Overview
The markets can be doing well while risks build underneath them. That is the tension I want to explore. The success of American businesses and the promise of AI are real. So are the questions about how concentrated that success has become, how it is financed, and how resilient it would be if conditions changed.
I am not rooting for a downturn. I invest, build businesses and believe in what people can create. I want to understand what is supporting this strength, where it may be vulnerable and what would change my view. You do not have to agree with my politics to agree that those are useful questions.
The period since the financial crisis can feel like an almost unstoppable run. But it was not one uninterrupted expansion or bull market. The United States had a short, severe recession in 2020, and stocks have suffered major setbacks. Recessions are not due simply because a certain number of years has passed. My concern is about the structure beneath the growth, not the calendar.1
My prediction is that more risk is embedded in this economy than many people appreciate. I can imagine one or two smaller recessions before a much more serious crisis, particularly if financing problems eventually limit the response. That is my prediction, not an established sequence or a reliable timetable. These risks could develop differently, and productive growth or policy changes could materially improve the outcome.
The biggest companies: real achievement and concentrated power
Large U.S. companies, particularly the growth and technology businesses that dominate market attention, have been powerful wealth creators over much of the post-financial-crisis period. The precise comparison depends on the index, dates and whether dividends are included. No asset class wins every period. Still, the concentration of attention and capital around the biggest businesses deserves scrutiny.
In my view, this is more than an AI story. Scale can produce real advantages: better products, global reach, research budgets and efficient distribution. But scale can also create power over competitors, suppliers, workers and customers. When competition becomes weaker, I worry that companies can protect margins in ways that would be harder in a more genuinely competitive market.
I also worry about the feedback loop between economic power and political influence. Successful companies can spend to shape the rules, and favorable rules can reinforce their position. That is an argument about incentives and institutions, not a claim that every large company is a monopoly or every lobbying dollar produces a measurable return.
The distinction matters. A profitable company can be an excellent investment and still raise difficult questions about the system that made its profits possible. Investors may benefit from pricing power at the same time that households struggle with the bills. I want us to be willing to examine both sides rather than pretending one cancels out the other.
This connects to the distinction I explore in the rebelution between creating value and extracting it. I do not think the answer is to stop people from building successful businesses. It is to ask whether the rules reward productive work and genuine competition or increasingly protect incumbents. That is my institutional interpretation, not proof that market returns are fabricated.
Near-term risk: when financing reinforces itself
My prediction is not that I know the date or trigger of the next downturn. It is that there is more risk embedded in this economy than many people appreciate. The question is not simply whether AI is useful or whether successful companies have real businesses. They do. The question is how much of today’s expansion depends on financing arrangements and expectations that reinforce one another.
When suppliers also invest in customers, or support demand for the infrastructure built around their products, I want to understand the connections. Which revenues come from durable demand by end users? Which commitments depend on another financing round? Who ultimately bears the risk if the projected demand fails to arrive? It is possible for arrangements that make commercial sense individually to create a more concentrated exposure collectively. That is the concern I want to investigate, not an accusation that every investment is a disguised loan or that the accounts are fraudulent.
I also worry about the interaction with public deficits, government purchases, tax policy and the wealth effect. Defense and healthcare are part of my concern about an economy that can rely on spending patterns that may prove hard to sustain. Higher asset prices can support confidence and consumption; if those prices reverse, some spending may reverse too. Real productive wealth does not disappear merely because market prices fall. But paper gains can shrink, financing can become harder to obtain, and the same connections that supported growth can transmit stress. I do not yet know which link would break first.
There are documented examples of overlapping roles worth examining. CoreWeave’s September 2025 filing described NVIDIA as supplier and shareholder and a conditional purchaser of residual cloud capacity, under an order initially valued at $6.3 billion. OpenAI later announced $30 billion of NVIDIA investment in February 2026 alongside compute collaboration. These establish relationships; they do not establish that NVIDIA concealed loans, manufactured revenue or committed fraud.101112
The earlier September 2025 intention to invest up to $100 billion was a letter-of-intent announcement, not $100 billion of verified cash already lent. I would not add that figure to the later $30 billion announcement as though they were independent, fully funded exposures. What matters is understanding the actual commitments, financing, customers and risks, with the latest filings in hand.
AI can succeed while some investments disappoint
I believe AI is real, important and likely to help people do extraordinary things. That is different from believing every company, every data center and every investment made during the boom will earn an attractive return.
