The Great Unmixing Has Begun

We're witnessing something that doesn't happen very often: a market literally rewiring itself in real time, sorting assets and capital into fundamentally different buckets based on what they actually do rather than what narrative investors hope they'll become.

It started with gold. Last week, precious metals experienced what analysts politely call a "reversion," which is Wall Street speak for "someone finally asked the question everyone was avoiding." The question was simple: If a newly nominated Fed chair has signaled he believes in sound money and fiscal discipline, why would you hold an asset whose entire bull case rested on currency debasement?

Gold plunged 11% on Friday—a historic sell-off—and by Tuesday was trading in the $4,900 range after rebounding 6%. Silver fell 31%. These weren't corrections. These were capitulations. The kind of moves that happen when a consensus trade gets abruptly exposed as a consensus hallucination.

What followed was predictable but instructive. The same momentum-chasing capital that had crowded into precious metals pivoted to the next available narrative. In this case, it was the wholesale dumping of software stocks. Tech stocks faced sustained pressure this week, with the Nasdaq down as traders rotated out of technology giants and into broader market categories. ServiceNow and Salesforce both fell close to 7% each as software stocks continued their 2026 tumble, with the iShares Software ETF down 20% for the year.

The stated reason: the release of an AI automation tool from Anthropic that services legal work exacerbated fears about risks to the core businesses of the software industry. In other words, someone built something that does what software companies sell, and investors collectively realized for the first time that maybe—just maybe—this is a problem.

This pattern has a name in financial history. It's what Kindleberger called the "revulsion" phase of a speculative cycle. After the mania, the crash. But before the stabilization, there's a period of pure panic liquidation where everything gets sold because nothing feels safe. The 1987 crash, the Asian contagion of 1997, the 2008 meltdown—they all followed this script.

What's unusual about the current moment is how geographically and sectorally diversified the panic is.

South Korea's Kospi index sank 5.26% on Monday, its worst day since April, after the index soared 76% in 2025 on investor enthusiasm about AI. The Kospi isn't some backwater emerging market index. It represents Samsung, SK Hynix, NAVER—real companies with real earnings that had gotten swept up in a global mania for anything remotely related to artificial intelligence. When the mood shifted, they got hit. Hard.

Bitcoin tumbled from above $83,000 to as low as $74,570, hitting its lowest level since April. The crypto faithful will tell you this is healthy. A shakeout. The weak hands getting flushed out. They're not wrong, but that framing obscures what's actually happening: risk assets are being repriced lower across the board because the underlying assumptions supporting their valuations have come into question.

Here's what makes this moment different from your typical market correction: the dispersion is revealing something true about market structure today.

We have, loosely, three groups of stocks. First, the genuine AI beneficiaries whose business models actually improve because of the technology—the Palanit's, Super Micro Computer, companies whose earnings will genuinely be higher in 2026 and beyond. Eli Lilly jumped more than 8% after earnings per share easily surpassed Wall Street expectations, with the company guiding for fiscal 2026 EPS and revenue above analyst consensus. Super Micro Computer shares jumped 10% after strong demand for its AI-optimized servers helped fiscal second-quarter results top expectations, with the company raising its annual revenue forecast to at least $40 billion. These stocks are holding up because their story is coherent and defensible.

Second, the software companies whose business models face genuine disruption. Novo Nordisk slid 5% early Wednesday as the company guided for 2026 sales and profit to fall between 5% and 13%, worse than expected. This is what a structural headwind looks like.

Third—and this is the important one—the carry trades, the momentum plays, the assets that were held because everyone else held them. Gold, silver, certain international equity indices, crowded cryptocurrency positions. These aren't being repriced based on new information. They're being repriced because the psychology that supported them evaporated.

What's fascinating is watching which assets are falling fastest. Shares of stocks with significant private credit market holdings were diving on fears about exposure to the industries being disrupted by artificial intelligence, with Blue Owl, TPG, Ares Management and KKR all down by double digit percentages. The private credit guys got squeezed because they've loaded up on software companies expecting steady returns in a stable world. But the world stopped being stable the moment people realized that Anthropic's Claude Code can literally replace some of what those software companies do.

This matters because it suggests the repricing is spreading from one asset class to another through financial plumbing rather than through economic fundamentals. That's how contagion works. That's how corrections become crashes.

And yet—and this is crucial—the broad market didn't break. The S&P is down 0.7% trying to hold onto its 50-day moving average. The Dow actually advanced. The real economy data remains solid. Manufacturing activity expanded the most since 2022, with mid-sized firms adding 41,000 jobs while large employers shed 18,000. This is exactly what you'd expect in a transition period: goods production improving, labor shifting, some companies thriving while others adjust.

The earnings calendar will be critical here. Wall Street this week will get a slate of corporate earnings results including from tech stalwarts Alphabet and Amazon that could influence market moves. If Amazon and Alphabet show that their immense capex is actually generating incremental revenue and margin expansion, the software rotation pauses. If they show that they're spending with abandon while returns languish, the repricing deepens.

Historically, periods like this precede either violent rallies or actual bear markets. The difference is usually determined by whether earnings growth can keep pace with valuation compression. Right now, consensus estimates suggest 11% blended earnings growth for the S&P 500. That's not nothing. It's not world-beating either, but it's respectable.

The real lesson from the past week isn't that markets are broken or that the sky is falling. It's that consensus can unwind very quickly when the narrative supporting it gets questioned. Gold was a consensus trade built on the assumption that central banks would remain irresponsible. When a Fed chair nominee emerged who signaled otherwise, the trade collapsed. Software was a consensus trade built on the assumption that AI would disrupt everyone except software companies. When evidence surfaced that it actually might disrupt software companies, the crowd scattered.

This is how markets work when they're functioning properly. They're not supposed to be stable. They're supposed to be mechanisms for discovering truth, and truth discovery is often violent and painful for the people betting on falsehoods.

What happens next depends on whether corporate earnings can validate the current level of the S&P 500—which is approaching all-time highs—without needing the help of precious metals appreciation, software multiple expansion, or crypto narrative momentum. That's a narrower foundation than we had two weeks ago. Whether it's stable enough is the only question that matters now.