AI大基建 + H-1B签证乱局 + TikTok备战“被提”

Let me ask you this. Settle a bet. What do you think is a better name for a golden retriever: Bubbles or Potato?

Male or female?

Female.

Bubbles.

That’s what I said. My brother and his family were getting a new dog, and they sent me pictures of this golden retriever. My nephew said, “We’re thinking Bubbles or Potato.” And I said, “Well, I think it has to be Bubbles because dogs want a nickname.”

Yes.

And if it’s Bubbles, then it’s Bubbs, it’s Bubba.

Sir Bubbs-a-lot.

Sir Bubbs-a-lot.

Yes.

You can imagine—

We could go on, yes.

If it’s Potato—

Yeah.

It’s what? Tate? Ta—

Tater.

Tato, Tater.

Popo.

Popo. I wouldn’t call it that.

That one’s taken, yeah.

I lost, and now the dog is named Potato.

No.

Yeah.

Come on.

Mm-hmm.

We gotta change this. It’s not too late.

It might be too late.

What? The dog doesn’t care.

That’s probably true.

Yeah.

Potato is just a puppy, so my dad saw a picture and he said, “It’s really more like a French fry at this point,” and I thought that was cute.

That is such a good dad joke.

Right? I’m Kevin Roose, a tech columnist at The New York Times.

I’m Casey Newton from Platformer.

This week—

An eye-popping deal between OpenAI and Nvidia has us asking: Is this AI boom turning into a bubble? Then, the chaos over visas. Immigration expert Jeremy Neufeld joins us to explain how Trump’s new $100,000 visa fee could hurt the economy. And finally, the rapture that wasn’t. I explain to Kevin how TikTok went wild for the end of days.

1. The Great AI Buildout

Heaven help us. Well, Casey, it’s time to put on our work boots and roll up our sleeves because today we’re talking about the great AI buildout.

Oh, I thought you were going to say we were joining the Village People.

No. This week we got some news that I think is thematically related to what I have started calling the Great Buildout. This is our big national infrastructure project to build all of the data centers, acquire all of the compute, and make all of the investments that are necessary for the next phase of AI development.

So, on Monday, Nvidia and OpenAI announced that they had struck a $100 billion deal. This deal will be an investment that Nvidia is making in OpenAI, $10 billion at a time until the full $100 billion is reached. Nvidia is giving OpenAI cash and getting equity in return. This is designed to help OpenAI build out massive new data centers that will run on Nvidia’s AI chips.

Yeah, and just to put that number in perspective, if I walked into a bank right now and said, “Can I have $100 billion?” they would say no.

I don’t know. I hear the newsletter business is very lucrative.

So that was announcement number one: this new Nvidia-OpenAI deal. Then, on Tuesday, we got announcement number two, which was that OpenAI had announced 5 new sites for Stargate, their huge infrastructure effort. They said that they were developing 2 new sites with Oracle in Shackelford County, Texas, and Doña Ana County, New Mexico, along with a third site in an undisclosed location in the Midwest.

They’re also developing 2 additional Stargate sites in partnership with SoftBank, and those sites will be located in Lordstown, Ohio, and Milam County, Texas. Altogether, these new sites will bring Stargate to nearly 7 gigawatts of planned capacity and over $400 billion in investment over the next 3 years.

So today we’ll talk about some of this new infrastructure news and what it means for the direction that AI is heading, just the vast, unprecedented scale of the investments that are being made in AI right now. And then we’ll talk about the strangeness surrounding these various deals, why we are seeing the biggest companies racing forward with no plans of slowing down, and what it means for the future. And before we do that, I think we should make our AI disclosures.

Sure. What’s yours?

Mine is that The New York Times Company is suing OpenAI and Microsoft over copyright violations.

And my boyfriend works at Anthropic.

Okay. So, Casey, that’s a lot of big numbers that I just threw at you and threw at our listeners. Maybe we should, before we talk about what this means, just give some context for the scale that we’re talking about here.

Yeah. One number that I would throw out is $71.4 billion, which is the amount of money that OpenAI had raised before this week, right? So, in the entire history of the company, they had only made it to $71.4 billion. And now along comes Nvidia, and in one fell swoop promises to eventually invest more than OpenAI has invested to this point.

