When AI Eats the Bottom Rung of the Career Ladder
Ben Lorica and Evangelos Simoudis analyze three pivotal AI trends: the "Great Hollowing Out" of entry-level jobs, the financial disconnect between AI hardware depreciation and its useful life, and OpenAI's strategic shift to router-based models in the quest for a sustainable business model.
Evangelos Simoudis and Ben Lorica explore several critical trends shaping the AI industry, from its impact on the workforce to the financial and business model challenges faced by leading companies.
The Great AI “Hollowing Out” and the Future of Work
A significant phenomenon, termed the “great hollowing out,” is emerging where AI-driven automation is making it increasingly difficult for new college graduates and early-career professionals to secure entry-level positions. This trend disrupts the traditional talent pipeline, as companies risk losing their primary source for future senior talent.
This hollowing out affects different segments of the white-collar workforce in distinct ways:
- Entry-Level, Low-Skill Jobs: Roles involving tasks like answering routine emails or writing simple scripts are already being replaced by capable AI agents.
- Entry-Level, High-Skill Jobs: Positions such as paralegals, customer support agents, and certain types of programmers are beginning to be replaced, though complete automation is not yet feasible with current technology.
- High-Skill Professional Jobs: While roles like lawyers or investment bankers are not being replaced, AI tools are making them hyper-productive. A junior investment banking analyst, for example, can now produce pitch books much faster, meaning firms may need fewer analysts overall.
Recent data from the New York Fed highlights a concerning rise in underemployment for recent graduates (41%) and college graduates (33%). This is particularly worrisome as these individuals often carry significant student debt and the first few years of a career are critical for building a strong resume.
The automation of so-called “grunt work” also removes a crucial learning phase. Historically, these tasks served as a rite of passage where junior employees learned the intricacies of their profession through a form of apprenticeship and socialization with senior experts. As experienced professionals retire, there’s a risk that essential skills could be lost if this interaction model is broken, potentially faster than AI can replicate them.
While AI does create new jobs, such as specialized roles in advanced manufacturing or teleoperator positions for autonomous vehicles, a mismatch exists. These new roles often require highly specialized skills that entry-level candidates don’t possess, and furthermore, fewer of these new positions are created than are displaced. This leads to an “AI doom loop”: automation eliminates jobs, applicants use AI to apply for the remaining ones, and companies use AI to screen the flood of applications, making it harder for everyone, especially those at the start of their careers.
For corporations, the short-term financial gains from reducing headcount through AI must be weighed against the long-term strategic risk of a depleted talent pipeline. This may require companies to completely redefine what an entry-level position entails and reimagine modern apprenticeship programs.
The Hidden Hardware Crisis: AI Asset Depreciation
A critical financial issue, highlighted by short-seller Jim Chanos’s observations on Meta, is the discrepancy between accounting practices for AI hardware and its real-world value. Companies like Meta are depreciating specialized chips over 11-12 years, a standard practice for traditional server hardware. However, the cutting-edge useful life and resale value of these chips are closer to two or three years, after which they are technologically surpassed.
The counterargument is that older hardware can still be used for less demanding tasks, such as inference, much like consumers continue to use older iPhones. Organizations will likely develop a tiered approach, using the latest, most expensive chips for training and cutting-edge research while repurposing older hardware for production inference and other tasks.
The key indicator to watch over the next few years will be the allocation of capital budgets. Observing how much capital is spent on replacing existing hardware versus adding new capacity will reveal the true economic lifecycle of these assets and whether current accounting methods are accurately reflecting their value.
The Router Revolution and the Search for a Business Model
OpenAI’s recent redesign of the ChatGPT interface, which replaced granular model selection with a router-based system, signals a pivotal strategy shift in the AI industry. This move is a clear indication that foundation model providers are aggressively experimenting to find a sustainable business model beyond simply growing user numbers.
This search is analogous to Google’s journey in the early 2000s to develop its advertising model. Foundation models are currently unprofitable on a per-user basis, even with subscriptions, and the ability to raise massive capital rounds will not last indefinitely.
A router allows a provider to:
- Value Queries Differently: A query for a local restaurant has a different commercial value than a query for a medical specialist. The router can dynamically assess a query’s intent and potential for monetization.
- Optimize Resource Allocation: It can route simple queries to smaller, cheaper models and complex ones to more powerful, expensive models like GPT-4, thereby managing computational costs effectively.
- Enable New Revenue Streams: This system paves the way for introducing paid recommendations or sponsored content for high-value queries, creating a business model that captures the value delivered to third parties (e.g., the business being recommended).
This strategic shift is not just about cost containment; it’s about building a long-term economic engine. It is predicted that other major players, including Google, Anthropic, and xAI, will inevitably follow suit, adopting similar router-based approaches to manage costs and monetize their powerful AI systems.