THE MACHINE IS EATING ITSELF, AND WALL STREET IS CALLING IT INNOVATION
THE MACHINE IS EATING ITSELF, AND WALL STREET IS CALLING IT INNOVATION
February 26, 2026
Let me walk you through what happened this week, because if you read the headlines in isolation you'd think the markets were being run by a chaos engine fed on espresso and existential dread.
Monday morning. IBM down 13.2%. Worst single-day drop since October 2000. The dot-com bubble era. The company that has spent three decades selling "AI transformation" to Fortune 500 executives, slapped with a 25-year low by a blog post. Not a leak. Not a scandal. A blog post — from Anthropic, explaining that Claude Code can now modernize COBOL systems. Automate the exploration. Compress years of consulting into quarters.
COBOL. The programming language from 1959 that still quietly runs 95% of ATM transactions in the United States. Hundreds of billions of lines. The scaffolding beneath the world's financial plumbing. IBM has made a religion out of being the only high priest of that temple. Their entire mainframe consulting empire — the armies of contractors, the multi-year migration projects, the seven-figure transformation engagements — all of it rests on the assumption that COBOL is too sacred, too arcane, too dangerous to modernize without IBM as an intermediary.
One blog post. Thirteen percent. Gone.
And here's the thing that should make you uncomfortable: IBM's stock was technically already pricing in some AI disruption. The stock had lost 25% year-to-date before Monday even happened. This wasn't a cold-water shock to a complacent market. This was a second punch landing on someone already bleeding. RSI at 26.98 as of today. Money Flow Index at 18.34. UBS rushed out an upgrade — from Sell to Neutral — which tells you everything about the confidence level in the room.
The week before that, it was cybersecurity. Claude Code Security launches — a tool that can scan codebases for vulnerabilities and suggest fixes. CrowdStrike: down 10%. Zscaler: down 10%. Netskope: down 12%. SailPoint: down 9%. The iShares Cybersecurity ETF hit its lowest level since November 2023. George Kurtz spent the weekend posting on LinkedIn about moats. There's something almost poignant about a cybersecurity CEO defending his company's perimeter on a social network.
Before that, SaaS. Before that, logistics. Trucking stocks. Commercial real estate. Financial services. Anthropic releases a tool, a sector collapses. The Kobeissi Letter put it cleanly: every new Claude capability has been erasing over $100 billion in market cap from the targeted industry. Citrini Research drops a note warning that widespread AI adoption could push U.S. unemployment toward 10%. American Express fell 7% on Monday. Mastercard dropped 6%. The fear has jumped the fence from software and is now roaming freely across anything with a labor-intensive moat.
We are watching a sequential deindustrialization of knowledge work in real time, priced into equities one sector at a time, every two weeks, with the reliability of a newsletter drop.
Meanwhile, the other story swallowing the week — the tariff chaos — has its own particular flavor of absurdity.
The Supreme Court ruled 6-3 that IEEPA doesn't authorize the president to impose sweeping tariffs. Fine. The market rallied briefly. Exporters exhaled. Then, hours later — hours — a new executive order lands. 10% global tariff under Section 122 of the Trade Act of 1974. A provision so obscure that the Justice Department's own lawyers dismissed it as a viable option last year. By the weekend, it's 15%, announced via Truth Social, "effective immediately," with threats of more to follow.
Section 122 allows the president to impose tariffs for up to 150 days citing a balance-of-payments crisis. The problem: the United States is exhibiting exactly none of the conditions a balance-of-payments crisis requires. No currency collapse. No capital flight. No spiking bond yields from foreign disinvestment. RSM's chief economist noted, politely, that by no reasonable reading of the law does the current economic situation qualify. The DOJ's own prior briefings said as much.
It doesn't matter. The 15% tariff is live. The EU postponed ratifying its trade deal, waiting for clarity that isn't coming. Steve Madden withdrew its 2026 earnings guidance — the first major corporate surrender to tariff uncertainty since the Supreme Court ruling. The U.S. average effective tariff rate is sitting near 17%, the highest since the early 1930s, and the administration has signaled it will use the 150-day Section 122 window to build out longer-term tariff architecture under Sections 232 and 301. The clock is ticking. Congress — which would need to approve any extension — goes to the polls in November.
The Tax Foundation estimates U.S. households absorbed roughly $1,000 in tariff costs in 2025. The 2026 figure is running closer to $1,300. Consumer sentiment is weakening. Affordability is a problem. And the same administration that created the problem is now scrambling — tariff rollbacks on some steel and aluminum, the India deal dangling at 18% down from 50%, Xi meeting pencilled in for late March — to look like it's solving it before midterms make everything a referendum.
The Supreme Court handed the administration a wound, not a kill shot. That's the takeaway. The legal architecture for sweeping tariffs is narrower now. But it's still there, and the 150-day Section 122 clock gives the White House just enough runway to construct something more durable — or to use the tariff threat as a negotiating posture for bilateral deals. The ambiguity is the point.
So here's where we actually are.
Salesforce reported last night. Revenue: $11.20 billion, 12% year-over-year growth. Beat on earnings — $3.81 adjusted versus $3.04 expected. The stock dropped 5% in after-hours anyway, because the fiscal 2027 revenue guidance trailed consensus. The stock is down a third since January. The company is growing at double digits, integrating Informatica, poaching ServiceNow clients, and the market is treating it like a sunset industry. The fear premium now exceeds the fundamental premium in enterprise software.
Gold is at $5,240. Fifth straight session of gains after the historic crash-and-recovery earlier this month. Bitcoin slumped to $74,570 at its low — down sharply from the $126,000 October record — and has since partially recovered, though the recovery has the energy of someone who got hit by a bus and is insisting they're fine.
The 10-year Treasury is hovering around 4.03%. The BoJ raised rates to a 30-year high in December and is signaling more. The RBA hiked for the first time since 2023 in February and warned of further hikes if Q1 inflation disappoints. Two of the last major holdouts in global monetary loosening are now tightening at the same time the U.S. is navigating an election cycle that demands rate cuts. The Fed is frozen — not by data, but by the political physics of the moment.
Nvidia reports tonight. The options market is pricing a 4% single-day move in either direction. Short interest is up. The stock has gone essentially nowhere in a month.
Here's the structural question nobody is properly asking: what happens to corporate earnings — and by extension, equity valuations — if the AI disruption narrative is directionally correct even at half the speed Anthropic's pace implies?
Every two weeks, a new industry wakes up to the possibility that the last twenty years of software investment, consulting moats, and knowledge-worker pricing power might compress faster than any restructuring plan can accommodate. The AI buildout keeps getting validated — Nvidia's capex cycle is real, the hyperscalers are spending, the infrastructure is going in. But the revenue destruction on the consumption side of the equation is now being priced in, erratically, sector by sector, blog post by blog post.
Constructive destruction usually happens over decades. This is moving in quarters.
The market doesn't know what to do with that. Neither does anyone else. The honest position is: nobody does. The IBM trade, the CrowdStrike trade, the Salesforce trade — they might all be overdone. Structural disruption stories, as any technician will tell you, can stay depressed far longer than the fundamentals would suggest. Or the fundamentals catch down to the fear.
Both outcomes are live.
The only certainty is that the machine is currently eating its own supply chain, the tariff regime is being held together with legal duct tape and 150-day statutory clocks, and Nvidia is about to tell us whether the infrastructure boom that funds all of this is still intact.
Watch the guidance, not the beat. The beat is already priced in — the question is whether Jensen's $700 billion capex ecosystem is still accelerating, or whether the first signs of digestion are appearing at the edges. If they are, the whole stack re-rates.
Stay hedged. Stay awake.