Uncanny Valley Podcast: Episode Show Notes & Transcript
This week on WIRED’s Uncanny Valley, co-hosts Brian Barrett and Zoë Schiffer break down the biggest takeaways from Nvidia’s annual developer conference, unpack why Tesla just sparked backlash among its most diehard online fans, and explore what Meta’s original plan to shutter Horizon Worlds on its Quest VR headset says about the end of early metaverse hype. (To clarify: Meta later walked back the full shutdown, announcing it will keep the platform running on limited support for the foreseeable future.)
Articles referenced in this episode:
Nvidia Prepares to Launch an Open-Source AI Agent Platform
Tesla Influencers Are Walking Away From the ‘Cult’
Meta Is Pulling the Plug On Horizon Worlds for Meta Quest
You can follow Brian Barrett on Bluesky at @brbarrett and Zoë Schiffer on Bluesky at @zoeschiffer. Reach out to the show team at [email protected].
How to Listen
You can stream this week’s episode directly via the audio player on this page. If you want to subscribe for free to get every new episode automatically, here’s how:
- On iPhone or iPad: Open the default Apple Podcasts app, or tap this link to go straight to the show. You can also use third-party podcast apps like Overcast or Pocket Casts—just search for “Uncanny Valley” to find us. We’re also available on Spotify.
Episode Transcript
Note: This is an automated transcript, so it may contain minor errors.
Zoë Schiffer: Brian, hi! It’s so fun getting another way to chat with you that isn’t me pinging you on Slack every five seconds.
Brian Barrett: It really is, because Slack doesn’t have the whole voice thing going for it.
Zoë: You’re not wrong.
Brian: I do have to say, it’s such a bummer Leah isn’t joining us today.
Zoë: I know, it really is. But when the cat’s away, the mice will play—we’re talking about all the topics Leah hates, so just wait and see.
Brian: Just to be clear, she’ll be back next week! She’s just out sick today.
Zoë: Yep. It’s allergy season, after all.
Zoë: Welcome to WIRED’s Uncanny Valley. I’m Zoë Schiffer, WIRED’s director of business and industry coverage.
Brian: I’m Brian Barrett, executive editor.
Zoë: This week on the show, we’re digging into key announcements from Nvidia’s annual developer conference, why a growing number of Tesla influencers are stepping away from the brand, and why Meta just pulled the plug on its Horizon Worlds metaverse platform for Quest. Let’s kick things off with Nvidia: the company held its annual conference this week in San Jose, which is the biggest yearly event for the entire AI industry. Some people even call it the AI Super Bowl. Developers, CEOs, researchers, even WIRED reporters all show up, and everyone waits to hear what CEO Jensen Huang has to say about where the company is headed next.
Brian: One really interesting thing about Nvidia’s conference is that most of it is geared toward business, not regular AI users who mess around with tools like Claude. You might not connect with a lot of the announcements as an end user. But even with that, Jensen dropped a pretty wild number: he said the total revenue opportunity for AI chips just at Nvidia could hit at least $1 trillion by 2027. Take that with a grain of salt, of course—he’s the one who stands to gain from that projection.
Zoë: Pocket change, really.
Brian: Right? It’s nothing for Nvidia at this point. But the most exciting update was that he actually launched a tangible new product, not just a promise of something coming down the line. A while back, Nvidia struck a licensing deal with a company called Groq—spelled with a q, not like X’s Grok that had all that non-consensual deepfake controversy.
Zoë: Yep, Groq with a q.
Brian: Exactly. So they’re pairing Nvidia’s AI processing chips with Groq’s chips, which add components that boost how Nvidia’s chips run. That $20 billion licensing deal is finally paying off: this new pairing will make AI inference faster, cheaper, and way more efficient for all of Nvidia’s customers.
Zoë: That’s right. I talked to a bunch of industry folks about this this week, and one thing they pointed out that’s obvious to AI researchers but totally not obvious to people like me is that we haven’t actually had purpose-built chips for AI this whole time. Everyone’s just been using general-purpose Nvidia chips for both training and inference this entire time. This is the first year we’re actually going to start seeing specialized chips built specifically for AI.
Brian: It’s wild to think about—even though we joke about my age, it wasn’t that long ago that Nvidia got to where it is today making GPUs for gaming PCs. Those GPUs just happened to work really well for the workloads AI needs, so the company sort of stumbled into this AI dominance by accident. That makes this moment a really big deal. Hey, we’ve said “inference” a couple of times already—can you break that down for listeners who don’t work in AI so we’re all on the same page?
Zoë: Totally. If you’re an AI researcher, you’re probably yawning right now, but this isn’t common knowledge outside the industry, so it’s worth spelling out. First you have the training process: that’s when you let an AI model digest huge amounts of data from across the whole internet, so it can learn patterns and information. Then inference is the step when you, as an end user, ask ChatGPT or Claude a question, and the model generates an answer for you in real time. Right now, most of the money that AI companies and big tech firms are investing goes to inference, not training.
Brian: That’s because they’ve already digested basically the whole internet at this point.