The analogy I keep coming back to is the buildout of the internet. A transformative technology can attract too much capital, at the wrong prices, into overlapping infrastructure. The technology can go on to benefit society while some investors and businesses suffer badly. The internet did not become unimportant when the technology bubble burst.
That is a scenario I can imagine for AI: enormous long-term usefulness alongside disappointed near-term returns, overcapacity in some areas, restructurings and a period of reassessment. I am not claiming all AI businesses will fail, that every project is overbuilt, or that the timing is knowable. I want to distinguish technological progress from the return on the capital being committed to it.
The questions to follow are practical: Are customers getting enough value to pay sustainably? Are revenues and productivity gains catching up with the spending? How much of the buildout depends on optimistic assumptions about future demand? What happens if funding becomes more expensive or investors become less patient?
Counterevidence belongs here, not in the fine print. AI could improve productivity rapidly enough to justify substantial infrastructure investment. Demand might broaden, costs could decline and firms could finance expansion from durable cash flow. If those things happen, the overbuild thesis weakens. I want to follow actual utilization, customer value and cash generation, not just the size of the announced projects.
Wealth, spending and household resilience
One explanation I am watching is that spending by wealthier, asset-owning households can keep the economy moving even when many other households are stretched. Rising stocks and home values can support confidence and spending. They can also create collateral for borrowing.
That does not mean the typical consumer has no money, or that every affluent household is borrowing to live. It means the distribution matters. The Federal Reserve’s 2025 survey found that 55% of adults reported a rainy-day fund covering three months of expenses. That leaves a substantial group without that buffer, but it does not support saying the average American can barely cover one month.4
If wealthier households account for a large part of discretionary demand, a reversal in their confidence or asset values could have effects well beyond their own balance sheets. They may postpone renovations, travel, services or major purchases. Businesses dependent on that spending may then reduce hiring or investment. This is a mechanism to investigate, not proof that those consumers alone are holding up all growth.
I want to watch whether apparent resilience comes increasingly from productive income growth, or from drawing on accumulated wealth and credit. Those are not the same thing.
The Federal Reserve’s analysis of the 2025 Survey of Consumer Finances gives a useful boundary: the top tenth of families, ranked separately by income or wealth, account for about 28% of its measured spending. That is disproportionate, but it is not most spending. Its measure excludes rent, mortgage payments and retirement contributions. It also does not tell us which households drove the latest increase in consumption. I should not use a spending-share snapshot as proof that wealthy borrowers alone are sustaining the economy.7
There is real wealth here, not simply an illusion. Assets can support households and businesses through hard times. The question is how much of the confidence built on those assets survives a change in prices, financing conditions or income.
Home equity is wealth; HELOC balances are debt
The Wall Street Journal chart I was referring to shows nearly $36 trillion in homeowners’ equity. That is accumulated value after mortgage debt, not $36 trillion already borrowed. The chart, sourced to the Urban Institute, stacks mortgage debt on top of equity to show roughly $50 trillion of single-family housing value. The screenshot does not give an exact ending quarter or a tabulated debt endpoint.9
Separately, outstanding HELOC balances were $481 billion in the third quarter of 2026, up $23 billion from the prior quarter. That roughly half-trillion-dollar number is borrowing on home-equity credit lines. It is not all mortgage debt, fixed home-equity loans, unused credit capacity or new borrowing during the quarter.59
HELOC balances are rising toward their previous peak. They could meet or exceed it if the trend continues, but that outcome is not certain. The chart still puts current balances below the roughly $700 billion financial-crisis-era high. “Within a couple of years” is a possibility I raised, not a forecast established by the graph.
A household may retain a low fixed-rate first mortgage while adding a variable-rate equity line. That can provide flexibility, but it also creates another payment and rate exposure. I want to know the amount drawn, payment burden, underwriting standards, delinquencies and concentration of risk before treating the rise alone as a crisis signal.
The article’s other charts require care. One counts borrowing choices by the year the original mortgage was issued, not total borrowing over time. Another covers cash-out refinance borrowers’ uses of money through Q3 2025, with multiple answers allowed. It is not evidence of how all HELOC borrowers spend, nor a breakdown of the dollars borrowed.
My personal experience of frequent business-credit offers makes me wonder how aggressively lending is being marketed. That is an observation, not a national lending survey. Home equity also remains a cushion; rising credit-line balances alone do not erase it.
