency in this layer may prove as durable as SWIFT's four-decade hold on institutional messaging.
Late entrants face structural disadvantages, not merely operational ones
Primary Source Signals — Stripe Annual Letter 2025
Exact Collison Quote
"Agents will most likely soon be responsible for most internet transactions."
— Patrick & John Collison · Feb 25, 2026
Infrastructure Mandate
"Meeting AI demands will likely require a horizontal architecture of multiple, interacting chains."
— Stripe Annual Letter 2025
Live Congestion Event
One memecoin frenzy: 12-hour Bridge settlement delay, 35x fee spike. This is the baseline today.
— Verified · Chainspect Feb 2026
Stablecoin Acceleration
$400B stablecoin volume — doubled YoY. 60% B2B. Bridge grew 4x. The velocity is already accelerating.
— Stripe Annual Letter 2025
All quotes above are direct primary source material from the Stripe 2025 Annual Letter (Feb 25, 2026) by Patrick and John Collison. TPS data from Chainspect (Feb 2026). No editorial interpretation applied — these are verbatim institutional signals.
All data figures cited with sources. Stripe Annual Letter 2025 (published Feb 25, 2026, Patrick and John Collison). Chainspect TPS data as of Feb 2026. Fintech market sizing: Mordor Intelligence 2025 Report. TAM illustrations by iProDecisions Research are clearly labeled as illustrative order-of-magnitude models. This report represents the thought leadership of Kishor Akshinthala / iProDecisions and does not constitute investment advice.
Section 04
Stakeholder Implications
What This Means for Each Actor
🏛
Traditional Banks
Legacy payment rails face structural obsolescence unless re-architected for machine-native settlement within 5 years
Banks designed for batch settlements lack the latency, programmability, and micro-denomination capabilities required for agent-scale commerce. The risk is not incremental market share loss - it is becoming the highway while the internet becomes the city.
->Invest in programmable money infrastructure: tokenized deposits, smart settlement layers
->Acquire or partner with L2/L3 settlement orchestrators before the 2027 strategic window closes
->Redesign risk models for continuous micro-transaction flows, not periodic batch clearing
Critical
⚡
Fintechs and Payment Processors
First-mover advantage in agentic payment primitives may prove more durable than any prior fintech moat
Stripe and Adyen sit at a strategic inflection. Their API-first architectures position them well - but only if they extend toward embedded agent-native payment primitives before specialized competitors emerge. Stripe's Tempo blockchain and OpenAI partnership signal the direction clearly.
->Launch agent-native SDKs and payment primitive libraries targeting AI developers today
->Build micro-settlement infrastructure capable of sub-cent, sub-millisecond transactions
->Position as the payment layer of the agent economy with dedicated infrastructure commitments
High Priority
☁
Cloud and Compute Providers
Cloud giants will become de facto financial intermediaries as compute and commerce converge in the agent layer
AWS, Azure, and GCP are uniquely positioned - agent workloads already run on their infrastructure. The convergence of compute billing (pay-per-inference) and commerce settlement creates an opportunity to become the financial substrate of AI-driven markets.
->Develop native agent commerce APIs that abstract settlement into compute workflows
->Consider strategic acquisitions in programmable settlement layers to avoid rail dependency
->Engage regulatory frameworks early - financial infrastructure designation becomes likely
High Priority
⚖
Regulators and Policymakers
The Fragmented Scenario - lowest probability, highest economic harm - is the direct product of regulatory inaction or incoherence
Regulators face a narrow window to establish coherent frameworks before balkanization sets in. The 15% Fragmented Scenario emerges directly from jurisdictional patchwork. Inaction is itself a policy choice with quantifiable economic costs to global GDP growth.
