Here is a sentence that would have sounded insane five years ago: the near-term direction of the American stock market, a meaningful slice of the economy, and by extension a piece of your retirement account will be decided Wednesday afternoon by what one man says about graphics chips. Wall Street knows it, which is why Nvidia just fell seven days in a row — its longest losing streak in four years — while everyone squares up their bets before Jensen Huang speaks.
Monday was the tell. The Dow rose 140 points. Quiet day, right? Except the Nasdaq lost 0.8% and the Philadelphia Semiconductor Index — the scoreboard for the chip complex — fell 2.7%. Sandisk (SNDK) dropped 6.5%. Ciena (CIEN) fell 6%. Micron (MU) lost 5.8%, Marvell (MRVL) 3.3%, Intel (INTC) 3.1%. Over in Seoul, Samsung tumbled 8.7% after announcing a shareholder-return plan worth up to $79 billion that somehow still disappointed people. When a $79 billion give-back reads as a letdown, expectations have left the atmosphere.
The company that buys its own customers
What makes this Wednesday different from the last eleven blowout quarters is the structure underneath. Nvidia is no longer just selling shovels to the gold rush — it has started financing the miners. This month it teamed up with six of Wall Street’s biggest firms on a $500 billion AI-financing plan, pledging to backstop lending to customers who otherwise couldn’t afford its chips. It took a stake in Cloverleaf Infrastructure, which arranges power for data centers. And it struck a $6 billion deal with a startup called Poolside — $1 billion invested at a $12 billion valuation, plus $6 billion to license the technology and hire most of the engineers.
Poolside’s own shareholder letter tells you why they said yes: at the end of last year the company had six weeks to raise $2 billion to pay for a cluster of 40,000 Nvidia servers. It missed the deadline and lost the cluster. The company selling the chips became the only buyer able to keep the customer alive. That is either vertical integration or a snake eating its tail, and Wednesday’s call is where analysts get to ask which.
The strategic logic is real, to be fair. The Poolside deal funds a new version of Nvidia’s open-weight Nemotron model, aimed squarely at China’s DeepSeek and Kimi K3 — models that trail the American frontier by only months and cost a fraction to run. Huang has argued, in his first-ever post on X, that America wins the AI race “not by one frontier AI model” but by building the open ecosystem everyone else builds on. He may be right. He is also now competing with OpenAI and Anthropic — the biggest buyers of his chips.
What we’re actually watching
Not the revenue number. Nvidia will report a record; it always does. We’re listening for three quieter things: how much of the demand is financed by Nvidia itself, what Huang says about the pace of orders from the debt-funded data-center builders, and whether the guidance leans on customers whose own finances are wobbling — OpenAI just told investors its revenue grew a tepid 18% last quarter while losses deepened.
And the market’s own posture is worth reading. The volatility index (VIX) rose almost 5% Monday to 15.85 — still low, but rising into a day when the Dow went up. That’s the options market quietly buying rain gear while the sun is out. The crowd isn’t panicking; it’s hedging. So are we.
Our position is unchanged and deliberately boring: we own the AI theme through the model books, we own the power and infrastructure that feeds it through companies like Chevron (CVX) with contracted data-center power, and we are not adding to any of it in the forty-eight hours before a binary event. If Wednesday goes well, we’ll still be here Thursday. If it goes badly, we’ll be very glad we were patient. That’s not a forecast; it’s a seatbelt.