Happy Halloween, y'all. The scariest costume in tech this week was a quarterly earnings report. Intel got gutted, Nvidia is thinking about cosigning somebody else's construction loan, an OpenAI-backed holding company bought your neighbor's alarm company, and the Justice Department is still fighting to keep federal agencies off Claude. Let's go.
01 of 04
Intel just had the kind of day you don't forget
Intel stock fell roughly 28 to 29 percent on Friday after it warned that fourth quarter revenue would land between $15 billion and $16 billion. Analysts were looking for about $17 billion. It also expects gross margin to slide to 37 percent from 43 percent and said it does not expect positive cash flow because it is still spending heavily. CNBC called it Intel's worst day since the 1970s.
Here is the AI twist. Intel's data center revenue actually grew 78 percent year over year. The problem is PCs. AI companies are hoovering up memory and other components, prices for everything else go up, and people stop buying laptops. CEO Lip-Bu Tan said it plainly on the call: "Our customers began signaling a PC slowdown based on the memory prices and availability."
And you and I are along for the ride. The U.S. government took a roughly 10 percent stake in Intel back in August 2025, and Friday's drop knocked about $12 billion off the value of that stake, per CNBC. Nvidia, which also bought into Intel in 2025, dipped too.
What gets me is the irony. AI demand is real enough to grow Intel's data center business by 78 percent, and it is still eating the company alive from the other side. That is what a boom looks like when it starts pulling resources away from everything that is not the boom. If you own a "safe" chip stock because AI is hot, this is your reminder that the supply chain cuts both ways.
Sources: TechCrunch, CNBC
02 of 04
Nvidia wants to cosign Nscale's loan in West Virginia
The Information reported Thursday, and Bloomberg corroborated, that Nvidia is in talks to guarantee billions of dollars of debt for Nscale's planned data center campus in West Virginia, which The Next Web identifies as the Monarch Compute Campus. Nobody has reported an exact dollar figure for the guarantee, and the deal is not signed.
The structure is a residual value guarantee on the first 460 megawatt building. If Nscale defaults on construction payments, Nvidia steps in, temporarily takes over the facility, and helps find a new buyer and tenant. That is different from the backstops Nvidia has given other neoclouds, which mostly covered leases and chip purchases, not bricks and concrete.
The tenant is the whole reason this matters. Anthropic signed a $44.6 billion contract for that first building, according to Nscale's IPO filing, and the two are talking about a second deal of roughly $45 billion. Per the Financial Times, Anthropic can walk if the building's financing misses a deadline, and Nscale has not yet lined up the money to build or equip it. Nscale also posted a $1 billion net loss on $141 million of revenue in the first half of 2026, per the FT, and is trying to IPO at a valuation of up to $35 billion.
Let me put it in plain Texas terms. When the guy selling you the truck offers to cosign your loan, he is not doing it out of love. He is doing it because he needs the truck to sell. Nvidia propping up a customer so that customer can keep buying Nvidia chips is a loop, and loops are fine right up until somebody in the middle can't pay. I'm not saying it breaks. I'm saying I'd want to see the guarantee number before I call this healthy.
Sources: TechCrunch, The Next Web
03 of 04
An OpenAI-backed company is buying ADT for $15.3 billion
Thrive Holdings, the holding company spun out of Josh Kushner's Thrive Capital, agreed Thursday to buy ADT for $10 a share in cash. That is a 45.6 percent premium over ADT's $6.87 close on October 27, about $7.6 billion in equity value and $15.3 billion in enterprise value once you count the debt. Apollo funds are providing all of the financing.
Despite some headlines saying it already happened, it has not closed. ADT shareholders and U.S. antitrust regulators still have to sign off, there is a 30 day go-shop running through November 28 where a better offer could show up, and the target close is the first quarter of 2027. After that, ADT goes private and leaves the New York Stock Exchange.
Where does AI come in? Thrive Holdings owns more than 75 businesses in accounting, IT services, and the built environment, and TechCrunch reports that OpenAI is both a partner and an investor, sending models and forward deployed engineers into those companies. The press release says Thrive's engineers will work alongside ADT's team to find spots where tech can improve its products. It does not name a single new product. Thrive partner Anuj Mehndiratta added that Thrive is "committed to upholding the highest standards of privacy and cybersecurity," and the release says ADT's data policies stay the same and no new data sharing happens without customer permission.
What I want to know is what "AI for home security" actually means once the ink dries. ADT's real asset is not cameras. It is the trust of people who let a company watch their front door. Pairing that with an AI lab's engineers is either a smart upgrade or a privacy story waiting to happen, and a promise in a press release is not the same as a promise in a contract. I'll believe the data policy when it survives the first growth target.
04 of 04
The DOJ is still fighting to keep agencies off Claude
On Monday, Bloomberg reported that the Justice Department appealed a ruling from U.S. District Judge Rita Lin in the Northern District of California that lifted the federal ban on agencies using Anthropic's AI. AI Chat Daily reports the appeal went to the Ninth Circuit and targets Lin's 59 page order from August 27.
Quick backstory. On February 27, the White House told federal agencies to stop working with Anthropic on a six month phaseout, and Defense Secretary Pete Hegseth designated the company a supply chain risk. The fight started when the Pentagon wanted Anthropic to agree its models could be used for "all lawful use cases," and Anthropic wanted contract language ruling out autonomous weapons and mass domestic surveillance. Anthropic sued on March 9, and Lin temporarily blocked the government in March. The Justice Department appealed that order in April too.
According to AI Chat Daily, the August order found the government engaged in First Amendment retaliation, denied Anthropic notice it was owed under the Fifth Amendment, and acted in an arbitrary and capricious way under the Administrative Procedure Act. Lin wrote that "the empty invocation of national security is not a blank check to punish and retaliate against government critics." She did side with the government on the separation of powers claims.
Here is where I land. You can think Anthropic is being precious about its rules, and plenty of folks do. But the government's move here was to blacklist a vendor for saying no to a contract term. If that is allowed, every AI company learns the lesson fast: never put a limit in writing. That is a worse outcome for everybody, including the Pentagon. One more thing to watch: this is the case about the agency ban. The supply chain designation itself is being fought separately in the D.C. Circuit, so a win in one court does not end this.
Sources: Bloomberg, AI Chat Daily, PYMNTS