How much AI can earn depends on how many employees the company lays off?
```
AI commercialization is entering the “revenue realization” period, with the most direct signal being: job replacement is being quantified.
According to Guojin Securities’ latest report citing Anthropic’s actual observation data, in the United States, AI has already affected about $1.45 trillion in salary scale and 18.35 million jobs; according to OpenAI’s theoretical exposure criteria, the corresponding figures soar to $5.68 trillion in salaries and 68.28 million jobs.
For comparison, total US payroll is about $10.83 trillion, with total employment around 155 million people. Calculated this way, the proportion of jobs AI can replace or reshape ranges from as little as 10% to as much as 40%.
With the accelerating rollout of AI coding, enterprise-level agents, and B-end applications, AI is evolving from an “efficiency tool” into a “cost variable” — savings are now directly converting into business profits, and the logic of capital returns is getting firmer.
Yet at the same time, structural contradictions are becoming more prominent: corporate costs decline, profits rise, while employee jobs shrink and roles are replaced. The deeper AI goes, the clearer the capital gains, but also the heavier the employment squeeze. Like a seesaw, when one end rises, the other must sink.
Technical layoffs have gone from theory into reality
Data shows the US tech sector has re-entered a high-layoff cycle. The single-round layoff rate among large tech companies generally falls between 5% and 10%; some software, SaaS, and cloud companies even approach 20%. But what’s noteworthy isn’t just the scale of layoffs itself; the capital market is now clearly “rewarding” this behavior.
In the current valuation logic, “AI cost-reduction + layoffs” is interpreted by the market as a positive signal of efficiency improvement and profit enhancement. By using AI to replace manpower, companies not only cut operating costs but also gain direct returns in their financial reports and stock prices. This positive feedback is turning AI job replacement from a technical option into a capital-driven strategic inertia.
In other words, technical layoffs are no longer just a side effect of efficiency tools, but are actively pushed into a positive feedback loop by market mechanisms—the more layoffs, the better the profits look, the stronger the stock price, and the greater the motivation companies have to keep laying off.
If this incentive mechanism doesn’t change, AI’s job replacement effect will not be a one-off hit, but a sustained, self-reinforcing, long-term process.
High-skilled jobs aren’t necessarily safe; low-skilled job impact is also huge
Guojin Securities, based on an analysis of 755 occupations from the US Bureau of Labor Statistics (BLS), shows AI’s impact on the job market does not simply follow a linear “low-skill first” logic, but exhibits a dual divergence trend.
In terms of absolute numbers exposed, low-skilled jobs number about 8.83 million, high-skilled jobs about 7.09 million, and mid-skilled jobs about 2.44 million. But in terms of employment percentage, high-skilled jobs affected reach 19.5%, significantly higher than low-skilled jobs at 10.8% and mid-skilled at 8.3%.
This shows that high-skilled jobs are also under substantial pressure and are not the “safe zone” the market had expected.
However, in the “actual exposure/theoretical exposure” dimension, which reflects real-world deployment speed, low-skilled jobs lead. The exposure progress for low-skilled jobs is 31.2%, higher than high-skilled at 27.9% and mid-skilled at 20.4%.
This means low-skilled jobs are first to face the reality of AI application, while high-skilled jobs face more reconstruction pressure in the long-term expectation layer, with a clear mismatch in their stages of risk.
Facing customers doesn’t determine job safety
Facing customers is not a natural shield for job security. The report further indicates that what truly determines AI’s impact intensity is not “whether it’s a front-office job,” but whether tasks can be standardized and digitized.
Divided by work object, jobs can roughly be categorized into three types: front-office jobs directly interacting with customers, like retail sales, customer service, cashier; back-office jobs focused on R&D, data processing and administrative operations, like software engineers; mid-office jobs responsible for coordination and management, like HR and supply chain management.
But this traditional classification is becoming obsolete. Analysis shows front-office attributes don’t equate to “job protection.” For low-skilled jobs, being front-office means higher AI exposure, with sales jobs being typical — standardized scripts, process-based communication, and decision paths are quickly being covered by model capabilities.
Contrary to intuition, the “front-office attribute” for high-skilled jobs forms a kind of buffer. For instance, professions like teachers, doctors, and lawyers, which rely on high-frequency interpersonal interaction and complex judgment, are more difficult to fully replace in the short term, thus delaying the direct transmission of AI impact.
What AI truly tests may not be efficiency, but distribution
Guojin Securities believes that current productivity gains from AI are mainly from capital deepening, i.e. improvement of labor tools, rather than genuine total factor productivity growth. Data shows that since 2024, US labor productivity and total factor productivity have diverged.
Therefore, a more worthy concern than efficiency is how technology dividends are distributed.
US households already face multiple pressures: declining labor income share, increasing share of government transfer payments in income, and weak actual income growth. If AI dividends further concentrate in capital and a few top talents, both employment and income distribution pressures may intensify accordingly.
The report suggests overseas debates about universal basic income (UBI) may resurge, but their precondition is government sharing in AI-generated returns. The report notes that the US’s current idea is for government to directly hold equity in tech companies, thus partaking in AI dividend distribution, possibly as a way to relieve social income stress.
For capital markets, this means the AI story is no longer just about computing power, models, and revenue growth, but will gradually enter new stages of employment, distribution, and social governance. The amount of profit AI can generate will likely depend on how much it can replace human labor; whether AI investment continues to attract capital market favor will also increasingly depend on whether society can bear the cost of this process.
Risk DisclaimerThe market carries risks and investment requires caution. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are applicable to their circumstances. Investing accordingly is at your own risk. ```