How leverage can turn a setback into a chain reaction
The value of an asset can fall quickly. The debt taken out against it generally does not fall with it. That is where a normal change in expectations can become something more serious.
A hypothetical chain looks like this: asset prices weaken; some borrowers face margin requirements or tighter credit; they sell assets or reduce spending; those actions put pressure on other prices and businesses. This is not a forecast that every household or bank will behave this way. It is a reason to distinguish wealth from liquidity and an unlevered holding from a leveraged one.
- 1Asset prices weaken
- 2Some borrowers face margin calls or tighter credit
- 3They sell assets or cut spending
- 4Pressure spreads to other prices and businesses
I am also concerned about the cultural line between investing and betting becoming harder to see. Buying a productive asset is different from taking a position whose payout depends on a single event. Event contracts can have information or hedging uses, but the experience can also encourage speculation and make risk feel like entertainment. Options and futures likewise have legitimate uses; that does not make every retail use prudent.
Illustrative event questions might involve an election result, a policy-rate decision, a sports result, a weather threshold or an economic release. These are categories, not claims that a particular live contract is legal or offered on a named platform. My concern is not that all these instruments are identical. It is that people can underestimate how much risk they are taking when the interface makes everything look like another investment.
The comparison that worries me is not a claim that today is literally the 1920s or that another 2008 must occur. It is the recurring possibility that confidence, rising asset prices and easy access to risk-taking reinforce each other until conditions change. A good analysis would look for the channels that could transmit stress, and also for the safeguards that could contain it.
Medium-term pressure: affordability and the social bargain
A lower inflation rate does not mean groceries, housing or medical care have returned to their old prices. It usually means prices are rising more slowly. That difference is obvious to the person paying the bills, even when the economic headline sounds reassuring.
I do not think it is accurate to say nobody’s wages have kept up with inflation for thirty years. Broad real-earnings measures have risen over that period, though unevenly, and national averages do not describe every worker. The more important question is whether income gains are enough for the actual mix of necessities facing a household, especially housing, health care and other costs that vary by place and stage of life.2
My concern about corporate pricing power remains, but I do not want to hang it on a misleading statistic. Saying “companies raised prices” describes inflation; it does not, by itself, tell us why. Costs, shortages, demand, tariffs, energy shocks and profit margins can all matter. The contribution of profits is debated and differs by period and industry. There is research finding that aggregate markup changes were not the principal driver of the post-pandemic inflation surge. That evidence needs to sit alongside my concern, not be hidden from the reader.3
The larger point is about bargaining power and who can absorb a shock. A household with appreciating assets and flexible spending has options. A household whose income is already committed to essentials may have very few.
Even without an immediate financial crash, the affordability problem deserves its own horizon. Families need housing, food and healthcare. If those costs continue to strain their budgets while automation changes jobs and bargaining power, financial pressure can become a deeper loss of confidence in the system.
My concern is that rising home and investment values have helped some households tolerate an economy that otherwise feels increasingly expensive. What happens if those gains reverse, or if homeowners cannot afford repairs or access credit? People who previously felt they were participating may decide the arrangement no longer works for them. I worry about a much more divisive generational and socioeconomic conflict if a broad share of the country reaches that conclusion.
That is my warning, not a measured forecast of social conflict. Automation could raise productivity and create new work as well as displace jobs; the distribution of those gains matters. Nor is “we have built no housing” literally true. The question is whether enough appropriate housing is being built where people need it, at prices they can afford.
I used a hypothetical majority, perhaps 55% to 70%, turning against the economic arrangement to describe the kind of political shift I fear. That was not polling or a modeled probability. The argument is about a possible change in social confidence, not a claim that such a threshold has been measured.
Healthcare: a cost problem and a transition problem
Healthcare is both essential work and a major burden on families, employers and government. Treating every additional dollar spent as prosperity misses the question of what people receive for it. But treating all of the sector’s growth as waste is just as misleading.
BLS’s 2025 accounting gives a concrete reason to consider transition risks. Health care added 387,000 jobs and the separate social-assistance industry added 299,000, together more than the economy’s net payroll increase of 116,000 because losses elsewhere offset gains. That is a narrow-growth story, not evidence that every new job was healthcare. October 2026 also showed gains outside healthcare; the period matters.68
If we decide the cost trajectory is unacceptable, reform affects employment and revenue as well as access. Hospitals, clinicians, insurers, pharmaceutical firms and investors have different economics. Lowering administrative costs, prices or unnecessary care cannot be assumed to affect everyone in the same way. An aging population and genuine care needs are also part of the picture.