Zoë: Exactly. And inference is just really expensive. Serving all those users in real time costs an absolute fortune.
Brian: For a little example of how much Jensen is leaning into inference, he closed his keynote with a pretty bizarre AI-animated music video all about it. Let’s recap that:
Archival audio: Once upon an AI time / training was paradigm / short talking models how / but inference runs the whole world now / Vera showed us who's the boss / at 35 times less the cost / Blackwell makes the token sing / Nvidia, the inference king.
Zoë: I really hope they used AI to make that, and didn’t pay a marketing firm millions of dollars for that.
Brian: Yeah, the quality is exactly what you’d expect from AI, honestly. For context, Blackwell and Vera are the names of two new Nvidia product lines.
Zoë: We should also mention that Jensen announced NemoClaw, an enterprise platform for building AI agents—essentially a secure, business-focused version of OpenClaw.
Brian: It’s so fun watching every company scramble to get in on this trend. Nvidia has NemoClaw now. The original creator of OpenClaw—what’s it been called lately, anyway?
Zoë: Right, it’s been Clawdbot, Moltbot, now OpenClaw.
Brian: Right, so the creator is now at OpenAI, and Meta bought Moltbook, that social network for AI agents. Everyone’s rushing to get a piece of this, mostly just so they can say they’re ahead of the curve, it feels like.
Zoë: Exactly. You can almost hear the backroom meetings: Mark Zuckerberg has to prove to investors that Meta isn’t falling behind on the latest trend. As soon as one company does something new, every other company rushes to copy it as fast as they can.
Brian: Yep. Speaking of which, I’m actually building my own AI agent social network focused on cooking. It’s a whole thing, don’t worry about it. I also wanted to bring up something I’m weirdly fascinated by: everyone’s obsession with space-based data centers. Just like we were talking about with companies rushing to hit the latest trend for investors, Nvidia also announced the Space-1 Vera Rubin Module, a GPU and computer built for use in space. There’s no actual timeline for when it’ll launch, but they’ve got their top people on it, Zoë!
Zoë: I love this topic. I talk about this with smart researchers all the time, and anyone who knows physics immediately starts sweating and swearing, asking how the hell you cool a data center in space. They get so worked up. Anyone who actually knows the space industry will tell you this is totally far-fetched—we’re decades away from figuring out how to power, cool, and run a full data center in orbit. But this is just a marketing play. Why? Because so many big tech companies are gearing up to go public soon: OpenAI, SpaceX, Anthropic—we’re going to see a ton of these big, out-there announcements ahead of IPOs.
Brian: Right. While all this big future-focused news is happening, there’s actual real-world competition popping up right now. Nvidia’s chips aren’t the only game in town anymore: Google is building its own custom AI chips, startup Cerebras makes specialized AI chips, Meta is working on its own too.
Zoë: Yeah, even companies like OpenAI and Meta are working with third parties to design their own custom chips. That means Nvidia is having to defend its spot as the leader of the AI industry in a way it hasn’t had to in years, if not over a decade. Jensen knows that, so he’s making moves to shore up his position. Do I think Nvidia is going to fall way behind or have to panic about its business any time soon? No, I really don’t. What do you think?
Brian: I agree. It’s just that we’re finally entering a new phase where the market isn’t just all Nvidia, all the time.
Zoë: All our GPUs in one data center, right?
Brian: Exactly! Wow, that was good.
Zoë: I know, right?
Brian: OK, speaking of disappointments, let’s shift over to Tesla. Don’t @ me, Tesla fans.
Zoë: Wow, he’s already on the BYD payroll.
Brian: Guilty, I love BYD. No, I’m not talking about Tesla as a whole, just the recent controversy the company found itself in. Tesla had offered a limited-time promotion that let you transfer your “lifetime Full Self-Driving” subscription to a new Tesla vehicle. First off, we should note that “Full Self-Driving” is a misleading name—there’s nothing fully self-driving about it yet, that term is super controversial. Still, super fans were excited because FSD is really expensive, so being able to transfer it for free to a new car was a big deal. Then Tesla changed the terms at the last minute: they said you have to take delivery of your new car by March 31 to qualify for the transfer. That’s a really tight timeline for a lot of buyers, so people got furious. Even the most loyal, ride-or-die Tesla fans are turning on the company now. There’s a growing group of ex-Tesla superfans and influencers who are distancing themselves from the brand, which I think is really interesting. This isn’t just the broader backlash against Tesla over Elon Musk’s politics, though that’s happening too in the US, Europe, and the UK. These are Elon fans who are saying they’re over it.
Zoë: This is such a fascinating shift. Tesla’s stock price has almost always outperformed its actual business fundamentals, and a big reason for that is Elon’s intentionally cultivated a rabid, loyal fan base that buys Tesla stock no matter what. It’s one of the most widely held retail stocks on platforms like Robinhood. If you’re a fan of Elon, you stick with Tesla through the ups and downs of the business, which gives Elon a ton of personal power. Even if this is just a small number of big-name influencers turning away right now, not a widespread fan base revolt, that’s still a really big problem for him.
