In Part One, I discussed some of the existential economic concerns that Artificial Intelligence forces us to consider. In this second part, I’ll focus more directly on the practical, near-term landscape of familiar economic forces.
The Economic Future from and of AI
This will be part one of a two part series. In the first part, I want to outline some of my views about how salient a set of what we might call existential concerns about AI should be. In part two, I want to discuss some more immediate interactions with today's economy.
In the near term, AI's integration into the economy will manifest through familiar channels, primarily impacting labor markets, financial investment, and industrial competition.
A more novel danger is what I call 'The Great Masking': the risk that intense investor optimism in AI is hiding serious, slow-building threats to the wider economy. Those threats are trying to hide their effects already, and are eager to utilize AI investment as another layer of protection, to mask the effects of tariffs or irresponsible fiscal policy.
Delayed Effects and Dangerous Unpredictability
Economists have warned of tariffs' impact on the US economy. The stock market has dropped when policies were announced, recovered when policies partially undone, or delayed, and then reached new records, while continued new announcements and implementations occurred.
Labor Market Disruption: Beyond Simple Job Churn
Historically, technological advancements have caused job disruption, not a net reduction in jobs. Some roles, companies, and even industries shrink or fail, while others are created to meet new or previously unmet needs. We don’t have reasons to think that AI is breaking this pattern. While we have reasons to wonder about this in the long-term, those are the existential concerns we covered in part one. In the present day, we’d need more concrete data. This data is limited and not conclusive at this point.
For instance, Derek Thompson supports the idea that AI is already shifting employment among recent grads, but he is careful to note that overconfidence is unadvisable.
When looking at recent data, it's crucial to distinguish AI's impact from other confounding factors. For instance, a perceived slowdown in hiring for software developers—a key data point in the discussion around recent grads—may be less a result of AI replacing jobs and more a consequence of the post-pandemic hiring rebalancing in the tech sector, which saw furious hiring in 2020-2021 followed by a correction. This highlights the difficulty in isolating AI's specific effects from broader economic trends.
To make the case that AI is breaking the prior patterns, you’d need to make the case for structural unemployment. This is not merely the temporary friction of workers moving between jobs, but a fundamental, lasting mismatch between the skills held by the workforce and the skills demanded by an AI-driven economy. If AI automates a wide swath of cognitive and manual tasks, a large segment of the population may find their skills devalued, creating a challenge that standard economic churn cannot easily resolve. This practical concern, if it grows large enough, becomes the mechanism for a more existential crisis.
While that conceptually is sound, it’s applicability depends on data. Weak data suggesting shifting employment among recent grads doesn’t suggest structural unemployment, much less attribute it to AI rather than other economic conditions.
Investment, Bubbles, and Financial Contagion
The development of AI is fueled by staggering levels of investment, creating both immense opportunity and significant financial risk. The source of this capital is critical to understanding the potential fallout.
Investment from Corporate Profits: When tech giants fund AI development from their vast cash reserves, the primary risk is an opportunity cost. If these investments fail to generate expected returns, shareholder value will fall, but the direct impact on the broader financial system is relatively contained.
Investment from Bank Loans: When investments are funded by bank loans and other debt instruments, the risk of contagion is much higher. If AI companies fail to meet lofty profit expectations, they could default on these loans. A wave of defaults could destabilize the lending institutions, forcing them to tighten credit across the entire economy. This is how widespread unemployment can occur. A person laid off may seek a new job or try to start their own business, but if loans for new ventures aren't available, job creation stalls. It is this failure to create new jobs, not just the disruption of old ones, that leads to prolonged unemployment.
This leads to the risk of an AI investment bubble. This doesn't require investors to be irrational, but rather for individually rational views to create a collectively irrational market. For example, one group of investors may rationally believe Company A will dominate the market, while another group rationally believes Company B will. If both groups invest heavily based on their beliefs, the aggregate market valuation can reflect a future where both companies win, an impossible outcome. The market as a whole may be pricing in a cumulative level of future profit that the industry cannot possibly deliver due to competition. The dot-com bubble of 2000 serves as a pertinent example of a financial collapse sparked by over-enthusiasm for a transformative technology, even one that ultimately delivered immense productivity gains.
