A Plausible AI Economic Scenario

By Deric Cheng @ 2025-10-19T06:19 (+10)

A draft of an economic scenario to be used in future writing. Fully open to thoughtful feedback.

The New AI Oligarchy

Over the next decade, it is becoming clear that 4-6 tech giants will dominate frontier AI development. The economics are unavoidable: building cutting-edge AI models requires immense computational resources and specialized talent that only the most well-capitalized organizations can afford. Only a few major players in the US and China can be competitive on the frontier. 

These technology behemoths will eventually succeed in building transformative AI capable of reasoning and conducting agentic tasks on a human level, and eventually a superhuman level. The eventual deployment of these systems will usher in an era of supercharged economic growth, as they bypass fundamentally human limitations in terms of time, energy, coordination, and intelligence.

There will still exist a sizable gap between the development of these systems and their diffusion into the broader economy, but this gap will be on the order of years, not decades. 

Because of competition and the relatively small costs of deployment, we'll also witness the commoditization of powerful AI systems, many of which will have AGI or beyond-human capabilities. Every person will have a personal superintelligence in their pocket for close to free – much as we are already seeing today. These will almost certainly be provided by the leading AI labs, but open-source alternatives will likely become prevalent as well.

At the same time, it’s likely that many organizations will begin to carve out niches deploying specialized commercial AI systems based on innovations pioneered by the leading labs, or open-sourced models. For example: quite a few corporations will specialize in productionizing AI robotics systems for specific sectors, which the leading AI labs may not have the capacity to develop themselves. 

While not competitive in frontier AI research or developing frontier technological breakthroughs, these firms will excel at: 

• Packaging and delivering AI services to specific markets.
• Conducting domain-specific research dependent on proprietary data collection or iterations.
• Solving the thorny problems of AI deployment and diffusion.

Together, these frontier labs and AI-driven corporations will form a new class of "superstar firms". 

Because of the incredible economies of scale created by existing digital infrastructure and intelligence-on-demand, these corporations will lack the structural barriers to growth that have encouraged more widespread competition in the past. Many of these superstar firms will rapidly end up as dominant entities in their specific sector and region (e.g. providing the majority of transportation services in the US, or dominating healthcare in Europe). 

Globally, individuals will still have independent access to superhuman-level AI models. However, they will simply not be competitive with these “superstar firms” in deploying AI systems at scale into the economy. In the same way it is feasible to run independent drop-shipping businesses, yet nearly all online shopping goes through major portals like Amazon, independent users of AI models will not see significant uptake compared to scalable, cheaper, and better alternatives from major industry players.

Unprecedented Concentrations of Power

Because of the incredible returns to scale, we might end up with only 10 - 20 superstar firms driving the majority of new economic growth in each major region of the world. This would break down into perhaps 1 to 3 organizations per sector (e.g. American healthcare providers) – rapidly consolidating new oligarchies. Many of these will likely be new organizations who have developed entirely novel technologies (e.g. Waymo), whereas some of these will be established incumbents who can leverage their massive institutional advantages and transition it effectively into an AI-driven economy (e.g. a distributed organization like Kaiser Permanente). 

The most striking feature of this transformation will be that these corporations will require remarkably few human employees. A typical AI firm deploying AI software might employ just a couple hundred people while generating billions in revenue. A transportation giant may require just thousands of technicians, rather than millions of human drivers.

AI robotics companies might require significantly more human capital due to raw material extraction, supply chains, and the complexity of manufacturing and physical capital. As a result, they will also see substantially slower growth than AI software deployment. However, they will still see economies of scale due to robotics automation surpassing the most efficient manufacturing processes today. Over decades, the marginal cost of an AI robotics system will eventually fall below the cost of human labor in a wide range of industries.

This will represent a profound shift in our economic structure. Previous generations saw corporations like General Motors or Walmart employing millions to achieve market dominance. Modern tech companies like Google and Meta still employ hundreds of thousands of employees. The coming AI corporations may need just hundreds to conduct remaining non-automatable tasks.

This acceleration of capital efficiency will create unprecedented wealth concentration. This class of superstar firms will grow slowly in human employees, but rapidly in capital, revenue, and share of economic growth. Superstar firms will achieve astronomical revenue-per-employee ratios as AI systems efficiently perform tasks that once required vast human workforces. The economic surplus will flow primarily to capital owners and a small set of talented human employees, but not to the general population.

The Transformation of Work

For human workers, the landscape of opportunities may narrow dramatically. Many jobs will remain in areas requiring human-to-human connection, but even these will be transformed. Consider the field of UX / UI: instead of large teams collaborating on projects, a single designer may serve as an intermediary between stakeholders and AI systems that execute most of the actual design work.

Humans will maintain advantages in face-to-face, offline interactions. Businesses that fundamentally require in-person service will demonstrate resilience to AI automation. However, even traditionally high-touch roles like therapists, teachers, and travel agents will eventually face massive disruption as AI systems handle the majority of routine cases. Human practitioners in many fields may need to be truly exceptional to justify their premium over increasingly sophisticated AI alternatives. For many average workers in customer-facing roles, demand will collapse as AI systems deliver remarkable performance at a fraction of the cost.

Importantly, nearly all remaining cognitive jobs for humans will become hybrids— a mix of leveraging AI systems via clear intent to handle complex tasks, and applying distinctly human capabilities in interpersonal communication.

In Summary

This transformation will dramatically boost productivity and GDP growth in absolute terms. The economy will see massive efficiency gains, and we will see incredible breakthroughs technologically, in scientific domains such as healthcare and material sciences, and overall economic improvements for both developing and developed countries.

However, the distribution of these gains will be profoundly unequal in the long-term, both accelerating trends we've already witnessed over recent decades and creating new dynamics for human labor: 

1. Drastically reduced demand for cognitive labor: As companies discover they can maintain or increase productivity with fewer employees, they'll stop hiring and eventually begin downsizing. Studies show that up to 90% of automation-related job losses occur in the first year of recessions – firms hold onto excess workers in good times, then cut jobs and replace them with machines when crises hit. We expect that a similar trend will occur with automation from AGI.
2. Downward social mobility: For the first time in history, higher education and intellectual capability may no longer guarantee economic security. A significant portion of the middle class will experience downward pressure on wages and career prospects. Many will be pushed into service or manual labor roles that better resist automation, creating oversupply and wage depression in those sectors.
3. Extreme wealth inequality: The overwhelming majority of economic gains will begin to flow to capital owners and the small class of specialized workers who design and maintain these systems. The wealth gap between capital owners and everyone else will reach historically unprecedented levels.
4. Shrinking of the middle class: In service economies, AI will lead to a decrease in demand for many middle-income jobs, across many sectors, all at once. Many of the desirable jobs in overlapping job clusters may see simultaneous collapses, leading to labor displacement with nowhere to go. “New jobs” created by AI automation may themselves be automated by AI systems, with the exception of those requiring human-to-human connection. Real wages may be driven down by this increasing competition.

The result of these trends will be a second "Engel's Pause"—a period where technological productivity gains dramatically outpace wage growth for ordinary workers. 

To be clear: our economy will not collapse. Society will not disintegrate. But without intervention, we will face an acceleration of existing problems—stagnant wages for most people, extreme inequality between capital owners and workers, diminished upward mobility, and increasing precarity for the middle class.

The massive benefits of artificial intelligence will not be distributed equitably through natural market mechanisms. We will need significant intervention to avoid exacerbating our most pressing social challenges.