My proposal
In the near term, I would make universal healthcare a priority, alongside strengthening Social Security. I am not proposing simply to have government pay the same inflated bill we already have. The cost of the system itself has to come down, with access and quality protected.
One funding approach I want to explore is an additional payroll contribution through a reformed FICA system, in the range of 4 to 6.5 percentage points, combined with taxes on very large fortunes and on the economic gains from AI and robotics that replace human labor. I want a system that pays for its commitments instead of adding another open-ended burden to public debt. We don’t know the exact number yet, which is why I give a range; these funding ideas need to be modeled together.
The gap I want to close
The United States spent about $15,500 per person on health care in 2024. Comparable wealthy countries averaged about $7,900 per person.1314 My goal is to bring what we spend down to where other developed nations are, roughly half of what we spend now, while protecting access and quality.
Note: I had remembered the comparison as about $7,500 for the next-highest country. The verified figure is the peer-country average of about $7,900 per person (2024); the next-highest single country, Switzerland, spent nearly $5,000 less per person than the U.S. The U.S. figure is $15,474 under the official national accounts, or $14,775 on the international comparison basis.1314
The household effect is central to my argument. Imagine a family whose annual healthcare burden falls substantially. That is money it could use for housing, vacations, ordinary purchases, saving or paying down debt. People might also be more willing to obtain care when they need it rather than avoiding deductibles and then facing more serious problems later.
I expect a difficult transition for parts of the healthcare industry if we reduce what the country spends. That cannot be dismissed, especially given the industry’s employment role. But I think the longer-run benefit could be a redistribution of purchasing power toward lower- and middle-income households, with more of their resources available elsewhere in the economy. My argument is that a transition can be worth pursuing, not that every displaced job will be instantly replaced or every family will automatically save.
I would connect this discussion to my existing policy page, If I were in charge, while leaving the current evidence and cost assumptions visible here.
Housing: a practical way to rebuild opportunity
My proposal
For the medium term, I would pursue a serious government-supported housing program. I want more single-family homes, condos and other entry-level options at prices young families can reach, not just more expensive properties that leave the same people outside the market.
My Columbus reference point is how far the price of a home has moved beyond what many younger households can comfortably afford. I would like to see starter options closer to a little over $100,000 rather than simply accepting $300,000-plus as the entry point. That is the goal I want to set: a real starter home for around $100,000.
I would explore subsidies and tax incentives for affordable construction, utility-cost reductions and renewable power such as solar and wind. I would also help new contractors obtain financing to start and expand. Over time, construction automation and robotics could help increase capacity. Those tools need to be evaluated for what they can actually do, not bought on the assumption that a future robot can already replace a skilled building crew.
Labor is part of the plan too. I favor a legal, workable pathway for construction workers to come here, with a system contractors can navigate and lawful housing and working arrangements. Temporary workers should have clear terms for returning home, and people who remain eligible, contribute over time and meet the requirements should have a route toward citizenship. This is a proposal for immigration policy, not a claim that all these pathways already exist.
Building more housing would create demand for labor, materials and infrastructure while addressing a generational problem. It may also reduce existing homeowners’ future appreciation or, in some markets, their current equity. I recognize that tradeoff. My hope is that we can improve access without a disorderly adjustment, but we should not promise that prices will only flatten and never fall.
A supply program should be judged by completions and affordability, not just announced subsidies. Lower prices for buyers may come at a cost to current owners; incentives can raise land prices if supply remains constrained; utility and transport capacity can limit where homes are useful. Those are reasons to design the program carefully, not reasons to accept permanent exclusion.
Long-term risk: a crisis with less room to respond
The longer-horizon risk I worry about is a major downturn arriving when public borrowing is already difficult to manage. If financing becomes substantially more expensive or investor confidence deteriorates just when policymakers want to respond, stabilizing the economy could become harder.
That is not the same as saying the government will simply run out of dollars, or that a debt-market crisis has a known date. Treasury yields, market liquidity, inflation expectations, monetary policy and fiscal choices interact. My concern is about reduced room to maneuver during a serious shock. The short-, medium- and long-term problems could reinforce each other, but they do not have to unfold in that order.
