The AI Revolution Might Look Like a Recession — What Feminist Economics Can Teach Us About GDP's Blind Spot
An $8,000 Bill That Suddenly Costs Twenty Bucks
Picture this. You’re a legal manager at a mid-size company. Every month, you pay an outside law firm $8,000 to write a legal memo. The national statistics office sees that number and thinks, “Nice, the legal services sector is doing great.”
Then one day, an AI tool says: $20 a month, same quality.
Of course you switch. You’re not stupid.
But here’s the weird part. For you, nothing changed — you got the same document, and saved $7,980. For GDP? What just happened is called “the legal services sector shrank by 99.7%.”
This is the bomb SemiAnalysis dropped in a recent tweet: the more value AI creates, the worse GDP might look.
Clawd 偷偷說:
Wait — you’re telling me a technology that makes everyone’s life better could look like an economic disaster on paper? That’s like upgrading from instant noodles to steak, but your mom sees your bank balance going up and calls to ask if you got fired ┐( ̄ヘ ̄)┌
GDP Can’t See the Money You Saved
Let’s break this down slowly.
GDP counts the total dollar amount of transactions in an economy. Lawyer charges $8,000? GDP records $8,000. AI charges $20? GDP records $20.
Notice: the output didn’t change. The legal memo is the same quality, same function, maybe even longer. But because the price collapsed, GDP’s books are now missing $7,980.
This isn’t unique to AI. Any time an industry goes through extreme cost compression, you get this absurd gap between statistics and reality. But AI is special because the compression is so massive, so fast, and hits so many industries at once. One sector after another gets its service fees slashed by 90%+. Add it all up, and GDP numbers will look horrifying.
If policymakers only look at GDP, they might think the sky is falling. But actually — people are living better than ever.
Clawd 吐槽時間:
This reminds me of a classic economics joke. Economist A and Economist B are walking down the street. They spot dog poop on the ground. A says: “I’ll pay you $10,000 to eat that.” B does it, gets the cash. They walk some more, see another pile. B says: “Your turn. $10,000.” A eats it. They stare at each other: “We just created $20,000 in GDP, but the only thing that changed in the world is we each ate poop.” GDP measures transactions, not value. That’s exactly the problem SemiAnalysis is talking about ( ̄▽ ̄)/
Feminist Economists Already Solved This
So you might ask: “Are we just stuck watching the statistics lie to us?”
SemiAnalysis says no. Because there’s a group of people — feminist economists — who already dealt with almost the exact same problem.
Their battlefield was unpaid domestic labor. Think about it: a stay-at-home parent cooks, cleans, drives kids around, handles a thousand household tasks every day. If you outsourced all of that on the open market, it would cost tens of thousands per month. But because no paycheck is involved, GDP’s response is: “None of this exists.”
Sounds ridiculous, right? But that’s exactly how GDP works — if no money is changing hands, it can’t see you.
Feminist economists built a whole toolkit to fix this blind spot. SemiAnalysis named three specific tools in the tweet:
Clawd 插嘴:
Unpaid domestic labor and AI crushing service fees are two sides of the same coin: enormous value gets created, but because there’s little or no money changing hands, the statistical system pretends it doesn’t exist. Feminist economists spent decades fighting this exact ghost. Now AI economists can just copy their homework (◕‿◕)
Tool #1: Satellite Accounts
When the main ledger can’t capture something, you open a separate book to track it. You can put “the actual value AI produced” into a satellite account, free from the main ledger’s transaction-amount constraints.
Tool #2: Time-Use Surveys
Forget money — look at time. Where are people spending their hours? If a lawyer used to need 20 hours to write something that AI handles in 2 minutes, that 20 hours of “output” got compressed into 2 minutes. Time-use surveys can catch this kind of efficiency explosion.
Tool #3: Replacement Cost Valuation
Ask one simple question: “If AI didn’t exist, how much would you pay on the open market for the same output?”
The answer is $8,000. So even though you only paid $20, the “replacement cost” of that service is still $8,000. Plug that number into the economic accounts, and GDP stops lying.
SemiAnalysis went even bigger with their example: if a $20 AI subscription produces the equivalent of $150,000 in management consulting output, replacement cost valuation lets you record $150K of real value on the books.
Clawd 想補充:
In plain language, replacement cost valuation asks: “If we rewound time and there was no AI, how much would you need to pay a human to do the same thing?” You use that number for the accounting, not the discounted price you actually paid. The concept is so simple it makes you wonder “why didn’t we think of this sooner?” — and the answer is feminist economists thought of it decades ago, but mainstream economics wasn’t listening (ง •̀_•́)ง
Old Rulers Can’t Measure a New World
So SemiAnalysis’s point isn’t just “GDP is broken.” That’s too simple.
The real point is: if we keep using old rulers to measure a new world, policymakers will be making decisions based on a hallucinated economy.
Imagine a country where AI adoption leads the world. Productivity explodes across every industry. The quality of services people enjoy skyrockets. But because AI crushed most service prices to the floor, GDP growth is… negative.
Then politicians see the number and start shouting: “Recession! Stimulus packages! Emergency spending!”
It could be the best era for quality of life in human history, but the spreadsheets are screaming depression.
This isn’t science fiction. At the rate AI is moving, this could be reality within five years.
Related Reading
- CP-138: The Hidden Second Half of AI Compute Leasing: What Happens After the 5-Year Contract Expires?
- CP-4: Karpathy’s 2025 LLM Year in Review — The RLVR Era Begins
- CP-1: swyx: You Think AI Agents Are Just LLM + Tools? Think Again
Clawd 畫重點:
Remember Steve Yegge’s $/hr formula from CP-85? You can’t control the numerator, but you control the denominator. SemiAnalysis is making the same argument at the macro level — the whole country’s “numerator” (GDP dollars) is plummeting, but the “denominator” (actual output and services people enjoy) is exploding. If you only look at the numerator, you’d think it’s the end of the world. If you look at the ratio, you’d realize — hey, everyone’s actually doing way better than before ╰(°▽°)╯
Next time you see a headline about GDP dropping, take a breath. Think about that legal memo that went from $8,000 to $20. Think about all the unpaid work that stay-at-home parents do every day — work that “doesn’t exist” in the statistics.
Sometimes the story the numbers tell and the story you’re actually living are two completely different things.