Tesla just secured $30 billion in new credit, and it hasn't touched a cent of it. With capital spending set to top $25 billion this year, the real question isn't whether Tesla can raise money. It's whether it will need to. Here's what to watch when Tesla reports next.

Nvidia just approved the biggest buyback increase ever — $150 billion on top of its existing program, for a total of $235 billion. But a buyback authorization is permission, not spending. To use it all by fiscal 2028, Nvidia would have to roughly double its record repurchase pace. Here's the math, the 3 scenarios, and the numbers that will decide whether this is a real capital-return machine or just a very large permission slip.

The Nasdaq hit another record, but beneath the surface, Meta's new AI agent split the market in two. Chip stocks surged while Schwab, Airbnb, Uber, and Lyft sold off on fears that AI agents could take over the customer relationship. The real question now: does that pressure spread?

AMD just crossed $1 trillion for the first time, but it didn't rally alone. Chip stocks surged, the Nasdaq set a record, and the AI trade may finally be broadening beyond Nvidia. Here's what's driving the move, and the test AMD faces next as it tries to hold the trillion-dollar line.

Nvidia reports Wednesday after the close with roughly $280 billion of market value in play — more than the individual market cap of 90% of S&P 500 companies. But the implied 5.4% move is smaller than Nvidia's recent average, and the stock just posted its seventh straight down day. The real variable isn't in the income statement. It's the 30-year Treasury yield sitting above 5% at a 19-year high, quietly repricing every data-center project the AI trade depends on.

Bloomberg reported Saturday that some of Nvidia's largest customers were told server prices will rise more than 15%, driven by soaring memory-chip costs. Nvidia hasn't confirmed it, and the stock hasn't traded on it yet. The real story isn't the price increase — it's that the AI bottleneck is moving from compute to memory, and the bottleneck always owns the margin.