Spyglass Dispatch: AI's Independence Race • EU's Bad Meta Ads • AI Chip Shenanigans • Netflix Ads Religion

The Spyglass Dispatch is a free newsletter sent out daily on weekdays. Feel free to forward it on to others who can sign up and view previous dispatches here. You can also view this dispatch on the web here.

I feel as if a significant percentage of stories pertaining to AI right now are about efforts by the various players to break from their dependencies on other players. This has long been the Apple playbook, of course. But it is seemingly happening across the board, and fast. Some companies want to break dependencies on models. Others on search agreements. And everyone on NVIDIA reliance. Though they'll never come out and say that. They don't dare. But actions speak louder...

The Race for AI Independence

Amazon, Apple, Google, Meta, and Microsoft all wish to break away from each other. As does OpenAI. With everyone needing to break free from NVIDIA.

Spyglass •M.G. Siegler

I Think...

🇪🇺 In Europe, Instagram Ads Are About to Get Less Personal – Look, I'm no fan of ads. And I get the data privacy elements here. But the reality of this situation is that it's going to make Meta products worse to use in the EU and likely harm Meta's business in the process. That feels awfully lose/lose. I much preferred the notion of paying to opt out of ads, and I wish you could do so in the US and UK, but that was not good enough for the EC, apparently. Meta simply must show you less targeted ads, they decree. And so they get what they want: "Some of the new ads—which will also be targeted based on age, gender and location—will cover the whole screen and be unskippable for a few seconds." Sounds great. Perhaps Mark Zuckberg's sucking up to Donald Trump and company was enough to persuade him to step in here as well? (In general, I'd bet the new administration issues a blanket "leave our companies alone" at some point next year. And because it's Trump, we'll also get an "or else...") [WSJ 🔒]

🍪 Inside the Murky New AI chip Economy – I pointed this out a week ago, but The FT Editorial Board goes more directly after some potential funny business – that's not particularly funny, but does seem to be a great business, for now... – going on up in the "neoclouds" where NVIDIA chips are being used to secure loans to buy more NVIDIA chips. The question is if/when the market corrects, what's the underlying value of those chips? This isn't like the railroads. Nor even guano. In some ways, the real value seems to be in the NVIDIA relationship to keep the latest and greatest chips coming. How do you value that? A drug dealer analogy is a bridge too far, but this whole market is wild right now. [FT 🔒]

📺 Netflix Says Ad Tier Has Surged to 70M Users, Christmas Day NFL Games Sold Out – Sort of wild how fast the narrative has turned here. Not that ads are starting to work for Netflix, but that Netflix would like everyone to know that ads are starting to work for them. And if they believe that – no surprise here – you have to believe the live events are going to ramp quite quickly. That "one-off" NFL game might be a half dozen "events" next year. More roasts. More award shows. More sports. Anything they can advertise against that all of their user base will see regardless of what tier they're on. Speaking of drugs, what starts with dabbling quickly turns into massive ad loads, I fear. Why make a buck when you can show double the ads and make two? Hopefully Netflix doesn't forget its roots. Remember when the key value proposition was no ads? [THR]

I Wrote...

This post from June holds up pretty well with the race well underway to undercut NVIDIA. At the same time, NVIDIA is currently worth $3.65T – a full $250B clear of #2 Apple at this point. That's more than two Intels. Which will be the first company to $4T?

Can NVIDIA Become Intel Faster Than Everyone Becomes NVIDIA?

As money pours into AI, can NVIDIA stay ahead?

Spyglass •M.G. Siegler

I Link...

• A tweak to the iPhone Camera Control settings in the new iOS 18 beta will let you enter into camera one with one click even if the screen is off – this would have probably been my number one feature request. [9to5 Mac]
• Andor – one of the few Star Wars properties that Disney has handled well – will be back in April for the second (and final) season (since it ends, presumably, right where the movie Rogue One begins). Four years will be covered in 12 episodes. [Deadline]
• No one seems more excited about Donkey Kong Country inside of Super Nintendo World than Shigeru Miyamoto in his video walk-through of the new section of the park opening next month in Japan. I'm going to guess that a Hyrule area is next. Could they do it in time for the Zelda movie? [Verge]
• Apple has made a hardcover book with Assouline to commemorate their '100 Best Albums' list. The book looks beautiful. The list is ridiculously bad. Recency bias, much? But I suppose all such silly lists are really just to get people talking. Though I also don't really recall any mention about this list until now? So maybe the next step to get people talking in a good or bad way is a $450 book? [MacRumors]
• The EC is now threatening Apple over their "geo-blocking" of services in EU countries. One can imagine Phil Schiller just throwing his hands up at this point. "Yes, we do that. On purpose. Nice job running a big investigation to find out what a web search could have told you." [9to5Mac]
• Europe is getting... a new search engine? [Wired]
• Gary Lineker stepping down as the host of Match of the Day is a big deal over on this side of the pond, as it's a post he's held for over a quarter century. And the former player is good at it. So much so that he's the BBC's highest-paid talent. Not to worry, he'll still do the World Cup in 2026. [THR]

I Quote...

"The 2010s were the age of scaling, now we're back in the age of wonder and discovery once again. Everyone is looking for the next thing. Scaling the right thing matters more now than ever."

-- Ilya Sutskever, the co-founder of OpenAI and more recently, Safe Super Intelligence, talking about the "plateau" that has been hit with the methods of pre-training with unstructured data for AI models.

While Sutskever is obviously conflicted on this topic given his departure from the leader in those "old models" and his desire for SSI to lead in any new field, as discussed yesterday, there's a lot of chatter about the slowdown and the race for the next breakthrough in AI.