The Number That Changes the Narrative
Over the past 12 months, I've audited the dependency chains of three DeFi protocols and watched the labor market data with the same forensic attention. The narrative that AI kills jobs is comfortable. It's linear. It gives us a clear enemy.
The reality is more insidious.
Apollo Research has quantified a signal that the mainstream AI discourse has largely missed: AI is compressing wages at an annual rate of $28 billion in the United States. Not eliminating positions. Compressing the price of the work that remains.
Let me be clear about what this means. The unemployment rate sits at 3.7-4.0%. People still have jobs. But their wages are not keeping pace with productivity gains. I've seen this pattern before in crypto markets—when a protocol loses 40% of its liquidity providers over seven days, the price action doesn't show up in the headline metrics immediately. The bleeding is real, but it's measured in spreads, not collapse.
The U.S. labor market is showing the same structural bleed.
Here is the key finding from my analysis of the Apollo research and its implications for the broader economy—and the crypto ecosystem that depends on human labor:
The $28 billion figure represents roughly 0.23% of the $12 trillion annual U.S. wage base. That's not a rounding error. But it is a signal. A signal that has been present in the data I've been tracking since my days auditing ICO smart contracts in 2017.
When I manually audited EthosCoin's smart contract source code, I identified a critical reentrancy vulnerability the whitepaper obscured. The market didn't want to hear it. The community was in full narrative mode. The same cognitive dissonance applies here: the market narrative says "AI creates jobs" or "AI destroys jobs," but the actual data shows a third path.
AI doesn't displace labor. It reprices it.
The Mechanics of Silent Devaluation
Let me break down the compression mechanism as I understand it, based on my experience building yield models for Aave and Compound during DeFi Summer 2020.
AI tools like GitHub Copilot, ChatGPT, and other large language model deployments boost individual worker efficiency by roughly 30% to 50%. When you give a developer an AI pair programmer, they ship more code in the same time. When you give a copywriter Claude or GPT-4, they produce more drafts in a day.
Simple supply and demand dynamics suggest what happens next.
The output supply increases. The demand for that output does not change. The market price for that labor decreases.
This is not a claim that jobs disappear—the function remains. But the pricing power shifts. The employer, holding the AI tool, captures more of the surplus value. The worker retains employment but loses leverage.
The Apollo research calls this "wage compression." I call it a repricing mechanism.
The Numbers Behind the Squeeze
Let's put some structure on the $28 billion figure.
- U.S. annual wage and salary disbursements: approximately $12 trillion
- Apollo's estimated AI wage compression: $28 billion
- Proportion of wage bill affected: 0.23%
- U.S. corporate deployment rate of AI tools: approximately 20%
At 20% penetration, the impact is already 0.23% of wages. This is the early innings. If AI penetration reaches 60-70% of U.S. firms over the next five years, the wage compression effect could triple or quadruple before accounting for the efficiency gains of the tools themselves.
I don't need to build a complex model to see where this goes. I built a yield model for Aave and Compound in 2020 that proved most high-yield pools were unsustainable arbitrage traps. The same "this looks too good" analysis applies to the labor market: wage growth is below productivity growth, and the gap is being filled by capital.
The Compression Mechanism in Action
The Skills That Get Squeezed
The compression is not uniform. I've categorized the labor market into three tiers based on my analysis of job posting data and AI tool adoption:
- High-skill, AI-augmented roles: Software engineers, data analysts, content strategists who use AI tools effectively. These workers gain efficiency. They might even demand wage premiums initially, as they're scarce.
- Mid-skill, task-automatable roles: Customer service representatives, junior analysts, entry-level programmers. These are the ones getting squeezed. AI tools can handle 50-80% of their tasks. Their bargaining power is eroding.
- Low-skill, physical roles: These are less directly affected by AI—robotics remains a different economics. But the indirect effect through service costs is real.
The wage compression is concentrated in the middle band. That's where the $28 billion is hitting hardest.
The data I've seen on this
- The Employment Cost Index (ECI) has been running at 3.5-4.0% annually.
- Productivity growth has been running at 2.5-3.5%.
- Real wage growth for the bottom 50% of earners has been flat or negative over the past 18 months.
This is not a coincidence. The marginal cost of labor is declining because the marginal cost of AI tools is declining. I've watched the price of GPT-4 API calls drop 80% over the last 12 months. This is a structural shift.
The DeFi Parallel
I'm writing this from the perspective of a token fund investment manager, and I see a direct parallel to DeFi.
In 2020, DeFi summer, we saw yield farming protocols offering 1000%+ APY. The market narrative was "decentralized finance will replace traditional finance." The reality was that most of these yields were unsustainable emissions schedules designed to bootstrap liquidity. The protocols were building on the same.
I published a report called "The Illusion of Yield" that showed how most high-yield pools were unsustainable arbitrage traps. I cited specific transaction volume anomalies. The market ignored it for months, and then the music stopped.
AI's wage compression is the same type of structural mispricing. The market narrative focuses on the headline AI milestones—GPT-5, Sora, autonomous agents. The underlying economics of wage repricing is the slow-moving variable that will determine the long-term health of the economy.
The 280 billion is the yield curve inversion of the labor market.
The Crypto Connection
This is a crypto publication, and you might be wondering why this matters.
The answer: AI and crypto are converging.
