Good morning.

The man who taught the Federal Reserve to speak in riddles has gone quiet. Alan Greenspan, who chaired the central bank for more than 18 years across four presidents, died this week at age 100. For a generation of traders he was "the Maestro," the man whose every mumble about "irrational exuberance" could move markets before anyone was sure what he'd actually said. He exits just as his old institution turns sharply hawkish under a brand-new conductor, which is its own kind of eulogy: the Fed Greenspan built now belongs to someone else, and that someone is in no mood to cut.

Big Tech

Mark Zuckerberg has watched the internet turn every election, ballgame and box-office weekend into a casino. Now he wants to own the table.

Meta is quietly building a standalone prediction-markets app, internally dubbed Arena, at Zuckerberg's direct order and treating it as a top priority, according to a New York Times report. The app would let users bet on real-world outcomes using video-game-style points rather than cash, at least to start, and would run separate from Facebook and Instagram even as those apps funnel a few billion users toward it.

The points-not-money design is the tell. It keeps Meta clear of the thorniest gambling rules while it builds the habit, and leaves the door open to real wagering once the regulators and lawyers catch up.

Meta is late to a party that has gotten very loud. Prediction markets have grown from a political curiosity into a genuine asset class, and the incumbents already command real money.

Combined monthly volume on Kalshi and Polymarket jumped from under $5 billion last September to more than $20 billion by April, and regulators have begun writing a formal rulebook for the sector, the kind of plumbing that usually shows up right before the institutions do. Meta, with a few billion users spread across its apps, looks at that and sees an engagement engine it doesn't yet own.

House Money: Points today, but few believe it stops there. If Arena catches on, Meta turns the world's largest social network into the world's largest betting floor, with all the engagement upside and every regulatory, addiction and election-integrity headache that follows. The prize is enormous, and so is the odds-on chance it becomes the next thing Zuckerberg gets called before Congress to explain.

Markets

The AI trade spent a year running on helium. On Tuesday a little of it leaked out, and the hiss started in South Korea.

The KOSPI, the world's best-performing major index this year, crashed about 10% and tripped a circuit breaker after a report that SK Hynix may shift capacity away from high-bandwidth AI memory back toward ordinary DRAM, a hint that the most fevered corner of the chip boom may be cooling. Samsung and SK Hynix, nearly half the index between them, did most of the falling. By the time New York opened, the selling had crossed the Pacific.

US stocks took the cue and ran with it. The Nasdaq Composite closed down 2.21% at 25,587.04, with Micron and SanDisk each tumbling more than 13%, while the Dow held nearly flat as money rotated toward staples and utilities. This was less a panic than a question, asked all at once across two continents: what if the spending never pays off?

That question has a number attached. Combined 2026 capital spending across the biggest hyperscalers now tops $452 billion, with Alphabet alone guiding to as much as $185 billion and Amazon flagging roughly $200 billion, even as the cash to fund it gets thinner.

By the numbers:

Buckle Up: A 10% air pocket in Seoul isn't proof the boom is over, but it is the first time the market has flinched at the one thing it refused to question all year: whether the spending earns its keep. Micron's report after today's close and the Fed's favorite inflation gauge Friday are the next two readings on the altimeter. Until then, expect the ride to stay bumpy.

Semiconductors

Qualcomm built an empire inside your pocket. Now it's spending billions to climb out of it. The chipmaker is in advanced talks to buy AI-software firm Modular for about $4 billion, whose tools let developers run AI models across rival chips. It's the loudest move yet in a rapid-fire shopping spree by CEO Cristiano Amon, who is trying to turn a company synonymous with smartphone modems into a credible challenger to Nvidia in the data center.

The logic is defensive and offensive at once. Phone growth has flatlined, and the only corner of chips minting real money right now is AI inference, the grunt work of actually running models once they're trained.

Amon has spent the past year buying his way toward an AI-infrastructure stack, and the tab is climbing fast.

Off the Hook: The reposition is real, and so is the risk. Buying your way into AI infrastructure is expensive, integration is hard, and Nvidia is not a gentle incumbent. Investors have rewarded the ambition so far; the harder question is whether all these logos add up to a coherent rival or just a pricey collection. Tuesday's signal was simple: Qualcomm no longer wants to be judged on handsets.

Markets

Most companies raise debt because they're short on cash. SpaceX did it with a vault that would make a small country blush, and investors fought to hand over more.

Less than two weeks after the largest IPO in history, SpaceX (SPCX) priced a $25 billion debut bond sale, upsized from a $20 billion target after drawing nearly $90 billion of orders at their peak. The five-part deal, maturing between 2031 and 2056 and rated investment grade by all three major agencies, will mostly refinance a bridge loan, with the rest going to fund an AI and satellite build-out that even an $85 billion equity raise can't cover.

The reception in the bond market couldn't have clashed more loudly with the equity tape. Even as SpaceX briefly dipped below its $150 debut price during the rout, credit investors treated the new notes like a velvet-rope event.

SpaceX is pitching itself as the next great AI infrastructure player, a story that burns cash and demands cheap capital. Bondholders, sitting on a company with roughly $100 billion in cash and recurring Starlink revenue, decided the risk was worth a yield that prices wide of comparable investment-grade names.

Mind the Gap: The split says everything about this market's mood. Credit investors are betting on present cash and a famous founder; equity investors are still pricing a future that's mostly slideware. A $25 billion bond that nearly quadrupled its order book, struck in the same week the stock cratered, is the cleanest snapshot yet of how differently the two camps see the same rocket.

This week buried a Fed legend, watched a circuit breaker trip in Seoul, and ended with Mark Zuckerberg deciding the one thing missing from your feed was a betting window. The AI trade lost altitude, SpaceX borrowed $25 billion it could practically find in the couch cushions, and Google kept leaking brains. Micron's number lands after today's close and the Fed's favorite inflation gauge hits Friday. Somewhere, a Maestro is reminding us that exuberance has a way of turning irrational. Place your bets accordingly.

That's the tape. We'll see you at the open. — AllThingsWallSt