->Develop regulatory sandboxes for agent-to-agent commerce with clear liability attribution
->Establish international coordination before national frameworks crystallize and diverge
->Define legal standing and financial obligations for autonomous AI agents as economic actors
Strategic
Section 05
Scenario Analysis - 2030 Horizon
Three Futures and Key Divergence Triggers
55% Probability - Base Case
Ordered Transition
Regulatory clarity unlocks institutional deployment
1M+
TPS Requirement - Stripe lower bound - 833x current blockchain max
Key Divergence Triggers
Regulatory sandboxes for agent commerce go live in EU, US, Singapore by 2026-2027
Stripe Tempo blockchain or equivalent achieves 100K+ TPS with sub-2 second finality
Enterprise AI agent deployments exceed 1 billion globally; B2B stablecoin becomes mainstream
OpenAI Agentic Commerce Protocol gains industry-wide adoption as the standard
Positioned to Win
Stripe / Tempo
Ethereum L2s
AWS / Azure
Circle / USDC
Chainlink
30% Probability - Bull Case
Hypergrowth Break
AGI-adjacent agents compress timelines 3-5 years
1B+
TPS Requirement - Stripe upper bound - 833,000x current blockchain max
Key Divergence Triggers
Next-generation AI models achieve true autonomous economic agency - self-fund, self-direct
Crypto infrastructure gains full regulatory legitimacy and institutional adoption accelerates
A major DeFi protocol achieves SWIFT-scale volume - proof-of-concept for machine-native rails
Agent-to-agent commerce creates emergent financial markets not originally designed by humans
Positioned to Win
Purpose-built rails
Solana
Novel L1s
AI infrastructure cos
15% Probability - Bear Case
Geopolitical Fracture
Balkanization prevents global interoperability
100K+
Fragmented across incompatible regional rails - still 83x current blockchain max
Key Divergence Triggers
US-China AI competition triggers incompatible financial infrastructure standards
EU regulatory patchwork creates compliance barriers to cross-border agent commerce
A major AI-driven financial incident triggers broad regulatory retrenchment
No settlement protocol achieves critical mass - liquidity fragmented across 20+ incompatible rails
Positioned to Survive
Regional rails
CBDC infra
Compliance tech
National champions
Across All Scenarios
Even the Bear Case demands 100K+ TPS - 83x beyond today's fastest blockchain. The floor has permanently moved. Building for anything less is not conservative; it is building for irrelevance.
The Asymmetric Bet
Building for Base Case (1M TPS) costs only marginally more than Bear Case (100K TPS) infrastructure - but positions you to capture 85% probability outcomes. Undershooting on the scale question is structurally irreversible.
The Regulator's Dilemma
Regulatory inaction is not neutral - it defaults toward the Fragmented Scenario. The absence of coherent global frameworks is itself a policy choice with material, quantifiable economic costs measured in GDP foregone.
Section 06
Central Thesis
Strategic Imperative for the Decade
Payment Paradigm Evolution
1990-2010
Human-initiated, batch settlement
Payments as discrete end-of-workflow events. Human navigates interface, triggers payment, settlement in days. SWIFT, ACH, card networks. Infrastructure designed for human scale.
2010-2024
API-native, real-time rails - We are here
Payments embedded in software. Stripe, Adyen normalize real-time rails. Humans still initiate. Settlement in seconds to minutes. Agent participation growing at edges. Stablecoin volume $400B and doubling.
2025-2035
Machine-native, continuous, planetary scale
Payments as embedded execution primitives inside agent loops. Pay-per-inference, per-millisecond, per-API-call. Millions to billions of agents transacting continuously. Settlement as a background process, not an event. 1M to 1B TPS required.
Decision Window
2025-27
Estimated period before winner-take-most dynamics in L1-L2 layers begin crystallizing
Infra Gap to Close
833x-833K x
Current peak blockchain TPS vs. Stripe's 1M-1B TPS projection. No incremental fix exists.
"The question is no longer whether existing rails can scale incrementally. It is whether the global financial stack can be re-architected to support autonomous economic activity at planetary scale. The answer will define the competitive landscape of fintech, crypto infrastructure, and cloud-compute marketplaces for the next decade."
- Kishor Akshinthala / iProDecisions Research - February 2026
Synthesizing: Stripe Annual Letter 2025, Chainspect Data, Mordor Intelligence Market Research
Strategic Decision Framework - For Institutional Actors
Q1: Is your infrastructure designed for human-scale or machine-scale commerce?
->If human-scale: you are building for the wrong paradigm. Estimated obsolescence window: 3-7 years.
Q2: Can your settlement layer handle sub-cent, sub-millisecond transactions at 1M+ TPS?
->If no: this is a foundational requirement, not a feature request. The gap cannot be bridged incrementally.
Q3: Do you have programmable money infrastructure that AI agents can natively interact with?
->If no: agents will route around you. Programmability is the interface layer of machine-native commerce.
Q4: Have you modeled your competitive position across all three scenarios?
->If no: you are not managing risk - you are assuming the Base Case. That assumption should be explicit.
The Inaction Cost
Every quarter of delay narrows the strategic window. Infrastructure decisions made in 2025-2027 are path-dependent for 10-15 years. The cost of early action is marginal. The cost of late action is structural and likely permanent.
#AutonomousAgents
#AIEconomy
#MachineToMachine
#ProgrammableFinance
#FutureOfPayments
#FintechInfrastructure
#BlockchainScalability
#DigitalEconomy
#AIInfrastructure
#AgenticCommerce
#NextGenInternet
This report is published for informational and thought leadership purposes only. It does not constitute investment, financial, or legal advice. All data cited from named third-party sources. iProDecisions analytical frameworks are clearly labelled as proprietary models. Past performance is not indicative of future results.