1. AI Agents are Coming for the Crypto Workforce
We're already seeing AI agents deployed in crypto operations. Automated market making, liquidity provision, and even governance participation. These agents don't demand wages. They consume compute. The marginal cost of an AI agent is the gas fee plus the API call.
When I look at the labor that goes into crypto projects today—community managers, business development, technical writers, junior engineers—I see the same compression pattern.
A community manager earning $60k-$80k a year can be partially replaced by an AI agent that provides 24/7 support, does sentiment analysis, and writes basic updates. The role doesn't disappear entirely. But the wage for the role will compress.
I've seen it happen. I audited a project that had a 12-person marketing team in 2022. In 2025, they're down to 5 humans and a suite of AI tools. The output is similar. The wage bill is less than half.
2. The $28 billion wage compression is a subsidy to crypto
Here's the counterintuitive angle: AI's wage compression is actually a form of capital for the crypto ecosystem.
When companies spend less on wages, they have more capital for other investments. A portion of that capital is flowing into crypto assets. I've seen this in the institutional flows data. As companies report better margins (due to AI efficiency), they allocate a small portion of that to crypto treasury positions.
The Bitcoin ETF flows in 2024-2025 were partially driven by this dynamic. Corporate profits were up, partly due to AI. They allocated a fraction of the incremental cash into digital assets.
3. The "Computational Sovereignty" thesis
I've been writing about a thesis I call "Computational Sovereignty"—the idea that institutional capital flowing into crypto ETFs creates a stable foundation for AI-driven on-chain agents.
The wage compression narrative is the macro backdrop for this. As AI compresses wages, the traditional employment model becomes less reliable. More individuals will seek alternative stores of value. Bitcoin as a non-correlated asset becomes more attractive. The "bankless" narrative gets a tailwind from the "AI wage compressed" narrative.
The Contrarian View
Now let me take the other side of this argument.
The $28 billion figure may be understated
I've seen this in my work with yield protocols. The stated APY was often the "straight-line" yield. The real yield, including the effects of impermanent loss and market volatility, was often much lower.
Similarly, the $28 billion figure is likely the "direct" wage compression. It doesn't include:
- The "hidden hours" effect: Workers spending time learning AI tools, retraining, and dealing with the cognitive load of AI integration. This is unpaid work.
- The "quality" effect: The shift from full-time to contract/zero-hour work. AI makes it easier for firms to hire on a task-by-task basis, reducing the wage premium for stable employment.
- The "algorithmic wage discrimination" effect: AI systems can now evaluate a job seeker's reservation wage in real-time. This allows firms to offer the minimum acceptable wage, rather than a fair wage.
When I add these indirect effects, the annual compression figure could be $40-$60 billion.
The "startup boom" is a mirage
The Apollo research suggests that AI lowers the cost of entrepreneurship. That's true. You can build a software product with AI assistance for $50,000 instead of $500,000. But the same AI tools are available to everyone. The barrier to entry is lower for everyone.
Lower barriers mean more entrants. More entrants mean more competition. More competition means lower margins. The result is a "startup bubble"—more businesses launched, but fewer survive.
I've seen this pattern in the NFT space. When I was tracking Bored Ape Yacht Club and other PFP projects in 2021, the narrative was "everyone can launch an NFT collection." They were right. Anyone could launch one. But that didn't mean they would succeed. The failure rate was over 90%.
AI is doing the same for startups. It lowers the floor. It does not raise the ceiling.
The policy response will be reactive
The American policy response to AI wage compression is likely to be reactive, not proactive.
I've been tracking the policy signals. The ECI data is being monitored. The Federal Reserve has mentioned "AI's impact on the labor market" in its minutes. But there's no concrete policy framework for addressing wage compression.
Expect 5-10 years of lag. By the time the policy response arrives, the structural shift will already be complete.
The Takeaway
The $28 billion wage compression is a data point that deserves your attention. It's not a job apocalypse. It's not a utopian productivity boom. It's a repricing of human labor.
The question is not whether AI will replace workers. It's who captures the value created by AI.
From my position as an investment manager, I see the following:
- The wage compression will continue. AI adoption is only 20%. The penetration will reach 60-70% over the next 3-5 years. The $28 billion figure will grow to $60-$100 billion.
- The crypto ecosystem will benefit. Lower labor costs mean higher corporate margins. Higher margins mean more capital for crypto allocation. This is a tailwind for the entire crypto market.
- The labor market will become more volatile. Wage compression is not evenly distributed. The middle-income workers will feel it the most. This creates social tension. I expect to see the first "AI wage protest" within 18 months.
- The "skill premium" is the only defense. Workers who can use AI tools will earn more. Workers who can't will earn less. This is the new dividing line.
A question for your portfolio
As you look at your crypto portfolio, consider this: Are you positioned for a world where AI compresses wages and crypto benefits? Or are you positioned for a world where the compression creates social unrest and political backlash?
The data suggests the first outcome is more likely. But I've seen enough black swan events to know that the second is never off the table.
Data over drama. Always.
About the Author: Ethan Johnson is a Token Fund Investment Manager with 17 years of industry observation experience. He previously published technical risk assessments on ICO vulnerabilities and yield models for DeFi protocols. His focus is on the intersection of macro trends, AI, and crypto markets.