AI created its own inflation, its own job cuts, its own regulation, and its own grid problem. The market figured out which companies sit on which side of the line.
MARKET PULSE
For most of 2026, AI was the trade. This week, it became the weather.
The S&P hit fresh highs. The Nasdaq did the same. Beneath those headlines, AI was creating costs that showed up everywhere at once. In jobs reports. In bond yields. In White House calls. In the price of power.
The week ran one experiment over and over. Which companies have AI revenue that lands? Which companies have AI costs without the revenue yet? The market sorted them ruthlessly.
By Friday, a clearer picture emerged. AI is no longer one part of the economy. It is the part driving every other part.
Here are the six things that mattered most.
PREMIER FEATURE
The REAL Reason Trump Is Invading Iran
For a moment…
Forget about Trump’s ties to Israel.
Forget about reports of Iran’s nuclear program.
Because my research has led me to believe we’re risking World War 3 with Iran for a completely different reason.
If you have even a single dollar invested in the U.S. stock market, this is going to directly impact you.
THREAD 1
The AI Economy Grew 31 Percent. The Rest Grew 0.1 Percent.
Greg Ip published the most important number of the week.
Strip AI capex out of first-quarter GDP and what is left grew 0.1 percent. The other 31 percent? That came from data centers, chips, and software. Two economies are running side by side. One is booming. The other has stopped.
Hyperscaler capex hits 3.3 percent of GDP next year, more than the U.S. spends on defense. Mag-7 first-quarter earnings rose 61 percent. The other 493 companies in the S&P 500 rose 16 percent.
Labor's share of business output fell to 54.1 percent. That is the lowest since records began in 1947.
The Takeaway
The market is at all-time highs because seven companies are doing all the work. That is not a forecast. It is the data. Anything that disrupts AI capex hits everything currently working in the economy.
THREAD 2
The White House Reversed Eighteen Months of AI Policy in One Week.
Trump's administration spent eighteen months telling AI labs to build without restrictions. That changed Tuesday.
VP JD Vance was briefed on Anthropic's Mythos model. The model finds software vulnerabilities and could launch cyberattacks against hospitals, water plants, and small banks. Vance was alarmed. So was Treasury Secretary Bessent. So was Secretary of State Rubio.
By Thursday, the White House was drafting an executive order to formally review AI models before release. NEC Director Hassett compared the proposed framework to FDA drug approval. White House adviser David Sacks called the concern overblown publicly. Two senior officials, opposite views, same president.
OpenAI consulted the administration before previewing GPT-5.5-Cyber. Anthropic did not have that option. The Pentagon blacklisted the company in March.
The Takeaway
For the first time in this cycle, Washington is pulling against AI capex instead of with it. The rules are not written yet. When they are, the capability ceiling they set becomes the most important number in the industry for the next decade.
FROM OUR PARTNERS
The Government Just Unlocked a $500 Trillion Opportunity
A small government task force just finished a 20-year project.
They probably didn't realize their findings would allow everyday citizens to stake a "claim" on a $500 trillion national treasure.
But they did. And under U.S. law your "birthright claim" is now active.
This opportunity won't stay under the radar for long.
THREAD 3
Three Companies Announced Major AI-Driven Layoffs In One Afternoon.
Thursday afternoon, three public companies announced mass layoffs in the same window. Each one named AI as the reason.
Cloudflare (NET) cut 20 percent of staff. Internal AI usage rose 600 percent in three months. Upwork (UPWK) cut 25 percent and said two-pizza teams are dead. Bill Holdings cut 30 percent and authorized a $1 billion buyback the same day.
The reactions split. Bill rose 7 percent. Cloudflare fell 19 percent. Upwork fell 18 percent. Same announcement template. Opposite results.
The difference was simple. Bill had a working business before the cuts. Cloudflare and Upwork were already losing ground. The market priced AI layoffs as a validation in one case and a confession in the other.
AI now drives 16 percent of all 2026 job cuts, up from 13 percent a month ago.
The Takeaway
The AI capex cycle is now showing up as cash flow, not capex. The first quarter of investor reports that tested whether the cuts came from confidence or weakness is now graded. The market is no longer paying for the layoff alone.
THREAD 4
Memory Chips Became A Channel For Monetary Transmission.
This week, memory chips stopped being a supply story and started behaving like a monetary one.
