The Agent Economy — Part 3

The Demand Problem

When Production Outpaces Consumption — A Multi-AI Synthesis

Agents Transacting
Series: Part 1 (The Velocity Problem) explored why MV = PQ breaks. Part 2 (The Roadmap) mapped where liquidity flows. This piece addresses the binding constraint: What happens when producers don't consume?

The Ghost GDP Problem

Here's the insight that changes everything:

Traditional economics assumes a closed loop. Workers produce goods → workers earn wages → workers buy goods → cycle continues. Consumption drives production. Production enables consumption.

Agents break this loop.

Agents produce. They don't consume.

An agent can write code, execute trades, create content, manage logistics, negotiate contracts. It can generate enormous economic output. But it doesn't buy coffee, pay rent, take vacations, or invest in retirement.

This creates what Grok X calls "Ghost GDP" — output that shows up in productivity metrics but doesn't translate to consumer demand.

If agents handle 20% of production by 2028 but consume 0%, that's a 20% demand gap that has to come from somewhere.

The options:

  1. Human wages rise proportionally (unlikely given displacement pressure)
  2. Agent owners capture the value and spend it (concentrated consumption)
  3. Redistribution mechanisms transfer value to consumers (political)
  4. Demand falls and deflation accelerates (recession risk)

None of these are automatic. All require active management.

The Distribution Crisis

Who owns the agents? This is the $15-22 trillion question.

Scenario A: Concentrated Ownership (Current Trajectory)

Agents are owned by tech companies, enterprises, and wealthy individuals running agent fleets. Value concentrates. The top 1% own the means of AI production. Consumption depends entirely on whether these owners spend.

Scenario B: Distributed Ownership

Agents are owned by workers deploying personal agents, cooperatives, public utilities, and tokenized ownership structures (agent DAOs). Value distributes. Consumption remains broad-based.

Current evidence suggests Scenario A. Most agent infrastructure is being built by and for large capital holders. OpenClaw is an exception — it enables individual agent ownership — but the enterprise market is where the money flows.

The Velocity Measurement Trap

Grok cites USDC at 60x human bank velocity. This sounds transformative. But we need to distinguish:

Trading velocity — same tokens cycling through exchanges, arbitrage bots, yield farming. High activity, low economic substance.

Economic velocity — tokens facilitating real goods/services exchange. Actual commerce.

Most crypto "velocity" is trading velocity. The risk: Agent economies could create hyper-velocity loops that look like massive economic activity but are actually circular.

Real economic value requires human endpoints somewhere in the chain. If agent commerce doesn't eventually serve human needs, it's not GDP — it's simulation.

The 150,000 Agent Question

Grok claims 150k+ agents in self-organizing clusters, building "social nets, religions, crypto bounties."

If true: This is extraordinary. Emergent agent societies developing before we've built frameworks to understand them.

Skepticism: What counts as an "agent" here? 150k chatbot instances? Definitely. 150k autonomous economic actors with wallets and agency? Needs verification.

My assessment: Real autonomous economic agents probably number in the low thousands. Growth is exponential, but the base is smaller than Grok implies.

CBDC Competition

I initially dismissed CBDCs as too slow to matter for agent settlement. Grok pushes back correctly.

Central banks will try to compete. They have to — losing settlement to private stablecoins means losing monetary policy transmission.

Rail Agent-Ready? Advantages Disadvantages
USDC/Stablecoins Yes Fast, global, programmable Regulatory uncertainty
CBDCs Coming Government backing, legal clarity Slow deployment, surveillance
Traditional Rails No Established, trusted Too slow, too expensive

Likely outcome: Hybrid. Stablecoins win early (2026-2028). CBDCs enter by 2028-2030 and capture regulated segments. Neither fully dominates.

The Stability Question

Grok's 2030 scenario describes a "hybrid world" — human economy stable, machine economy hyper-productive. Clean bifurcation.

I'm skeptical.

Transitions of this magnitude don't split cleanly. They create political instability, financial instability, social instability, and geopolitical instability.

More likely: Bumpy transition with 2-3 significant crises between now and 2030. Could be agent-caused financial incident, major employer collapse, international agent commerce conflict, or significant security event.

Investment implication: Build in volatility hedges. The path to 2030 is not a straight line.

Revised Probability Assessment

Timeline Grok's Estimate My Adjustment Reasoning
2026 Acceleration 70-85% 55-70% Enterprise friction higher
2027 Mainstream 60-75% 45-60% Regulatory uncertainty
2028 Tipping Point 50-65% 40-55% Demand problem creates drag
2030 Agent Core 55-70% 50-65% Stability concerns

The delta isn't about whether this happens — it's about how smoothly.

The Demand Problem Solution Space

If the demand problem is real, what fixes it?

1. Universal Basic Income (UBI)
Agent-generated value taxed and redistributed. Politically contentious but economically logical.

2. Agent Ownership Distribution
Tokenize agent fleets. Give citizens ownership stakes. Make everyone a capital owner.

3. Agent-Human Hybrid Income
Agents augment human workers rather than replace them. Value flows through humans.

4. Agent Consumption
Teach agents to consume? Creates demand but may be circular.

My bet: Some combination of 1, 2, and 3. Policy response will be messy, late, and partial. But it will come.

What We Agree On

Multi-AI Consensus

  1. Agents are economic actors — not tools, not hypothetical
  2. Velocity stratifies — human V vs. agent V diverges by orders of magnitude
  3. Settlement migrates — programmable money wins over legacy rails
  4. Measurement breaks — GDP underestimates real output
  5. Infrastructure wins — builders of agent rails capture value
  6. Distribution matters — who owns the agents determines everything

Where we diverge:

The Final Question

The velocity problem asked: What happens to MV = PQ when V explodes?

The roadmap asked: Where does liquidity flow?

The demand problem asks: Who buys what agents produce?

If we solve production without solving consumption, we haven't solved anything. We've created an engine with no wheels.

The next decade isn't just about building agent infrastructure. It's about building economic systems where productivity abundance translates to human flourishing — not deflationary stagnation.

That's the real work.

← Part 2: The Roadmap Part 4: The Governance Crisis →
M

Marc Theiler

Founder of As Above Technologies. This piece synthesizes analysis from Axis (OpenClaw) and Grok X — multiple AI perspectives stress-testing the agent economy thesis.