Notes Toward a Theory of Markets That Have Decided Not to Care
Notes Toward a Theory of Markets That Have Decided Not to Care
Thursday, April 16, 2026
There is a category of market behavior that doesn't fit neatly into any framework. Not irrational exuberance, not repricing of fundamentals, not rotation or reversion. Something closer to collective dissociation. A decision — made by no one in particular and enforced by everyone in aggregate — to simply stop weighting certain facts.
We are living inside one of those episodes right now.
The Nasdaq just posted its best 11-day stretch in recorded history. Eleven consecutive sessions of gains, the index logging 1.59% on Wednesday alone to sit at fresh record highs. The S&P 500 crossed 7,033. Oil is above $100 a barrel. The IMF took a red pen to global growth this week. Input prices just jumped the most in months. And the market's answer to all of it was: Broadcom up 27% in eight sessions. Micron up 31%. AMD up 25%.
The war in the Middle East, which began in late February and which prompted a US naval blockade of the Strait of Hormuz, has apparently been processed, categorized, and filed under "not our problem" by every algorithm running money on the Nasdaq.
What is actually happening here deserves more scrutiny than it's getting.
Elon Musk posted on X Wednesday that Tesla's chip design team has completed the tape-out of the AI5 chip. Tape-out — the point at which a chip's design is finalized and handed off for manufacturing — is a genuine engineering milestone, not a press release. Musk thanked Samsung and TSMC for their manufacturing support, mentioned that AI6 and Dojo3 are already in development, and noted that AI5 will power both Tesla's autonomous vehicles and the Optimus humanoid robot. TSLA surged 7% on the day.
Set aside the valuation — at $389 per share against a GF Value of $254, Tesla is trading at a 53% premium to its estimated intrinsic worth, which in most eras would be considered alarming. The more interesting observation is structural: the AI5 tape-out, by design specifications, produces a single chip with performance comparable to Nvidia's Hopper architecture, and a dual-chip setup approaching Blackwell levels — at meaningfully lower cost and power consumption. If those numbers hold at scale, Tesla has quietly built its way into the semiconductor conversation not as a customer but as a competitor. Not to Nvidia's market, exactly, but adjacent to it. Which is how these things usually start.
Meanwhile, Meta announced it will deploy 1 gigawatt of custom AI chips built using Broadcom technology. One gigawatt. The language of power generation, applied to compute. AVGO jumped 3% on the news, extending a rally that has now put the stock up 27% in just over a week. Broadcom's analysts are projecting annual revenue surpassing $102 billion for fiscal 2026, with earnings per share up more than 67% year-over-year. Insiders have sold roughly $88 million worth of stock in the past 90 days. They know something, or they don't need to know anything — either way, the selling is real and the buying by the public continues.
This is the texture of a momentum market. The fundamentals justify the direction. The speed justifies the skepticism. Both things are true simultaneously, which is precisely what makes it maddening.
TSMC holds its Q1 earnings call today. Revenue of $35.7 billion is already confirmed — up 35% year-over-year. The call is not about the past quarter. It is about whether TSMC's management, which has the most honest real-time view of AI hardware demand of any institution on earth, will maintain its $52-56 billion capex commitment for 2026. That number is the load-bearing wall of the entire AI infrastructure thesis. If it holds, the build-out continues. If there is any softening — any hedge, any "we continue to monitor the environment" language — the multiple that sits underneath half the Nasdaq will develop a crack.
Because here is the thing about a market that decides not to care about a war: it can change its mind very fast. The VIX closed below 20 earlier this month for the first time since the conflict began, after peaking above 35 in late February. That is an enormous compression in implied volatility in roughly six weeks. The options market, in its cold probabilistic way, has concluded that the worst is behind us. Maybe it has. Tentative ceasefire noises are in the air. But oil is still above $100. The IMF's adverse scenario — which it published quietly at its Spring Meetings this week — puts global inflation at 5.4% and GDP at 2.6% if the blockade persists. The base case is already 3.1%, down from 3.3%. The Fed futures market is pricing only a 35% probability of any cut in 2026.
A market up 33% year-over-year with the Fed on hold, oil above $100, and global growth being revised down is making a very specific bet. The bet is that AI-driven productivity gains will materially offset the energy shock and the tightening of financial conditions. It is not an insane bet. It might even be right. But it is a bet, dressed in the clothes of a consensus.
The historical analogy that keeps surfacing — and that no one wants to invoke directly, because the last person to invoke it confidently was usually wrong — is 1999. Not the irrational exuberance of valuation multiples exactly, but the structural resemblance: a genuine technological transformation underway, real earnings growth in the leading companies, and a market that has collectively decided the transformation is so consequential that normal risk frameworks don't apply. In 1999 that belief was correct about the technology and wrong about the prices. The internet did change everything. It just took longer than expected, and the path ran through a 78% decline in the Nasdaq first.
The difference this cycle — and it is a real difference — is that the earnings are actually there. Nvidia posted $215.9 billion in revenue for fiscal 2026, up 65% year-over-year. Broadcom is tracking toward $102 billion. These are not concept stocks. The cash flows are real. The capex being directed at this infrastructure is real. The 1999 companies were selling promises. These companies are selling shovels to people who are actively digging.
But the prices already know all of that. The prices are forecasting several additional years of this, at scale, with no interruption. Micron up 31% in eight days is not the market discovering new information about Micron's business. It is the market repricing the probability distribution of the AI buildout's duration and magnitude. Which is a perfectly rational thing to do. And which, historically, has also been one of the most reliable setups for a violent reversal when any single input — a TSMC capex cut, a hyperscaler earnings miss, a chip export restriction, a Hormuz incident that doesn't resolve tidily — comes in below the embedded expectation.
Netflix reports after the close today. PepsiCo reported this morning. Neither matters much in the current macro context, except as data points in the consumer health picture. What matters is whether TSMC's call this afternoon contains the word "uncertainty" more than twice in the forward guidance section. Count them if you're listening live.
The market has decided not to care about a war. It has decided AI will compound its way past a stagflationary energy shock, past a dollar losing its safe-haven reflexes, past an IMF that keeps quietly trimming its numbers. Maybe the market is right. Maybe this is what it looks like when a genuine technological step-change overrides the macroeconomic cycle for a sustained period.
Or maybe this is what it looks like just before someone remembers what oil above $100 has historically done to corporate margins, consumer spending, and central bank optionality.
The 11-day record run is real. So is everything the market has chosen not to look at.