We’ve already been at a point where people are saying, “Hmm, I don’t know. Sure seems like these companies have raised a lot of money and they’re not profitable, and everything they do is incredibly expensive.” And yet, instead of moderating at all, these companies are stepping on the gas pedal.

Yeah. The comparison I’ve been trying to use to illustrate the scale of investment that’s happening now is with the Interstate Highway System, which was built over a 36-year period at a cost of roughly $300 billion adjusted for inflation. That’s in today’s dollars.

On a recent earnings call, Jensen Huang, the CEO of Nvidia, estimated that this year alone, U.S. companies would spend about $600 billion on AI data centers. In other words, this year, the AI industry is projecting to spend basically twice as much as the cost of the entire Interstate Highway System in 1 year.

It’s almost like they’re building an information superhighway system.

Get out of here.

All right.

So that’s one comparison that, for me at least, helped drive home the scale of this. Are there any comparisons that have been helping you drive home the scale of this?

Yeah. I decided to do a little research. I wanted to try to reach our Gen Z listeners, because we have a lot of Gen Z listeners. And so I tried to estimate how much it would cost to build 1 Labubu factory, okay? Labubus are very hot right now. Everybody’s trying to get their hands on one of these things.

According to the estimates that I was able to put together, if you were to build a new Labubu factory from scratch, it would cost you between $15 million and $20 million. So that means that, Kevin, for this $100 billion that Nvidia may now invest in OpenAI, they could have built 5,000 Labubu factories.

Wow.

I hope that puts that into perspective for folks.

Yeah. I can’t believe the Trump administration is not making domestic Labubu production an urgent infrastructure priority.

It’s really my number 1 issue with the Trump administration.

2. The AI Deals Get Weird

So, obviously, the scale of these things is impressive and massive, and we could just go on talking for half an hour about the big numbers here and what they mean. But I think it’s also worth pointing out that some of these deals are just extremely weird.

Yeah. These are not the standard deals that you get when you sign up for Y Combinator with your startup and give a little piece of equity in exchange for some amount of money.

Right. One thing that makes them especially weird is that many of these companies that are making these huge investments in each other are also buying goods and services from each other. So, with OpenAI and Nvidia, Nvidia is investing billions of dollars in OpenAI. OpenAI is going to turn around and spend many of those billions on Nvidia chips.

This is sometimes called round-tripping. I’ve also heard it called “left pocket, right pocket,” since you’re taking money out of one pocket and putting it into another. Some people call it vendor financing.

What other names do you have for it?

Chicanery is what I call it.

Mm.

But this is one of these deals that is raising a lot of people’s eyebrows because they’re basically saying, “Wait, you’ve just created this giant circular money machine that’s moving money around in the economy.”

Yeah. But it’s also an incentive for OpenAI to keep Nvidia as its major chip provider, right? We know that Nvidia is the best in the business when it comes to making AI chips, but they do have competition. Google makes its own AI chips. Amazon makes its own AI chips. AMD makes AI chips. Earlier this year, OpenAI struck a deal with Broadcom to make its own custom chips.

So one way of looking at this is Nvidia is saying to OpenAI, “Hey, stick with us, and we’ll give you most-favored-customer status when it comes to getting state-of-the-art chips.”

Yeah. There are some other weird things about some of these deals that have been happening. I’m spending some time looking into the financialization of some of these deals and how they’re being financed.

Sam Altman this week had a blog post called “Abundant Intelligence,” where he hinted that OpenAI was coming up with some creative new financing strategy that they would share more details about soon to pay for all of this. When I hear phrases like “creative new financing strategy,” that’s when you know you should start looking for your wallet.

Well, one thing they might just do, though, is lease the chips, right?

Right.

So you can get them at a lower price because you’re not buying them outright. There’s a public company called CoreWeave. That’s their entire business model: leasing chips to other companies. So that might be one creative thing that they do. But I would agree with you, Kevin: the more creative we get from there, the more trouble we might be in.

3. The AI Bubble Question

Yes. So I have to ask: Is any of this—the giant announcements, the weirdness around some of the financial aspects of this—making you think any differently about whether we’re in an AI bubble or not?

Well, yes, because the numbers are, one, large in unprecedented ways, and two, we’re not yet connecting this to any reasonable expectation of profits, right? This isn’t the traditional Silicon Valley, “Hey, we lose money for a couple years, but then we turn on the monetization spigot and everything just works out.” This is, “We’re spending hundreds of billions of dollars well in advance of any expectation that that is going to be profitable.”