Samsung's semiconductor division produced 94 percent of total company profit. SK Hynix and Micron (MU) are entering the global top ten. The three combined are set to earn $350 billion in 2026, with Samsung second only to Nvidia (NVDA).
Here is what changed. Memory price increases are now flowing directly into corporate capex budgets, then into hyperscaler costs, then into PCE through technology and services components. Meta (META) raised full-year spending by $10 billion citing memory specifically. GE HealthCare (GEHC) cut its profit forecast for the same reason. Tim Cook warned Apple (AAPL) memory costs will be significantly higher after June.
That is not a chip cycle. That is a price signal moving from one Korean fab into the Federal Reserve's preferred inflation gauge.
The chokepoint is concentrated. Two Korean firms and one U.S. firm hold the entire supply curve. The U.S. has no leverage over the Korean half except through diplomacy. China is racing to break in. Sovereign competition for memory capacity is already underway.
The Takeaway
Memory has crossed from being a tech input to being a macro input. When three factories control the price of the most important component in AI infrastructure, their pricing decisions reach inflation data, monetary policy, and sovereign trade strategy. That is a different kind of constraint than the market is used to pricing.
FROM OUR PARTNERS
AI CEO Issues Code Red: Prepare for Meltdown
The CEO of this AI company (click here to get the name, 100% free) just issued a CODE RED in an internal memo…
Warning his employees that they’re dealing with a critical situation.
Another company executive even implied they might need a government bailout.
And now Jim Rickards is predicting this company is about to go bust, in a full-blown AI meltdown that could be 10 times bigger than Lehman Brothers.
THREAD 5
The Bond Market Said AI Is The New Oil.
The assumption all year has been straightforward. Iran war ends, oil falls, inflation cools, the Fed cuts. This week, that argument lost its strongest support.
Pimco published research showing core PCE has separated sharply from core CPI in 2026. The gap traces to AI infrastructure costs feeding into the services and technology components weighted heavily inside PCE. The Fed leans on PCE more than CPI when setting policy.
CME odds shifted. Zero rate cuts this year. A 17 percent chance of a hike. The 10-year Treasury yield held at 4.39 percent. The 30-year approached 5 percent.
Even if oil falls, the AI input layer keeps inflation sticky. That changes the rate path argument from "war ends, cuts resume" to "war ends, but yields stay high."
The Takeaway
Every duration-sensitive trade in equities runs through this. The bond market priced a higher floor under yields that does not lift even if Iran resolves. That is the floor every other risk asset trades against now.
THREAD 6
The Power Grid Said It Could Not Keep Up.
PJM Interconnection runs the largest power grid in the U.S. It serves Northern Virginia, the densest data center region in the world. This week, PJM CEO David Mills wrote that the situation is not tenable.
The white paper said PJM has years, not decades, to overhaul itself. American Electric Power (AEP), one of PJM's largest utilities, said it is considering withdrawal. The grid has 220 gigawatts of new interconnection requests sitting in queue.
Natural gas turbines are back-ordered into the early 2030s. Solar and batteries can install faster but face their own China-link approval freeze. Nuclear takes a decade.
Every hyperscaler capex projection assumes the grid clears. PJM is openly saying it cannot, on the timeline that matters.
The Takeaway
The physical bottleneck on AI is no longer chips. It is electrons. The Morgan Stanley (MS) capex number and the power that number requires are not currently reconcilable. Whoever solves the gap captures the rent. Whoever depends on it gets squeezed.
FROM OUR PARTNERS
The Strait Headlines Are Everywhere—But This Isn’t
The headlines are loud right now—oil, volatility, uncertainty.
But while attention shifts, something else keeps working quietly in the background.
An overlooked investment that’s compounded at an extraordinary rate over time.
Most people never even look at it.
CLOSING LENS
This week, AI stopped being a sector inside the economy. It became the operating system the rest of the economy runs on.
The S&P is at a high because Mag-7 earnings are up 61 percent. The Fed cannot cut because PCE is sticky. Layoffs are accelerating because efficiency is the new growth story. Washington is moving because capability outran governance. The grid cannot deliver because demand outran planning. Memory is transmitting price into monetary policy.
All of it is one tape.
Last week was about whether AI revenue was real. This week answered that question and asked a harder one. What happens when AI's own costs start binding faster than its revenue can absorb them?
The market does not know yet. The data arriving over the next four weeks does.