Market Structure: Layers and Concentration
The AI industry is developing in distinct layers: hardware providers (e.g., AI chips), compute providers (cloud services), model providers (foundational models), and workflow integration (applying AI to specific industries). Competition within and between these layers will shape the distribution of profits. A key uncertainty is who will capture the value at the workflow layer.
Startups: New, specialized startups could focus on different niches, leading to a differentiated market with high potential profits for the successful disruptors.
Cloud Providers: The large cloud computing companies could extend their offerings, creating products for each market. This would likely lead to less differentiation as providers feel pressure to match each other's offerings.
Existing Companies: Incumbents could develop their own AI solutions internally. In this case, profits that might have gone to startups would remain within these established firms.
While these layers are distinct, there will be pressure to break them down. Cloud providers, for instance, have a strong incentive to expand into the other layers. The economics of AI—with massive upfront R&D costs and powerful network effects—create a tendency toward market concentration.
It’s useful not to overstate this point though. Multiple layers allow for multiple points where concentration, and the effects of concentration can be weakened.
Model providers are only as good as their latest model, and so far current investments haven’t provided a clear moat. Mostly AI companies with early success have stayed in the forefront by ever increasing investment, not by resting on laurels. New entrants have arrived and shown results that would have fully obsoleted early investments, at a fraction of those costs.
Amazon, Azure and GCP may remain the primary compute providers, but three competitors are enough to retain a competitive market, and other avenues for compute will remain. In the extreme scenario that competitive forces were muted, the efficiency factor from that scale is still within a single order of magnitude, limiting the exploitative capacity here.
NVidia has had great success, but stays in that place through continued investments. If they ever did hit a brick wall in terms of ability to improve performance, there’s no reason not to expect competitors to reach that same wall quickly, and thus have complete competitive equality.
But above all, each lower layer can only charge something less than what higher layers can capture in terms of value. A higher layer cannot pay for services it has not gathered enough revenue to pay for. With workflow at the top, and the most diversified of all the layers, this is a strong force toward a more broad distribution of economic surplus.
A Counterforce: Defensive Competition and Economic Surplus
For some areas, like cloud providers, forces toward concentration aren’t demonstrably stronger than existing forces. Investing in AI is not simply an attempt to build a new business, but one to retain a competitive position in an existing one. This dynamic could ultimately lead to more economic surplus being more generally by society through usage, rather than by companies through the various means of capturing revenue. This surplus can be:
Direct: Users may get free services or receive far more value than the price they are charged for those services.
Indirect: Many smaller businesses, empowered by more accessible AI tools, become more efficient or capable and pass on those savings or enhanced capabilities to their own customers.
While that is a great deal for consumers generally, it’s a risk to investors. If they assumed a different outcome, or followed others without considering the outcome, their paper wealth will decrease. To the degree that investors are individuals or entities that can survive that risk, it’s not something we need worry about, but that’s not always the case with investors. To some degree we are those investors, but even more, we interact with a system that could be disrupted by turmoil.
The Great Masking: How AI Optimism Obscures Deeper Economic Risks
While the financial risks within the AI sector are significant, a more novel and under-discussed danger is how the AI investment boom may be masking other serious, slow-building threats to the economy. The intense market optimism and capital flows generated by AI can create a "sugar high," temporarily hiding the destructive effects of tariffs, trade disputes, and irresponsible fiscal policy.
The mechanism of this interaction is perilous. These external factors, such as tariffs breaking down global cooperation or budget deficits driving inflation, act as a persistent drag on the economy. They increase costs for businesses and reduce purchasing power for consumers. In a normal environment, the negative effects of these policies would be more visible in economic data. However, the powerful forward-looking optimism of the AI boom can overwhelm these signals, keeping investment sentiment high and stock market valuations buoyant.
This phenomenon aligns with observations from economists like Paul Krugman, who notes that markets are often poor at pricing in long-term policy risks, allowing them to build until a crisis becomes undeniable. An AI boom and tariffs both fit this narrative of market myopia, focusing attention on future technological gains while downplaying present-day policy costs.