And the bigger the numbers get, Kevin, the harder it is to make the money back. So I do think it is a moment to pause and reflect on the sheer magnitude of what is being invested.

Yeah, I think that’s right. What sticks out to me is just that you now have the leaders of the biggest technology companies in the world saying effectively in unison that they do not care how much it costs to build all the way to AGI or superintelligence, or wherever they think the final destination of this technology is going to be.

Mark Zuckerberg gave an interview to Alex Heath last week where he said that he doesn’t worry about overinvesting in AI. If anything, he’s worried about underinvesting in it. Other companies have said similar things. Larry Page, the co-founder of Google, has been reported to have said something like he would rather go bankrupt than lose the race that they are in.

So I think it’s really just unusual, not just in the sense of how big the investments are, but in the fierce, single-minded tunnel vision of all of the leaders of these companies who are saying in very clear terms that they know they are entering a bubble. They know that they are entering a crazy, unprecedented environment for investment in AI. They just don’t care, and for the most part, no one can stop them.

No, I mean, I think it’s worth saying just how much ego is wrapped up in this, right? No one wants to be the fourth-place finisher in the race to build AGI because that is going to dramatically affect how your obituary gets written, if it gets written at all, because, of course, you might be able to upload your mind to the cloud.

So yes, this is about business, and yes, this is about creating shareholder value, but a big part of it is, “I personally want to be the one who builds AGI because that’s going to make me feel super cool,” and that is justifying an enormous amount of spending.

I think that’s right. I think there’s ego. I think there’s FOMO. I also think there’s a—

Ego and FOMO: dangerous combination. The dreaded ego-FOMO combo.

I also think there’s a real logic to what they’re saying. If this is a winner-take-all technology, and we’ve seen several of those in recent decades, then you really do want to be first. Even if you only have a 6-month lead on your nearest competitor, if you do buy the logic of the intelligence explosion, where AI reaches this point at which it starts to be able to recursively improve itself, and you have robots building better robots and building factories to build more robots, then 6 months could actually amount to a meaningful difference in your trajectory.

And so, look, I don’t know that Larry Page and Mark Zuckerberg and Sundar Pichai and all these people, including Sam Altman, are running through—I don’t know if they have a detailed financial model anywhere on their computer for how all of this is going to pay for itself. But I do know that they think that it will at some point pay for itself.

Yeah, which is a fairly common thing for people spending insane amounts of money to say: “We have a plan. Don’t worry about it.”

Right.

Mind your own business.

Right. So, Casey, if I’m hearing you correctly, you are worried about a bubble but not so much that you think your alarm bells are going off quite yet.

Well, look, here’s the thing. If you can build a software system that can automate vast amounts of labor to the point that big companies no longer have to hire thousands of people, yes, I do think in the end you will make your money back on these investments, even the ones that have 9 zeros in them, right?

The question is how many of those companies are there going to be. I do think there will be at least 1. I’m not sure there are going to be 3. There are probably not going to be 7. So I understand why all of these folks are jockeying for a position here, because while this race may not be winner-take-all, it probably will be winner-take-most, and to some extent that should justify really, really huge investments.

Yeah, I mean, I guess what I would say is I am not convinced yet that we are in a dangerous speculative bubble when it comes to AI. But I am starting to move some of my probability mass onto these fears about not necessarily a bubble, but just the overall distorting effect that such a large, fast investment has on the rest of the economy. I’m starting to worry about that maybe a little bit more than I was a couple weeks ago.

That makes sense. I’m glad we are talking about this because I do think it is inevitable that some of these businesses are going to fail. And when that happens, I want our listeners to know that we’re telling you that is probably going to happen, right?

We talk a lot about AI on the show. We talk about some good things that it can do. We talk a lot about some bad things that it can do. But when it comes to the business fundamentals, they’re not all sound across the entire industry, and an unprecedented amount of money is being invested. It would also be unprecedented if everyone making those investments made all their money back. So we are really betting against that, and frankly, it’s a pretty easy bet to make.

Yeah, and I think if I were to leave our listeners with one thing about this great AI build-out, it is this: We are not talking about the future here. These investments are happening now. There are shovels in the ground. There are bulldozers going onto the construction sites. These buildings are going up. Many of them are already operational.

