Everything Is Fine and Nothing Is Real
Everything Is Fine and Nothing Is Real
Friday the jobs number dropped and the market had a small nervous breakdown. 172,000 nonfarm payrolls against an 80,000 consensus. The 2-year yield jumped 10 basis points. The 30-year cracked 5% again. The S&P 500 fell 2.6%, ending a nine-week win streak in the ugliest single session of the year. Chips led the selloff — the sector lost roughly 10% in a day. Everyone who had been ignoring the inflation data, ignoring the Warsh committee deadlock, ignoring the fact that CME FedWatch had quietly been repricing hike probabilities for weeks — they all looked up at once, saw what they'd been ignoring, and sold.
Then Monday came and they bought it all back.
Not all of it. But most of it. The Nasdaq 100 gained 1.5% at the open. NVDA caught a bid. The narrative reassembled itself almost immediately: it's a dip, the AI buildout is intact, the ceasefire in the Middle East is holding, and anyway look at the developers conference happening in Cupertino — Tim Cook is crying and Apple is telling everyone the future is going to be great. The healing was swift. The amnesia was complete.
This is the market in 2026: a psychological state wearing an Excel spreadsheet as a costume.
Let's talk about what actually happened at WWDC yesterday, because it deserves some honest examination.
Apple walked onto the stage at Apple Park and announced that Siri — the assistant that couldn't set two timers simultaneously as recently as 2023, the product that became a cultural punchline while OpenAI printed billions from ChatGPT — is now called Siri AI. The rebranding is so naked it should embarrass someone. The product upgrade is more substantive: Apple partnered with Google to build next-generation foundation models, burned down its old architecture, and rebuilt Siri to be conversational, context-aware, capable of multi-turn exchanges, and embeddable across the hardware ecosystem. Craig Federighi described it as "a deep collaboration" with Google on Gemini. The new Siri will also support third-party models — Gemini, potentially Claude, potentially others. Choice, interoperability, platform openness. The language of a company that knows it's behind.
Apple's stock rose 2% at the open and then slid during the keynote. Which tells you everything. The analysts wanted a come-from-behind victory speech. What they got was a competent concession that reframed the contest. Apple spent two years missing deadlines on features it announced at the 2024 WWDC. The new AI chief, Amar Subramanya, took over from John Giannandrea only months ago. John Ternus takes the CEO chair in September. Tim Cook cried. The last keynote of a fifteen-year run that turned AAPL into the world's second-most-valuable company — currently sitting at $4.5 trillion in market cap, a number requiring repeated mental verification — and the product centerpiece was a voice assistant with a new name and Google's models running underneath.
None of this is a disaster. It might actually work. The bet Apple is making — don't build the frontier model, instead integrate the best ones and own the relationship with 2 billion devices — is a reasonable strategy for a hardware-first company that has never been in the research-lab business. But it is a bet. And the market had been pricing AAPL as an AI winner for eighteen months based on a thesis that was never really articulated out loud.
The same day, OpenAI quietly filed a confidential S-1.
Confidential filings mean no numbers yet. But the structure is legible: OpenAI converting from nonprofit to for-profit, the valuation conversation happening around $300 billion, Sam Altman sitting on a company that has been burning cash at rates that would alarm anyone who looked closely. The IPO filing comes one week after Anthropic filed confidentially. These are the two companies that have convinced the world — and Silicon Valley's institutional capital — that the frontier model race is the most important industrial competition since the semiconductor itself.
Whether that's true is not the question. The question is what OpenAI is worth as a public company in a rate environment where the 30-year yield just touched 5.02%, where the Fed is more likely to hike than cut before year-end, and where the hyperscalers who are both OpenAI's customers and its existential competitors — Microsoft, Alphabet, Amazon, Meta — collectively spent something close to $610 billion on AI infrastructure last year alone. OpenAI's revenue is real. Its moat against the companies funding its compute is not.
But none of that matters yet. The S-1 is confidential. The roadshow is months away. The market will price the narrative before it prices the numbers, because it always does.
Here's the thing about Monday's rebound that should nag at you.
The jobs data that caused Friday's rout did not change over the weekend. 172,000 payrolls — more than double consensus — is still on the board. Average hourly earnings were up 3.4% year-on-year. The unemployment rate held at 4.3%. Nothing about this picture is consistent with rate cuts. Nothing about it is consistent with the forward P/E on the S&P 500 sitting at 23x. The 30-year at 5% is the bond market's editorial on all of this, written in yields.
The chip sector's 10% drop was a momentary honesty. The recovery was the market arguing that the AI demand cycle transcends macro, that Nvidia's data center revenue — $81.6 billion last quarter, up sharply year-on-year, with a backlog that remains intact — insulates the trade from rate sensitivity. Which is an argument worth taking seriously. These are real revenues from real customers writing real checks. The AI infrastructure cycle is not a fiction.
But the speed of the recovery is its own tell. Markets that have genuine conviction don't need to convince themselves back into positions over a weekend. They hold during the selloff because they believe the thesis. What happened Monday felt more like someone checking that the story still hung together and finding it provisionally held, so buying back in quickly before anyone noticed they'd sold.
The CPI data drops Wednesday. If it comes in hot — and the May jobs number suggests an economy that is not cooling — then Friday will look like a preview, not an overreaction. The 30-year at 5% is a pressure point. The Warsh committee meets June 17 with a market that has just re-priced hike odds to 72% on CME FedWatch. Apple slaps AI on Siri and OpenAI starts drafting its roadshow materials.
Everything is happening very fast and none of it is quite touching the ground.