Matthew Yglesias expands on this, explaining why this effect may be particularly severe now. The logic is that traders believe Trump is so sensitive to the stock market that if a sell-off occurs, Trump will back down. This creates a powerful incentive for investors to ignore the initial disruptive action and "buy the dip," confident that the policy will be reversed. This confidence, however, can prevent the very market crash needed to trigger the policy reversal. This dynamic, where the market's faith in its own influence creates a dangerous tolerance for risk, is amplified by the sheer optimism surrounding American AI companies. Since the major AI players are American, and are perceived as having the support of the political establishment, investors are willing to overlook institutional risks that would cause panic in other circumstances.
The danger is that these problems don't disappear; they fester beneath the surface. This sets the stage for a potential cascade of concurrent failures. An eventual, and likely inevitable, correction in AI valuations would not happen in a vacuum. Instead, it would act as a trigger, suddenly exposing the underlying weaknesses that the boom had papered over. This could lead to a multi-pronged crisis:
A Financial Shock: The AI bubble deflates, erasing wealth and crushing investor confidence.
A Corporate Shock: Businesses, already weakened by higher input costs from trade barriers, face a sudden drop in demand and tighter credit, leading to widespread insolvencies.
A Consumer Shock: Households, whose purchasing power has already been eroded by inflation, face the added threat of mass layoffs.
This "perfect storm" scenario, where a financial correction and a real-economy crisis hit simultaneously, is far more dangerous than either event occurring in isolation.
Too many triggers, and weakened foundations
What’s more, either can create the precipitating event. Bad policy could induce stagflation or a recession, spook AI investors and force them to accept that expected profits are now more distant. AI would still succeed, but the financing costs will still impact companies. Some may fail, but those that survive will still have a weaker financial position that merits a valuation change. The weaker financial position would cause investments to realign, and slowdown, removing the lift it’s been providing to the economy and exacerbating a policy induced recession.
On the other hand, an adjustment to optimism about long term effects can happen at any time. This could be a rational adjustment, or an irrational one. Normally an irrational adjustment would be temporary, but if it triggers a break in an unstable foundation, the normal recovery would not occur. Even a logical, healthy correction in AI stocks could trigger a catastrophe. Investors who thought they had safety nets would discover those protections were eroded by bad policy, causing a much wider economic collapse.
Finally, you have how the Trump administration would react to this. TACO (Trump Always Chickens Out) logic presumes that Trump can undo the damage done. That works far better for a policy announced yesterday that has not been put into effect. It does less for one that’s been slowly eating away at the American economy for months, maybe even years. In that case, which policy do they reverse? All of them? While that would be my choice, it presumes a lot to assume it would be the Trump administration's choice.
It might instead think it can placate markets as it has in the past with partial reversals of the most recent policies. This of course would not work, as something fundamental would have shifted, and so only something more fundamental can correct it. We also have signs that the Trump administration would take a page from China’s playbook and create the numbers it wants, and hope it can fool enough of the public to restore optimism. There’s a chance that might work once or twice, but ultimately if the bad policies remain, it only allows the foundation to erode further, creating a yet more future crisis.
Conclusion: The Risk of a Great Unmasking
While the long-term, existential questions surrounding AI command attention, the most immediate and tangible threats to the economy are rooted in the complex interplay between the current investment boom and other festering economic problems. The path to any future, utopian or otherwise, must first pass through a period of significant short-term risk, where the greatest danger is not a single point of failure, but a cascade of them.
The central, under-analyzed threat is that the AI investment bubble is actively masking the slow-building damage from irresponsible fiscal policy and the breakdown of global trade. Should the bubble burst, it will trigger a "Great Unmasking." The initial financial shock from deflating tech valuations will be immediately compounded by the sudden exposure of a real economy already weakened by inflation and supply chain friction.
This creates the potential for a multi-faceted crisis where a credit crunch, a collapse in business investment, and a surge in unemployment all happen concurrently. This is not just a tech-sector correction; it is the risk of a systemic economic downturn sparked by the tech sector but amplified by pre-existing conditions the boom helped to hide.
Therefore, while navigating the long-term societal transition is crucial, the immediate priority for policymakers must be to recognize and address this dangerous interaction. It requires looking beyond tech-sector regulation to the interplay of fiscal, trade, and financial stability policies. Successfully managing the practical, interconnected turbulence of today is the absolute prerequisite for realizing the profound promise of tomorrow.