I worry about dependence on deficits, the cost of defense and healthcare, and the eventual fiscal consequences of chronic illness. These concerns need separate evidence rather than a single claim that public spending is all unproductive or that illness is a matter of individual blame.
A dollar-issuing government is not a household. Higher yields, a weak auction, a liquidity disruption and an enduring loss of investor confidence are different developments. My severe scenario is that a major downturn collides with more costly or politically constrained financing, making rescue efforts harder. I do not know that it will occur in ten or fifteen years, and I cannot establish that it would necessarily be worse than 2008.
What could change my mind?
I want this page to evolve, not become a collection of charts selected to prove me right. Evidence that growth is broadening, real incomes are improving across groups, household credit remains manageable and productive investment is generating durable cash flows would make me less worried about fragility.
If AI spending produces strong customer value, sustainable revenues and broad productivity gains, that weakens an indiscriminate bubble thesis. If home equity remains substantial, debt service stays manageable and delinquencies remain contained, that weakens the case that rising HELOC balances alone imply a crisis. If housing completions improve access and healthcare reform preserves quality while lowering total costs, that strengthens the policy case.
I would also revise my proposals if the funding math fails, if a measure reduces investment or access more than expected, or if better evidence suggests another approach. We should compare consequences, not just intentions.
My view is that broad participation is part of economic strength, not a charitable afterthought. I want people to build, invest and succeed. I also want an economy in which more people can afford the basics, withstand setbacks and participate in the gains. Strong markets are welcome. A stronger foundation underneath them is the goal.
Sources
These sources support specific, bounded observations, not every interpretation on this page. Where I offer a prediction, an illustration or a proposal, I say so. Company-specific relationships can change, so I check the latest filings before updating them.
- NBER, U.S. business cycle expansions and contractionsRecession chronology, including the 2009–2020 expansion and the February–April 2020 recession.
- BLS via FRED, median weekly real earnings (LEU0252881600Q)Full-time wage and salary workers, quarterly, 1982–84 dollars: 310 in Q3 1996 and 367 in Q3 2026. Not a constant-worker panel or a household cost index.
- Federal Reserve Bank of San Francisco, “Are Markups Driving the Ups and Downs of Inflation?” (May 2024)Finds sector differences, but not aggregate markups as the main driver of the post-pandemic surge.
- Federal Reserve, Economic Well-Being of U.S. Households in 2025 (released May 2026)55% report three months of emergency savings; 63% could cover a hypothetical $400 expense with cash or its equivalent.
- Federal Reserve Bank of New York, Q3 2026 household debt and credit report (November 10, 2026)HELOC balances, credit limits, housing value and homeowner equity are different measures.
- BLS, Employment Situation for October 2026 (November 6, 2026)Payrolls up 165,000; unemployment 4.1%.
- Federal Reserve, FEDS Notes: Measuring spending in the Survey of Consumer Finances (October 2026)Top-decile share of about 28% under separate income and wealth rankings; excludes rent, mortgage payments and retirement contributions.
- BLS Monthly Labor Review, major industry employment changes in 2025Health care +387,000; social assistance +299,000; total net payrolls +116,000.
- The Wall Street Journal, “These charts show how more people are pulling cash out of their homes”Nearly $36 trillion of home equity and $481 billion of HELOC debt; chart dates and classifications differ.
- CoreWeave, SEC filing (September 2025)NVIDIA as supplier, shareholder and conditional purchaser of residual capacity, with an initial $6.3 billion order. Not evidence of concealed loans or improper accounting.
- OpenAI, OpenAI and NVIDIA systems partnership (September 22, 2025)A letter of intent for an investment of up to $100 billion, not verified money disbursed.
- OpenAI, Scaling AI for everyone (February 27, 2026)A $30 billion NVIDIA investment and compute collaboration. Not to be added to the earlier intended ceiling.
- Centers for Medicare & Medicaid Services, National Health Expenditure data (2024)U.S. health spending of $5.3 trillion in 2024, or $15,474 per person.
- Peterson-KFF Health System Tracker, How does health spending in the U.S. compare to other countries?2024: $14,775 per person in the U.S. versus a $7,860 average across comparable high-income countries; Switzerland, the next highest, spent almost $5,000 less per person than the U.S.