We are already starting to see the effects that these data centers are having on things like electricity prices for consumers. This is showing up at the highest levels of our national politics. This build-out is now our great American infrastructure project of the 2020s, and it remains to be seen what the effects will be.

But I don’t want to leave people thinking, “Oh, this is just Casey and Kevin talking about something that may or may not happen.” This is happening now, and while we don’t know what the consequences will be yet, we are now one nation under AI capital expenditures. I just think we need to internalize that, because I think it does really make it clear that every person in America has an interest now—whether it’s just a financial interest or otherwise—in what is happening in AI.

Mm-hmm. Yeah.

Should we go to a data center?

We’ve been invited.

Have we?

Yeah, OpenAI invited us to one.

Should we go?

How about dinner? Let’s go.

Let’s go. When we come back, we’ll sort through the chaos of Trump’s latest policy changes for immigrants in tech.

4. Trump’s Visa Chaos

Well, Casey, the other big story of the week in tech is this madness and chaos going on with visas.

Yeah. Now, I use a Mastercard, so I haven’t been following as closely, but is this important?

Different visa.

Oh.

This was a proclamation that President Trump released on Friday of last week, called “Restriction on Entry of Certain Nonimmigrant Workers,” which made a bunch of big changes to the H-1B visa program. This is the program that allows immigrants from high-skill occupations to come and work in the country.

President Trump announced that, going forward, there would be a new $100,000 fee for employees who are hired under the H-1B visa program. I started hearing about this right away from people who were saying, “Oh, my God, what does this mean?”

The tech companies started sending memos and emails to their employees who are on H-1Bs, saying, “Hey, we’re still trying to figure out what’s going on here, but if you’re outside the country, you need to come back immediately, and if you’re inside the U.S., don’t leave because we’re not sure if you’re going to be allowed back or not.”

So there was just a lot of confusion about this new program, the new fees, who it applied to, and who it wouldn’t apply to. It just seemed to be a big mess.

Yeah, it was a weekend full of absolute chaos. My own friends who are immigrants and I have been exchanging a lot of messages as we try to figure out what is going on here. And so we thought, well, we should try to tell that story on Hard Fork this week and tell you what’s going on as best as we can.

Yeah, so there is still a lot we don’t know as of the time of this recording. The White House has tried to clear up some of the confusion about these changes. They’ve said this new fee will apply only to new applicants for H-1B visas. It will be a one-time fee, not an annual fee, and anyone currently on an H-1B visa can come and go as they always have.

But, Casey, I think it’s important to talk about this this week because the technology industry in America is and always has been dependent on immigrants.

Mm-hmm.

Immigrants run many of our largest technology companies. In fact, Sundar Pichai, the CEO of Google, and Satya Nadella, the CEO of Microsoft, among many others, have had H-1B visas at various points in their careers.

AI, in particular, is a part of the tech industry that relies on a lot of foreign-born talent, and the tech industry was the target of these policy changes. And so I think we should examine them, try to figure out what’s happening, and bring in someone who can really help us unpack what’s going on.

Yes. So to help us navigate all these changes, we have invited on an immigration policy expert to walk us through what’s happening. His name is Jeremy Neufeld. He’s the director of immigration policy at the Institute for Progress, which is a nonpartisan think tank for accelerating scientific, technological, and industrial progress.

He has written a lot about how these changes will impact not just startups and big tech firms, but also nonimmigrant American workers and the wider U.S. economy. He’s got a lot to say. Let’s bring him in.

Yes. As the Trump administration is now saying no to applicants for H-1B visas, let’s bring him in.

Jeremy Neufeld, welcome to Hard Fork.

Thanks so much. Thanks for having me.

So you focus closely on immigration. Tell me about your weekend. Was this one of those where your phone is ringing off the hook starting on Friday?

Yeah, so on Friday, I first heard that there was going to be this proclamation and signing in the White House about something related to immigration policy. This was not on anyone I know’s radar before that happened.

And then, from the time that they made these announcements up until the time we’re talking now, I’ve just been in nonstop conversations about it, trying to figure out what’s going on, how it’s going to affect the United States, and what it all means.

I want to take a step back and understand the history of this visa in particular. We have a lot of different kinds of visas in this country. I have some friends who work in tech who are on what they call O-1 visas for individuals with extraordinary ability. What is the H-1B system? What was it originally designed to do?

