The total US market index sits near $74.7 trillion, which is about 230% of domestic GDP and the reason every valuation screen currently reads red. Divide the same numerator by world GDP of roughly $115 trillion and you get something closer to 65%. Neither number is wrong. They answer different questions, and the second one is closer to the question an owner of the index is actually asking, because the companies in it do not sell to the United States. They sell to the world and they book the profit in Delaware. The relevant scarcity is not American output. It is the supply of durable claims on global cash flow, and that supply is smaller than it looks.
Consider what has happened to the listed sector. The United States had somewhere near 8,000 public companies in the mid-1990s and has roughly half that today. The market has not grown by adding participants. It has grown by concentrating value into a shrinking set of survivors while the global pool of savings looking for a home has expanded without interruption. Pension systems, sovereign funds, defined-contribution flows that arrive on a schedule and do not consult a valuation model, foreign allocators treating US equity as the reserve risk asset. That is a demand curve meeting a supply curve that has been sloping the wrong way for thirty years. A high price is what that looks like. It is not obviously a temporary one.
Which points at the first mechanism for growth that involves no repricing at all. The largest private companies in the AI complex are not in the index. When they list, market capitalization rises by the full value of the new listing, the Buffett Indicator prints a worse number, and nothing has become more expensive. The metric counts listings, not valuations, and it treats an expansion of the investable universe identically to a speculative melt-up in existing shares. A meaningful wave of AI-era IPOs would push the ratio toward 250% while making the market broader, deeper and cheaper on a per-dollar-of-earnings basis. The screen would flash danger at the precise moment the opportunity set improved.
The second mechanism is duller and more important. Prices can stay flat and multiples can fall while market capitalization rises, provided earnings do the work. This is what 2011 through 2019 looked like for large parts of the index and it is the only healthy way out of an elevated reading. It requires the profit pool to grow into the price rather than the price to fall back to the profit pool. The question is therefore not whether 230% is high, which it plainly is, but whether the denominator of the earnings multiple is going to expand fast enough over a five year horizon to make the current level look like an entry point rather than a peak. That is a forecast about corporate profits, not a forecast about a ratio, and treating the ratio as though it settles the matter skips the actual analysis.
There is a third channel that most of the bearish commentary misses entirely, and it works on the bottom of the fraction. The dot-com comparison fails because that capex cycle was thin. Fiber and routers, a rounding error against national output, and it was over in two years. What is being built now is physical. Data center shells, turbines, transformers, grid interconnects, transmission, fabrication plants, cooling, water infrastructure and the construction labor to assemble all of it. Every dollar of that is counted in GDP, in the quarter it is spent, on the denominator side of the very ratio being used to argue the market is untethered from the real economy. A capex cycle of this composition mechanically raises nominal GDP. The ratio can compress from below without a single share changing hands, which is the opposite of how a bubble resolves and is not a possibility the 1999 analogy contemplates.
Where the headroom actually sits is worth being specific about, because it is not evenly distributed. The extremity of the aggregate reading is concentrated in perhaps ten names carrying an unprecedented share of index weight. Strip them out and the remainder of the market trades at multiples that are elevated against history but not remotely at record extremes. That concentration is usually presented as fragility, and it is, but it is also the shape of an unfinished move. If the AI buildout does what its sponsors claim, the earnings revisions eventually arrive in power generation, electrical equipment, industrial construction, memory, cooling, and the international suppliers of all of the above, none of which are priced for it. A market can rise a great deal on rotation alone, and rotation is the mechanism by which an index with an expensive top and a reasonable middle grows without the top going anywhere.
The honest limits on all of this should be stated. Sixty-five percent of world GDP is a lower number than 230% but it is still a record for that series, and using the global denominator is a legitimate correction rather than a get-out-of-jail card. The listing-supply argument cuts both ways, since new supply also absorbs the bid that is currently supporting existing shares. Rotation into industrials and power assumes those firms capture the spending rather than competing it away, which is a strong assumption about businesses that have historically earned their cost of capital and no more. And the physical capex that flatters GDP is the same capex being depreciated over schedules that may not survive contact with the obsolescence rate of the hardware, which means the earnings holding up the numerator carry a question mark of their own.
The decision-relevant series here is net equity supply. Buybacks have exceeded new issuance for most of the past decade, which has been a quiet and underappreciated tailwind, and the private AI cohort listing at scale would flip that for the first time since the last cycle. Watch the quarterly Fed Z.1 corporate equity issuance and retirement figures alongside the IPO calendar. If issuance stays suppressed while buyback authorizations keep clearing, the supply squeeze that produced 230% is intact and the number is not the ceiling anybody thinks it is. If the pipeline opens and net supply turns positive, the bid gets tested for the first time in years, and that, rather than any valuation screen, is what would mark the top.
