👋 Good morning! Stocks fell on Thursday as the optimism for an Iran resolution — perhaps foolish in hindsight, since Iran was denying negotiations — evaporated, sending them down into the red.
The S&P 500 (^GSPC) lost 1.7%, the Dow (^DJI) 1.1%, and the Nasdaq (^IXIC) 2.4% as Brent crude held at over $102.
After the close, however, President Trump said on social media he would extend the pause on strikes on Iranian oil infrastructure for another 10 days, characterizing talks as “going very well.” The new deadline? 8 p.m. ET on April 6.
On the agenda this morning:
👎 Social media sting
🏞️ ‘Buy land, AJ’
🍪 Random Access Memories
🥶 Microsoft’s hiring freeze may be a playbook
✈️ One picture says 1,000 words
📆 What we’re watching Friday: With President Trump urging Iran to “get serious” and make a deal before it’s “too late,” the tenor of the vague negotiations has gone from productive to fraught.
In addition to that big potential catalyst, we’ll be watching for sentiment data out of the University of Michigan, which will be our week’s big economic reading.
Meta CEO Mark Zuckerberg departs a meeting with Senate Majority Leader John Thune (R-SD) on Capitol Hill on March 26, 2026 in Washington, D.C. (Andrew Harnik/Getty Images) ·Andrew Harnik via Getty Images
The penalties that juries in New Mexico and Los Angeles ordered Meta — and, in the latter case, Google as well — to pay are nominally small. The penalties that investors charged these companies and their social media peers for these rulings were far larger.
Shares of Meta, Snap, and Reddit were big losers on Thursday as investors saw these rulings as more likely to be the beginning rather than the end of a new era of legal headaches around what responsibility these businesses have for the content on their platforms.
But this slide also brings to mind an old market adage, which is that price leads narrative.
Before Thursday’s losses, Snap and Reddit had seen their shares fall more than 40% this year. Meta stock — one of the biggest winners of the AI trade — was down over 12% for the year.
Negative legal headlines are rarely spun as a positive for a company, but the problems these stocks were already facing in what’s become a downright dour stock market weren’t zero before Wednesday. And they are greater than that now.
(L-R) Actors Tony Sirico, Steven Van Zandt, James Gandolfini, Michael Imperioli & Vincent Pastore in publicity still for HBO cable TV series The Sopranos. (by Anthony Neste/The LIFE Images Collection/Getty Images) ·Anthony Neste via Getty Images
Using cryptocurrency to turn digital wealth into a physical plot of land isn’t new. But using cryptocurrency to acquire a mortgage that conforms to Fannie Mae’s standards is. And now you can do that too.
Lender Better Home and Coinbase announced the new product on Thursday, which will allow people to pledge bitcoin or USDC to fund a cash-down payment for a conforming loan, which typically comes with a lower rate than other ways people might’ve previously used crypto to buy real estate.
Tony Soprano famously said, “Buy land, AJ, ’cause God ain’t making any more of it.”
The same is technically true of bitcoin; once the 21 millionth bitcoin is mined in 2140, that’ll be it. Another piece of investing advice for the boss to pass along. Perhaps one better aimed at Christopher.
Screenshot from Google’s TurboQuant website. We will not pretend to understand what this means, technically speaking, but it looks cool.
Memory stocks sank on Thursday, the latest development in what’s been an interesting week for the AI trade.
Bloomberg attributed a sell-off in memory names like SK Hynix, Micron (MU), SanDisk (SNDK), and Western Digital (WDC) to research published by Google earlier this week that appeared to lower demand requirements for AI models.
Google researchers unveiled a tool called TurboQuant, “a compression algorithm that optimally addresses the challenge of memory overhead in vector quantization.”
When you interact with an LLM, part of that response — in some cases, most or all of it — is coming from a memory cache that houses previous interactions with the model. In other words, every interaction with an LLM does not start from zero.
TurboQuant is aimed at lowering the memory intensity of storing these responses. A shortage of memory chips has been one of the key bottlenecks in AI development flagged by the industry over the last few months.
Thursday’s decline in memory stocks also cools off one of the hottest AI trades this year and is the latest pocket of stock market weakness tagged to new technological advances in AI.
Software stocks have also continued their sell-off this week, with this decline attributed both to new agentic tools released by Amazon and advances from Anthropic that will allow Claude to complete a wider range of tasks on a user’s computer.
AI has given a lot to investors over the last few years. What it’s starting to take away – or threatening to, at least – is adding up.
Satya Nadella, CEO of Microsoft, speaks on stage during the Microsoft AI Tour. (Sven Hoppe/picture alliance via Getty Images) ·picture alliance via Getty Images
We’ve noted above that unemployment claims have not surged despite some high-profile layoffs this year from Amazon, UPS, and Meta.
Though the headlines paint a trend, the overall data on a national scale does not show the needle moving significantly.
But we’d remind the labor market curious to remember the mantra of the current economic situation: “low hire, low fire.”
It’s a slow burn of a situation that comes from suppressed immigration, employers who remember how hard it was to hire in the post-pandemic moment, and hopes of doing more with less as AI potentially supplants the need for higher headcount.
Microsoft’s reported hiring freeze in various major divisions, including sales and cloud, is perhaps indicative of where the labor market could go amid an AI revolution.
Instead of an AI transformation precipitating a blizzard of pink slips, mere attrition followed by a hiring freeze might do the dirty work with far less PR drama and expense. It’s a playbook we’re watching.
At the very least, it’s a sign that even for a big, rich hyperscaler like Microsoft, the AI build-out is coming with a cash crunch.
“Not only did we invent the car, we have invented so many new technologies over the decades. We have one motto, if you will. It’s not technology for the sake of technology, it’s technology for the sake of humans.”
Long TSA lines OUTSIDE at Atlanta Hartsfield-Jackson International Airport on Thursday. (Megan Varner/Getty Images) ·Megan Varner via Getty Images
Given what it’s a picture of, we can only imagine those words are of the four-letter variety.
Economic data: University of Michigan sentiment, March final reading (55.5 previously); U. Mich. current conditions, March final reading (57.8 previously); U. Mich. expectations, March final reading (541. previously); U. Mich. 1-year inflation, March final reading (+3.4% expected previously); U. Mich. 5-10 year inflation, March final reading (+3.2% expected previously); Kansas City Fed services activity, March (6 previously)
Earnings calendar: Carnival Corporation (CCL), Legence Corp. (LGN), Perpetua Resources Corp. (PPTA), TMC the metals company (TMC), Standard Lithium (SLI), Nano Labs (NA)
Hamza Shaban is a reporter for Yahoo Finance covering markets and the economy. Follow Hamza on X @hshaban.