Brian: How big of a problem do you think this is, really? Elon keeps saying Tesla isn’t even a car company anymore—it’s a humanoid robot and robotaxi company. I always forget the robotaxi part, even though that’s the one that’s closer to being real. The stock price is already priced for that future, right? Is there a world where this dissatisfaction with the car business doesn’t matter at all because the future is all robots and self-driving fleets?
Zoë: You could be right. Elon is first and foremost an incredible marketer and storyteller. If he can sell the story that Tesla is now a robot company, and investors buy that, then this might not matter. But the thing is, a lot of these super fans aren’t just Tesla car fans—they’re Elon Musk fans, and Tesla is just the product he makes that they can latch onto, promote, and hype up on X all the time. Maybe that loyalty transfers over to the robot story, but if his popularity with his core fan base is fading, and he keeps doing things that push his own community away, that’s a problem. This started for a lot of people when he got deep into partisan politics. I talked to a bunch of Cybertruck owners last year for a story, and multiple people told me they love the car, but it’s such a bummer to drive it in public now because it’s become a political lightning rod. They didn’t buy it to signal they support Trump, but Elon got involved with Trump and the DOGE movement, so now people yell at them and even try to cut them off on the road when they drive it.
Brian: If you haven’t read Zoë’s Cybertruck photo essay from last year, go check it out right now. She interviewed dozens of owners, and there’s one quote that’s stuck with me forever: “Women don’t like the vehicle.”
Zoë: Yep, that’s exactly what he said.
Brian: So, Zoë, you’ve had thoughts on this before—what do you think of people who bought their Tesla before Elon’s very public turn, and now put a bumper sticker on it that says something like “I bought this before Elon went crazy”? How do you feel about that?
Zoë: I’m so glad you asked, and I’m sorry if this offends anyone, but I think it’s so embarrassing. You have two options: either drive your Tesla loud and proud and skip the sticker, or sell it and buy another EV. Teslas aren’t even the best EVs on the market right now! I live in the Bay Area, so I see these cars everywhere, and this middle ground is the worst possible choice. Just pick a side!
Brian: To be fair, there are fewer EV options than ever right now. Honda canceled three planned EV models for the US, and a lot of domestic automakers are pulling back on EV plans. It’s a weird time where Tesla is still one of the most widely available options for people who want an EV.
Alright, let’s move on to another big call from Meta this week. Meta announced Tuesday that starting March 31, Horizon Worlds will no longer be available on the Quest store, and it’ll shut down entirely on June 15. For anyone who forgot, Horizon Worlds was Meta’s big metaverse bet: it was the flagship virtual world the company built to be the face of the metaverse, designed to be experienced on Meta Quest VR headsets. You’ll still be able to access it on your phone after the shutdown, but that’s not really a metaverse—that’s just… a regular app, at that point.
Zoë: Right! It’s so wild that the company changed its entire name to rebrand around this product.
Brian: I know, right? I checked when the news broke: it was only four and a half years ago that they rebranded from Facebook to Meta.
Zoë: Meta poured so much money into this. If you talk to Meta executives and other metaverse true believers, they’ll say they were just too early, that the metaverse vision could still happen someday, the Ready Player One future is still possible. But the reality now is that AI blew up and stole all the attention and investment. Meta is playing catch-up on AI right now, and we still haven’t seen what will come out of all the money they’ve poured into Alex Wang’s new AI labs.
Brian: Meta lost a reported $77 billion on its Reality Labs division over four years, for context. That’s a ton of money to lose on a bet. But I have a unifying theory for consumer tech, do you want to hear it?
Zoë: Absolutely, hit me.
Brian: It’s simple: every big new consumer tech bet is 3D TV until proven otherwise.
Zoë: OK, you have to unpack that.
Brian: Back when 3D TVs were the big hype, for two whole years every TV manufacturer insisted 3D TV was the future. You couldn’t even buy a high-end TV that didn’t have 3D built in—every top model came with half a dozen pairs of 3D glasses. So this fits two big red flags: first, companies will insist something is the future no matter what all the evidence says, until they lose tens of billions of dollars and have to backtrack. Never take a company’s “future of tech” hype at face value. Second, people don’t like wearing extra things on their faces.
Zoë: That’s such a good point.
Brian: And Horizon Worlds and the metaverse checked both boxes. On top of that, it created more separation between people instead of bringing them together, which was Facebook’s whole original mission! Wild.
Zoë: Yeah, that hasn’t gone well for Meta on that front either. But that vision of the future where everyone sits alone in their house with a headset on, interacting with virtual avatars of friends or strangers? I hate to sound preachy, but that’s such a grim, depressing vision of what life should be. It’s no wonder it never caught on with most people.
Brian: Let’s shift back to AI, since that’s the new big bet for Meta and every other company. It’s easy to have déjà vu here: everyone’s saying AI is the future, they’re spending hundreds of billions building new data centers, that’s way more than Meta lost on the metaverse. What’s the difference between AI and the metaverse, though? Why isn’t AI just another 3D TV?
Zoë: I don’t think you can really compare the two. AI is a foundational innovation that you
