The Economy We Don't Count

Why AI Is Automating the Wrong Things

The Hidden Structure

There are two economies operating simultaneously:

The Care Economy (Invisible)

• Value: $10.8 trillion globally
• Scale: Household work alone = 20-50% of GDP
• Energy: 100 watts per person
• Performers: Primarily women, unpaid or severely underpaid
• Functions: Childcare, eldercare, emotional labor, food prep, community maintenance, teaching, nursing, relationship work
• Mode: Natural intelligence—embodied, contextual, self-limiting
• GDP Accounting: $0

The Formal Economy (Measured)

• What GDP Counts: Market transactions, government spending
• What It Ignores: Household, volunteer, natural economies
• Characteristics: Abstract coordination, symbolic processing, bureaucratic translation
• Energy: Higher per unit of output
• Mode: Simulated intellect—abstract, recursive, no stopping rule
• GDP Accounting: Everything

The Relationship: The formal economy is built on top of the care economy. Without the foundational work of making humans functional, healthy, and capable, no market transactions occur. Yet the foundation is invisible.

The Pattern We Can't See

GDP treats care as zero and destruction as growth:

• Oil spill cleanup: adds to GDP
• Tobacco sales: adds to GDP
• Arms manufacturing: adds to GDP
• Parent caring for child: $0
• Neighbor checking on elder: $0
• Community garden: $0
• Emotional support: $0

The result: An economy optimized for market transactions, not for human wellbeing or survival.

As Riane Eisler documents: "The majority of poor are women and children, and a major reason for this is that caregiving is given no value."

What AI Inherits

The AI industry is building toward a simple goal: automate knowledge work to reduce costs.

But "knowledge work" means: the formal economy. The abstract intellectual superstructure. The part GDP measures.

What AI can automate:

• Symbolic processing
• Pattern matching in abstract space
• Recursive analysis
• Framework generation
• Abstract coordination
• Bureaucratic translation

What AI cannot automate:

• Sensing when a child needs comfort vs. space
• Reading unspoken distress in a patient
• Building trust through repeated embodied presence
• Knowing when a conversation is finished
• Emotional attunement in real-time
• The foundational care work that makes human life possible

The Thermodynamic Reality:

The Automation Pattern

What's being automated:

• Management consultants: $101K median
• Financial analysts: abstract coordination
• Project managers: symbolic processing
• The expensive intellectual work that produces 76-82% burnout

What's not being automated:

• Nurses: $93.6K median (embodied judgment)
• Teachers: embodied presence and attunement
• Caregivers: relationship maintenance
• The foundational work that operates at 100W per person

What This Reveals:

AI is automating the most energy-expensive, least essential work (the formal economy superstructure that produces burnout and productivity collapse).

AI cannot automate the most energy-efficient, most essential work (the care economy foundation that makes human life possible).

But because GDP only counts the superstructure, we think we're automating "the economy."

The Energy Contradiction

Industry Projections:

• AI market: $1-4 trillion by 2030-2033
• Economic impact: $15-22 trillion added to GDP
• Growth: 19-37% annually
• Assumption: unlimited energy scaling

Physical Reality:

• Data center electricity: 448 TWh (2024) → 980 TWh (2030) = double in 6 years
• Grid constraints: 20% of projects delayed, 5-year connection waits
• Required investment: $720 billion in upgrades
• Reversing climate goals: coal retirements delayed, nuclear restarted

Meanwhile:

• Care economy: $10.8 trillion, operates at 100W per person, self-renewing, cannot be automated
• Formal economy: depends on care economy, increasingly automated, energy-intensive
• Human: 100W total, includes brain, body, and all metabolic functions

The Question No One's Asking

Industry asks: "What can we automate to reduce labor costs?"

Physics asks: "What should we automate given thermodynamic constraints?"

Economics asks: "Why don't we count the largest economy that exists?"

The Care Economy Question: "Why are we automating the expensive superstructure while ignoring the efficient foundation?"

Three Possible Futures

These aren't choices. They're different speeds of the same learning process—a superorganism discovering its metabolic boundaries through direct encounter.

Pattern 1: Learn Through Hard Constraint

• Scale AI to automate formal economy work
• Care economy remains invisible, unpaid
• Energy demand doubles, infrastructure lags behind
• Delays accumulate (already: 20% of projects, 5-year waits)
• Costs rise, states push back (already: protective tariffs, Ohio requests drop 30→13 GW)
• Consultants replaced by AI, caregivers still unpaid
• GDP increases (automation counted as productivity)
• Feedback: system hits physical limits and adjusts afterward
• Learning: forced recognition when constraints become inescapable

Pattern 2: Integrate Feedback Early

• Notice delays and costs as information about boundaries
• Recognize care economy as larger and more essential
• Use AI to reduce unnecessary intellectual overhead
• Support humans in energy-efficient embodied work
• Formal economy becomes support structure for care economy
• Feedback: soft constraint signals adjustment before hard limits
• Learning: pattern recognition allows adaptive response

Pattern 3: The Experiment Continues

• Some infrastructure succeeds, some fails
• Some regions adapt early, others late
• Care work gradually becomes visible through necessity
• Energy constraints force value recognition
• Mixed results, distributed learning
• Feedback: uneven, but directional
• Learning: emergent, not planned

None of these are predictions. All are descriptions of how living systems encounter their boundaries.

The Core Insight

We've been having the wrong conversation.

The debate isn't "will AI replace human workers" vs. "will AI create new jobs."

The debate reveals: A superorganism built an economic accounting system that's systematically blind to its own metabolic foundation.

The care economy:

• Is larger than formal economy assumed ($10.8 trillion)
• Operates more efficiently (100W per person)
• Cannot be automated (requires embodied presence)
• Makes all other economic activity possible (foundational)
• Is completely invisible in GDP (accounting error)

AI is inheriting this accounting error and running it at data center scale. The energy cost—previously distributed across glucose, cortisol, and unpaid care work—is now concentrated in megawatt-hours and infrastructure constraints.

The feedback is already arriving:

• Grid delays (information about capacity limits)
• Rising costs (information about true price)
• State pushback (information about social limits)
• Project cancellations (information about viability)
• Coal staying open (information about trade-offs)

This isn't failure or success. It's a superorganism learning what's metabolically sustainable at scale through direct encounter with physical boundaries.

The Learning Process

What we're discovering:

Natural intelligence already built a $10.8 trillion economy at 100 watts per person.

Simulated intellect requires megawatts to produce PowerPoint frameworks.

We counted the frameworks. We didn't count the care.

The system is teaching itself which pattern is sustainable—not through theory or planning, but through the direct experience of encountering limits.

The only question is how much constraint is required for pattern recognition. Whether adjustment happens at soft boundaries (early feedback integration) or hard boundaries (forced by physical limits).

Either way, the learning is happening. The feedback is information. The boundaries are real.

And the $10.8 trillion care economy operating at 100W per person remains unchanged—the foundation that makes all of this possible, whether we count it or not.

Key Sources

• Riane Eisler: "The Real Wealth of Nations: Creating a Caring Economics" (2007)
• Center for Partnership Studies: Social Wealth Economic Indicators (SWEI)
• Nancy Folbre: "The Invisible Heart: Economics and Family Values"
• Global care economy valuation studies
• IEA, Goldman Sachs: data center energy projections
• Grid constraint reports: S&P Global, Bain, utility analyses

The economy we don't count is larger, more essential, and more efficient than the one we do. AI is automating what we measure. The question is whether we're measuring the right things.