How much money AI can earn depends on how much it can take from the human wage pool.

```

How much money can AI large-model companies really make? Guojin Securities’ latest report provides a disruptive answer: Stop focusing on the software market—look at humanity’s payroll.

Companies buy AI not to be trendy, but to save money. Replacing part of human labor with AI, increasing efficiency, and cutting costs—this is the real reason why companies are willing to pay. Therefore, the real ceiling of AI revenues is not the size of the software market, but the size of the wage pool that can be repriced by AI. Guojin Securities calls this the "wage pool that can be repriced by AI."

Guojin Securities' latest research report did the math: Within the annual total payroll of about $10.83 trillion in the US, $1.45 trillion is already exposed to the impact range of AI—that is, the work content of these jobs can be done by AI, or AI can assist in a significant portion.

How much do AI companies earn from this money? Take leading company Anthropic as an example: annualized revenue is about $47 billion, only 3.2% of the $1.45 trillion. In other words, it’s barely a drop in the bucket.

The wage pool, not the software market, is the ARR's valuation anchor

Guojin Securities' report points out that the most intuitive way to understand the "epic growth ceiling" of this round of AI revenues is to calculate how big the "wage pool that can be repriced by AI" actually is.

The report matches the exposure of different occupations to AI technology with the 830 positions in the US Bureau of Labor Statistics (BLS) 2025 Occupational Employment and Wage Survey (OEWS 2025). Results show that, among about $10.83 trillion in total US payroll, based on Anthropic’s actual observed exposure, about $1.45 trillion in wage costs are within AI technology’s exposure range, accounting for 13.4%; using OpenAI/Eloundou’s theoretical exposure, the potential impact could reach about $5.68 trillion, more than 52%.

Based on employment numbers, among about 156 million employed people in the US, actual exposed population is about 18.35 million, accounting for 11.8%; theoretical exposure reaches about 68.3 million, accounting for 43.9%.

The report emphasizes that $1.45 trillion in wage costs should be understood as "the ideal upper limit of ARR income under current penetration and technological capability," and that this ceiling is discounted—companies may only need $10,000 in AI spending to equivalently replace $100,000 in labor costs. Even so, the ARR of current large-model businesses, in the tens of billions range, is still at a very low penetration rate compared to the size of the wage pool above.

AI impact shows "high wage bias," knowledge jobs are first in line

Unlike past automation, which mainly impacted manufacturing and repetitive manual labor, this round of AI more directly touches high-wage, knowledge-intensive and service jobs.

Report data shows that the theoretical exposure of occupations to AI technology is markedly skewed right compared to annual average salary distribution—high-income groups face significantly higher AI exposure than low/mid-income groups. For example, among the lowest income percentiles (such as laundry workers, bakers, tire mechanics), AI exposure is generally low; among high-income groups, Financial Product Managers (income percentile 96.6%, exposure 78.6%), HR Managers (income percentile 95.3%, exposure 76%), and Aerospace Engineers (income percentile 92.5%, exposure 89.3%) all face higher replacement risk.

By industry, the top three industries in theoretical exposure are computer & math (87.6%), business & finance (78.2%), and law (78.0%). However, the actual observed exposure ranks differently: the highest are computer & math (35.3%), office & administrative support (33.2%), and sales-related positions (24.6%).

This gap reveals that AI’s replacement of labor is not solely determined by model capabilities, but also constrained by job attributes, responsibility, and organizational processes. Legal work involves coordinating interests, litigation strategy decisions, and lifelong responsibility; financial services rely on client relationships and non-standardized information judgment. In contrast, programming jobs, with clear objectives and short feedback loops, see faster actual replacement progression.

Computer industry treats "all equally," finance sector shows obvious differentiation

Among the top 20 jobs with the highest actual exposure, 8 belong to computer and math, involving about 1.59 million employed, accounting for 30.2% of the industry total. The report states that for the computer industry, wage level and AI exposure are not necessarily linked—when facing AI impact, the whole industry is treated nearly "equally," highlighting overall vulnerability under technological iteration.

The financial sector, meanwhile, shows distinct differentiation. Some roles require responsibility (e.g., audit, accounting), and the degree of standardization varies greatly among positions, so overall actual exposure in finance is low, yet there is noticeable internal differentiation. Of these, Market Research Analysts have 64.8% actual exposure, Financial and Investment Analysts 57.2%, both facing high replacement risk; while other roles needing client relationship maintenance and non-standard judgment have lower exposure.

Looking at total wage exposure, the $1.45 trillion in actual exposed payroll is mainly concentrated in five sectors: office & administrative support ($289.6 billion), business & finance ($247.4 billion), management ($221.7 billion), computer & math ($215.2 billion), and sales-related roles ($199.5 billion). The report suggests that this offers direction for specialized large model B2B business development: for certain success, focus on administration, computer, finance—sectors where replacement is already apparent; for "zero-to-one breakthrough," education and medical diagnosis still have major potential.

Replacement does not equal unemployment, but wage restructuring is underway

The report clearly distinguishes "exposure" from "replacement": exposure means tasks may be AI-assisted, automated, or reorganized, but it does not mean these wage incomes will disappear proportionally. The real determinant of AI’s economic impact remains the speed of enterprise adoption, boundaries of model capability, organizational process transformation, and regulatory constraint.

However, the report also notes, AI’s macro impact will not simply show as a linear decline in jobs. The more likely path: some single-duty jobs replaced, many multi-duty jobs restructured; some wage costs compressed, more labor processes repriced. Especially since AI agents have the feature of "the higher the wage, the higher the replacement rate," the potential impact of AI on consumer-side income may be even deeper.

For investors, the report’s core conclusion is: the medium-term space for AI revenues should not be calculated only from the software market size, but should be anchored to the much larger labor cost pool. Current ARR penetration for large-model companies remains extremely low, but the flip side of this coin is that human wage structure is facing a systemic restructuring that has yet to be fully priced in.

Risk DisclaimerThe market has risks, investment needs caution. This article does not constitute personal investment advice and does not take into account individual users' special investment goals, financial situation, or needs. Users should consider whether any opinions, views or conclusions in this article fit their specific situation. Invest at your own risk. ```