5. How the H-1B System Works

Yeah, so the H-1B visa is our flagship high-skilled immigration visa. It’s intended for people in specialty occupations. And there are 85,000 of these H-1Bs that are available each year for the private sector, and then universities and national laboratories get access to as many H-1Bs as they want without being subject to that cap.

But I think one way to think about the H-1B system is that it’s really a couple of different talent pipelines under the same umbrella of the H-1B. So, on the one hand, it is an incredibly valuable source for top talent, for international scientists, and for top founders of tech companies like Elon Musk.

And then, at the same time, it’s the same visa that’s being used for lower-level, lower-skilled talent, like IT consultancy services work. And so I think one of the things that happens with a lot of the debate around this visa is that the defenders of the program are pointing to very true things. They’re pointing to the amazing contributions that H-1B holders have made for the United States, for our scientific and technological leadership.

The critics are also pointing to real cases of certain companies, like Infosys and Cognizant, taking a huge number of these visas and bringing in relatively low-skilled workers who might be undercutting Americans. And both of these sides have a point, because the H-1B program is not one thing. It’s a lot of different kinds of things.

Yeah, I think it’s worth drilling into that a little bit more, because what I’ve been hearing from people who are defenders of the H-1B program is, you know, all these famous tech people came here on H-1B visas at some point. Sundar Pichai, the CEO of Google, was on an H-1B visa. Satya Nadella from Microsoft, Eric Yuan, the founder and CEO of Zoom—I mean, you can go on and on and on.

We have so much top tech leadership in this country that was reliant at some point in their careers on this visa program. But then you also have these outsourcing and IT firms that sometimes people in tech will pejoratively refer to as body shops. What are they doing, and how much of the overall H-1B program do they represent?

Yeah. The basic reason all of these different pipelines exist is because employers want far more H-1Bs than there are H-1Bs available each year. So last year, about 400,000 workers got entered into the lottery. There are only 85,000 slots, and so what that means is that there’s an incentive for certain kinds of business models that can play into that basic lottery.

These body shops have a business model of spamming the H-1B lottery with lots of petitions because they don’t really care which individuals are going to win the lottery.

Hmm.

All that they care about is getting some number of visas so that they can then contract out those workers to third-party companies. So they’ve ended up increasing in scale.

Right now, H-1B-dependent companies—that’s companies where a significant share of their workforce are on H-1Bs—represent about a third of all of the visas that get given out. Outsourcing companies are a subset of those companies. These are companies that specialize almost entirely in contracting services. They are about 15% of all of the capped H-1Bs.

Now, where do we think this $100,000 number came from? I saw some speculation on X that someone had tweeted this out a while ago and said, “Oh, they should fix this program by making everyone pay $100,000 to get an H-1B visa.”

I don’t know whether that was actually in the administration’s feed when they decided on that number, but do you have any better sense of where that came from? And why would charging companies or workers $100,000 to get an H-1B visa result in fewer of the kinds of immigrants that the Trump administration doesn’t want and allow others to come in?

Oh, I think I can answer that one. Wait, can I guess? I want to see if I get it right, to see if I’ve been following the discussion.

Yes.

Is it because the outsourcing companies would never pay $100,000 for one of these workers because their wages just aren’t that high and wouldn’t be worth it to them?

That’s the basic theory.

Yes.

Whether that ends up being true remains to be seen. The first thing I just want to mention is that the proclamation does 3 different things at once.

6. The H-1B Fee Backfires

The first is the fee. The second is a prioritization within the lottery for people who are more experienced and are older. And then the third is it has the Department of Labor try to beef up requirements about what people are required to be paid.

And the intention is supposed to be to make it harder for the outsourcers to gobble up so many of these scarce slots. That’s what the administration says that they’re trying to do.

But what these policies would end up doing in effect is cutting out a lot of the high-value talent and potentially leaving the outsourcers untouched. And the reason for this is, first of all, the outsourcers have access to another kind of visa called an L visa for intracompany transfers.

And the fee has a major loophole, which is that people who are switching from one visa type to an H-1B, they don’t have to pay the fee as long as they’re already in the United States.

And so what I think we'll see is that the outsourcers can adapt. They can bring people here on L visas—

Hmm.

