😬 Palantir +50%. He's Short

Wall Street just bet billions on a small collection of stocks.

And after a volatile first half of 2026, it looks like they’re about to shift even more.

MarketBeat’s updated 10 Best Stocks to Own in 2026 report reveals the 10 names attracting fresh capital right now.

Send My Free Report

Michael Burry bet against Palantir. It went up more than 50% in a month.

That's the trade that defines his August. The man who called the housing crash spent the month watching almost everything he shorted go up.

Seven of his nine short positions rose. One was flat. Caterpillar was the lone stock that fell, and only by about 2%.

Burry's thesis is that Palantir is wildly overvalued and the poster child for an AI hype cycle that ends badly.

He didn't just hold that view. He doubled down with out-of-the-money puts running as far out as March 2027, strikes in the low-to-mid $100 range.

Then the stock went vertical. A 50% monthly move is not a wobble against your position, it's the position getting steamrolled.

On Nvidia, he bought December 2026 and June 2027 puts with strikes in the low $100s, plus December calls in the mid-to-high $200s as an earnings hedge. Nvidia rose 10% in August anyway.

Micron is the bolder one. He shorted it on 1 July 2026 at around $1,051.87, calling the memory rally speculation rather than fundamentals. He added to it in July. Added again in August.

Micron then gained 16.49%. Adding to a losing short twice takes a specific kind of conviction. Or a specific kind of stubbornness.

The rest of the short book behaved the same way:

Of 13 bullish positions, seven advanced, five fell, one was unchanged.

Veeva was the star, up 40%. Adobe added 17%, Fannie Mae 5.7%, MercadoLibre 3%, and Zoetis sat still.

The drags: PayPal fell 8%, Sprouts Farmers Market lost 6%, Flutter dropped 4%, and Fiserv slipped 1%.

His two loudest value calls barely moved. He called Lululemon "screaming cheap" and got about 1% for the trouble. He added to Molina Healthcare around $198, betting on the health insurer long term with the election season as a possible catalyst, and got 1.6%.

Buried under the pain is the position change worth watching. Burry exited Alibaba completely a few months back and significantly increased his bullish bet on JD.com instead.

His price for coming back? Alibaba has to fall roughly 50% from here.

One month of red on a short book proves nothing about whether the thesis is right. Burry was famously early on housing too, and early looks identical to wrong until it doesn't.

The useful signal is what he's targeting: AI infrastructure names with vertical charts, hedged with options dated into 2027. He's positioned for a timeline measured in years, not weeks.

The lesson for a normal portfolio is about timing risk. Being right eventually and being solvent throughout are two different problems, and options with expiry dates are how that problem bites.

In a gold rush, sell shovels. Every SpaceX launch runs on parts from public companies trading far below the IPO hype. See the 3 picks our analyst named. Get the Free Report.

Pizza Hut just sold for $1.5 billion.

The entire brand. Every red roof, every Stuffed Crust, every ball pit you fell into in 1996. Priced like a mid-size office block.

LongRange Capital has completed its acquisition of Pizza Hut from Yum! Brands $YUM ( ▼ 0.84% ). Mainland China stays with Yum.

So how does the biggest pizza chain in the world end up on the discount rack?

In 1958, brothers Dan and Frank Carney borrowed $600 from their mum and opened one store in Wichita, Kansas.

It worked. Aggressively.

By 1971, Pizza Hut was the world's largest pizza chain by both number of outlets and total sales. It listed on the NYSE in 1972 and went international later that decade.

Then the corporate parents started swapping it around like a fantasy football pick.

Pizza Hut's golden run was the late 1980s. By 1990, systemwide sales hit $4B, powered by two of the most successful menu items in fast food history: Personal Pan Pizza and Stuffed Crust.

Sit with that number for a second. The brand just sold for less than half of what it rang through the tills in a single year, three decades ago.

What happened in between is the boring part, and the important part. For the last decade Pizza Hut has been quietly bleeding market share to Domino's $DPZ ( ▼ 0.49% ) and Papa John's $PZZA ( ▼ 3.28% ), who built their businesses around delivery apps and tight ops while Pizza Hut was still defending dine-in restaurants nobody wanted to sit in.

Interim CEO Eduardo Luz is leaning into the fresh start, saying the chain is now fully focused on guests, franchisees and teams, and that LongRange's support and resources put it in a position to invest in the brand.

LongRange isn't a restaurant specialist. Its other holdings include 24 Hour Fitness, Batesville, US Synthetic and the Alpin Unlimited platform. This is a turnaround shop buying a tired brand cheap, not a pizza company buying a pizza company.

For Yum, the read is simpler. It keeps KFC, Taco Bell and the Mainland China Pizza Hut business, and drops the segment that's been dragging the numbers around.

For Domino's and Papa John's, a newly funded competitor with no public shareholders to please is a different animal to a neglected division inside a conglomerate. Private owners can cut prices, close sites and rebuild without explaining themselves every quarter.

Whether $1.5B turns out to be a bargain or a very expensive nostalgia purchase depends entirely on whether anyone still wants to order from the red roof.

SpaceX ran AI data centres for months with no backup cooling.

No backup power either. Just vibes and hope that the Tennessee weather stayed polite.

That detail comes from The Information, and it explains why the company has quietly torn up its data centre org chart.

The reliability problems hit sites in Tennessee and Mississippi, where SpaceX was sprinting to add computing capacity faster than it could safely wire it.

So management brought in people who normally build rockets.

Wesley Salandro, a senior production leader on the Falcon and Dragon programs, has been reassigned to infrastructure. So has Logan McConnell, a product-support director from Starbase.

They've also recruited a Neuralink engineer and posted several civil-engineering jobs.

When your fix for a cooling problem is "get the orbital launch people on it", the problem is not small.

Meanwhile, new data centres outside the Memphis hub are on ice, including proposed projects in Texas.

The shake-up follows a clear-out at the top.

Infrastructure chief Jake Palmer walked in late July, taking data centre leaders Zach Wells and Pablo Mendoza with him. Other managers spent the past year leaving for OpenAI and Anthropic.

Anthropic being, awkwardly, also a paying customer.

Because the AI business is the whole growth story now.

AI revenue hit $2.6 billion in Q2 2026, up from $737 million a year earlier. That's roughly 3.5x in twelve months, driven by deals renting computing capacity to Anthropic and Google.

The capacity numbers are the promise investors bought: about 1.4 gigawatts online in June, more than 2 gigawatts targeted by year-end, and further expansion penciled in through 2027.

For scale, 2 gigawatts is about a Hoover Dam's worth of power, pointed entirely at making chatbots think.

There's a safety valve too. SpaceX can reclaim the capacity it committed to Anthropic and Google if it needs the compute for itself. Which it might, having just spent $60 billion in all-stock money on coding startup Cursor in August.

This capability was front and centre in the pitch before June's public offering.

Gigawatt targets are forecasts. Backup cooling is a fact, and for months at some sites there wasn't any.

The bull case survives if the rocket engineers do what rocket engineers do and the 2 gigawatt target lands roughly on time. Falcon reliability wasn't an accident, and that operational muscle is genuinely transferable.

The bear case is simpler: outages and delays cost money twice. Once in repair bills, once in revenue that never shows up because a customer's workload wouldn't run.

What would change the view? The year-end capacity figure. Hit 2 gigawatts and this was a management wobble. Miss it, and the growth curve everyone underwrote gets a lot flatter.