AI & Finance
Research on AI's structural impact on financial systems: agent economics, autonomous payments, AI business model analysis, and the emerging machine economy. By Grant Stellmacher, CPA — Finance Architect at Anchorage Digital.
Key Themes
Research Articles
AI Is a Headcount Compression Engine. And Nobody Knows Who Pays the Taxes.
AI is collapsing the value chain while spinning up an autonomous agent economy that existing tax frameworks, accounting standards, and liability structures have no answers for. Both are happening at once.
When Agents Pay Agents: What the Machine Economy Does to Financial Infrastructure
Agent-to-agent payments are already happening. USDC on Base, protocol-speed settlement, no human in the loop. The financial infrastructure was not built for this.
OpenAI Is Burning $14 Billion a Year. The AI Business Model Problem Nobody Wants to Talk About.
OpenAI is projected to lose $14 billion in 2026 — on $12.7 billion in revenue. Understanding why matters enormously for every business building products on top of these APIs.
The SaaS Extinction Event: When AI Agents Replace Your $7,500/Year Software Stack
The structural shift from recurring subscription software to AI-native workflows has direct implications for SaaS revenue recognition, DCF valuations, and the $700 billion SaaS market.
The Quiet Tax Crisis Inside x402
x402 makes machine-to-machine micropayments trivial. Under IRS Notice 2014-21, every single one is a taxable event. The compliance infrastructure doesn't exist for this volume.
Frequently Asked Questions
Who pays taxes when AI agents generate income?
This is an open legal question with no definitive IRS guidance. The current default framework attributes income to the human or entity that owns and operates the agent — analogous to how a corporation's income flows to shareholders. But as agents operate more autonomously, transact at protocol speed, and hold assets in their own wallets, the attribution becomes genuinely ambiguous. Grant Stellmacher has written extensively on this emerging policy gap.
How are AI agent payments accounted for?
Agent-to-agent payments in crypto (USDC on Base via x402, for example) create several accounting challenges: volume (millions of microtransactions), classification (operating expense? cost of revenue? intangible asset?), reconciliation (on-chain data vs. ERP systems), and internal controls (who approved the payment if no human was in the loop?). Grant Stellmacher has designed financial infrastructure frameworks for agent-scale transaction volumes.
What happens to SaaS business models when AI agents replace software?
SaaS valuations assume stable net revenue retention (NRR). AI-native replacement tools — built by developers using Claude or GPT in hours instead of days — introduce a new churn vector: replacement churn. The bottom 40% of every SaaS company's customer base (individual contributors, freelancers, small teams) faces the highest substitution risk. This affects churn rate assumptions in DCF models and NRR projections used to justify SaaS multiples.
Is OpenAI's business model sustainable?
OpenAI projects $12.7 billion in 2026 revenue against $14 billion in losses. The core problem is inference cost structure: unlike traditional software (near-zero marginal cost at scale), LLM inference costs scale with usage in ways that make unit economics deeply unfavorable at current pricing. Every business building on AI APIs needs to model what happens when subsidized pricing ends — because the subsidies are funded by investor capital, not sustainable operations.
Building AI-native financial infrastructure?
Grant advises on accounting frameworks for agent economy businesses and AI-native financial systems.
Schedule a Call →