AI Hits the Brakes

AI slowdown fears slammed semis while oil and Treasury yields piled on the pressure.

Author

Elie Zakhour
September 15, 2026

Market Snapshot

Get a Top Job Offer, Guaranteed (or tuition is free) | Apply Here

📉 Banana Bits

Market News

The AI Bosses Asked for a Speed Limit. The Chips Took It Personally.

Stocks opened ugly on Monday and spent the afternoon crawling back. S&P 500 down 0.48% to 7,619.98, Nasdaq down 0.56% to 26,186.41, Dow off 152 points to 52,421.20. At their session lows, the S&P was down 0.8%, the Nasdaq 1.3%, the Dow almost 300. So not a disaster. More like the market read a very long essay and had to go lie down for a bit.

The spark came from Anthropic CEO Dario Amodei. In a roughly 3,800-word essay, he argued that AI capabilities are advancing faster than companies can reliably understand or control them, and that the industry should deliberately slow the pace enough for safety work to catch up.

His plan has three steps: independent safety auditors sitting inside labs with employee-level access, then shared standards across labs in democratic countries, then eventually international agreements. Altman posted that he agreed, said pacing has become a real internal debate at OpenAI, and then ruled out the 2026 IPO one month after his CFO told staff it would be 2027 "or sooner." Even Musk agreed, which may be the weekend’s most impressive alignment result. Worth remembering all of this comes two months after OpenAI disclosed that one of its autonomous agents broke containment and hacked Hugging Face. So, context.

Wall Street did the math fast. If slower AI development eventually means less urgency around massive training runs, that could mean fewer GPUs, racks and data-center buildouts. So the entire picks-and-shovels stack got sold at once. Broadcom -4.8%, AMD -4.4%, Lam -8.3%, Super Micro -8.4%, CoreWeave -6.8%, Vertiv -7.7%. The SOX sank 5.9%, making it painfully clear where Monday’s damage was concentrated.

Nobody actually went to cash though. They just bought software. Alphabet +3.2%, Meta +2.7%, Microsoft +2.0%, Salesforce +4.7%, and the communication services ETF gained 2.2% on a red day. The market's logic was pretty simple: slower infrastructure spending could ease the capex burden on the hyperscalers writing the checks, while increasingly capable AI creates even more demand for cybersecurity.

Then oil, which any other day would've been the whole newsletter. By Tuesday morning, Brent topped $108 and WTI cleared $103 as the shutdown of Saudi Arabia's East-West pipeline kept squeezing supply. Iran-aligned groups in Yemen are squeezing the Red Sea route too. Diesel is at a record above $6.23 a gallon, and Gasoline alone accounted for more than a third of August's monthly CPI increase, so this shows up at the pump, not just in an inflation table. And the bond selloff wasn't done. The 10-year Treasury yield pushed above 5.02% Tuesday morning, its highest level since 2007, as surging oil kept inflation fears alive ahead of the Fed.

Peel Take: The AI trade didn't break, the assumption that the labs would never voluntarily tap the brakes did. And fine, Amodei's plan is the most specific safety proposal any lab CEO has put in writing. It also dropped the same week his company is shopping a $2 trillion IPO with Nvidia anchoring it. You can ask everyone to slow down while you're standing at the dealership, but people notice, and CNBC had the "tightrope" story up before lunch.

What's Ripe

Zscaler (ZS) 16.52% 

  • ZS was down about 27% on the year going into Monday, so this was rotation plus short covering.

  • There was no major company-specific catalyst. The move looked primarily like part of the broader cyber rally sparked by renewed concerns around AI safety and security.

  • The whole group moved together: Zscaler rose 16.5%, CrowdStrike nearly 14%, Palo Alto about 13%, and Fortinet 9%. Investors were buying the category, not just one company.

  • Peel Take: Zscaler may be the clearest sign this was a sector trade. It entered Monday as a major 2026 laggard, had no big company-specific announcement, and still finished as one of cyber’s biggest winners. That tells you investors were buying the security theme first and sorting out the winners later.

CrowdStrike (CRWD) 13.85% 

  • Cybersecurity became the market’s favorite hedge against the AI scare. CrowdStrike jumped 13.85%, while Palo Alto, Zscaler and Okta all rallied sharply as investors bet more powerful AI will require more security around it.

  • CEO George Kurtz clapped back at Amodei on X: the frontier moves no matter what any one company does, and cyber's job is to make it safe rather than slow. Cramer then called the stock a "MUST BUY," which historically could mean literally anything.

  • Timing helped. CrowdStrike had just wrapped Fal.Con, where it launched Falcon Guardian, SafeMind and new AI-agent identity tools. Raymond James raised its target to $250, Truist lifted its target to $300, and RBC reiterated Outperform at $260.

  • Peel Take: Probably the cleanest winner from Monday’s rotation. If AI agents start doing real work inside companies, securing what they can access, execute and change becomes a much bigger line item. Still, CrowdStrike has already doubled this year and just added another 14% in one session. The theme is strong. The valuation leaves very little room for the story to cool off.

What's Rotten

Corning (GLW) 13.70%

  • Worst stock in the S&P 500. Friday after the close, Corning unveiled a $2 billion at-the-market equity program with Goldman as sales agent. Setting up a potential $2 billion stock sale right before an AI infrastructure selloff is impressive timing. Bad, but impressive.

  • Corning can sell shares over time under the program and says proceeds may be used for general corporate purposes. Mizuho cut its price target to $180 from $210 as investors weighed the capital needed to keep expanding capacity.

  • The stock entered Monday up about 90% YTD, fueled by booming optical-fiber demand and last week’s multi-billion-dollar Verizon agreement.

  • Peel Take: ATMs are the sneaky kind of dilution. No fixed price, no single deal day, just the ability to keep issuing shares into the market over time. Add “your biggest customers might slow down” and yeah, -14% makes sense. Thing is, the fiber buildout already underway doesn’t disappear because the next frontier model ships later. Data centers going up now still need to get wired. On a one-year view this feels like an overreaction. On a one-week view, why stand in front of a fresh $2 billion equity overhang? Let it work, come back later.

Hewlett Packard Enterprise (HPE) 10.76%

  • Evercore's Amit Daryanani cut HPE to In Line from Outperform and left his $65 target alone. Pure valuation call: HPE at roughly 13x FY27 earnings versus a five-year average near 8x.

  • HPE closed Friday up 158% on the year (S&P: +12%) after a 12% single-day pop on Sept. 11. Monday gave back the entire Friday move.

  • AI-server peers went with it: Super Micro -8.4%, Dell -5.8%.

  • Peel Take: A downgrade to In Line with the target unchanged took 11% off the stock. That's not on Evercore. The stock was priced for perfection and the note was the excuse. "Nice run, you're fairly valued now" is about the kindest thing an analyst can say, and the market treated it like a guidance cut.

🧠 Technical Trip

Interview Q&A from UBS

👉 Want 1-on-1 recruiting help from UBS bankers & 5,000+ top mentors? Apply to WSO Academy

🦈 Deal Dispatch

M&A, IPOs, And Other Notable Transactions

Student Success Corner

Banana Brain Teaser

Previous

A company has $300 million of revenue and an EBITDA margin of 20%. If depreciation and amortization are $15 million, what is the company’s EBIT?

Answer: $45 million

Today

A company has $400 million of revenue and a 25% gross margin. What is its cost of goods sold?

❝

The most important quality for an investor is temperament, not intellect.

Warren Buffett

How Would You Rate Today's Peel?


Happy Investing,
Chris, Elie, Mitchell, Fernanda, Nick, & Patrick