—and then they can transfer them over to H-1Bs without paying the fee in a way that a lot of other companies can't adapt because they're not multinationals.

Hmm.

And then the second way that the outsourcers are helped is, I think, entirely unintentional. The Department of Homeland Security was instructed to prioritize higher wage levels in running the lottery, and what they've settled on is something called a wage level classification constructed by the Department of Labor.

But the important thing to recognize is that wage levels are not the same as actual wages. Wage levels are dividing the salaries within an occupation. The most experienced, most qualified acupuncturist in the world might be paid $60,000, and so they'd be classified as wage level IV, the highest wage level, because they're really highly paid compared to other acupuncturists.

A recently minted Ph.D. coming out of a top university is at the start of their career. They're about to go get a job at OpenAI paying $300,000. But because they're right at the beginning of their career, they're going to be classified as wage level I or wage level II because it's entry-level.

And so the way this wage-level system works is it would favor the acupuncturist over the AI scientist because it looks only within the occupation. The way that helps the outsourcers is that the outsourcers bring a lot of older, midcareer staff. So even though they're paying less than other employers, they bring in midcareer people, so they get classified at higher wage levels.

The end result of all of this is that the outsourcers will come out of this actually ahead. They'll get about 8% more visas under the proposed policies, even though the administration says they're trying to clamp down on the outsourcers.

Got it. You mentioned that universities, research institutes and hospitals have not been subject to this lottery system that we have. Are they going to have to pay the $100,000 fee? And how do we think it will affect jobs in those categories?

Yes. This $100,000 fee is essentially a major tax on universities and national laboratories for recruiting foreign-trained talent. There's no problem in principle with using some mechanism to shift scarce visas to the people who are going to contribute the most, but for universities and other organizations that aren't playing in the same zero-sum game that the lottery creates, this is just a pure cost.

While big tech might be able to afford a $100,000 fee, universities trying to bring in the best associate professor they can to lead up some new research area aren't going to be able to do it. I'm already seeing top Ph.D.s no longer be that interested in coming to the United States because they know they're not going to be able to build their research career here.

I was just talking with someone who was accepted into an MIT Ph.D. program in neuroscience, and now she doesn't want to come anymore—

Hmm.

—because of the legal uncertainty that this causes.

Wow. What about the startups that don't have millions of dollars of runway to pay these fees?

Yeah, that's such a good point. They don't have the liquid cash to pay $100,000. This will definitely hurt their ability to expand, and that is hugely important for the future of the tech ecosystem.

Right now, 60% of the top AI startups in the United States are founded by immigrants. That would not be possible without the H-1B program. We're threatening to undermine that huge success story.

Wait, that is an insane statistic.

Yeah.

I did not know that 60% of the top AI startups in America are started by immigrants.

Sixty percent of the top AI startups.

That's incredible. “Immigrants get the job done,” Hamilton said that. Jeremy, what are you hearing from people in the tech industry about how they are planning to deal with this going forward?

Yeah, I'm hearing a lot of different answers. I think some people are just hoping that the legal dispute can happen as quickly as possible because there are a lot of questions about whether this fee is legal. Most of the people I've talked to think that it probably isn't, and so it's a matter of time before it gets struck down. People want to ride this out until that happens.

Others are coming up with strategies to try to deal with this new fee and shift their recruitment toward people who are already here, whom they might not have to pay it for. Others are looking at this like it will just fundamentally change whether or not they will be able to participate in this program at all.

I can imagine some supporters of this $100,000 fee believe that maybe, in some ways, this could help to increase the wages of American workers. Is there any evidence that that's true?

Yeah, for the private sector. Putting aside the fact that we're going to cut out the universities and make it much harder for people to do basic research, in the private sector, the thinking is that only the companies who really value particular people at least $100,000 and are willing to pay that on top of their salary are going to be willing to go forward with using the H-1B system.

But the problem with this is, like we mentioned, that the end result of all of this is likely to be that the outsourcers are going to be able to get around it and continue to bring in the kinds of people who have led to the problems of undercutting Americans. Top talent who might actually be creating jobs for Americans are going to be hit particularly hard. Startups who are going to be creating jobs are going to be hit particularly hard.

The net effect of all of this is that this incredibly huge national asset, which benefits the American worker, is going to take a massive blow, and the people who are undercutting jobs can adapt.

Tell us about the exceptions to this policy. I've heard people talking about the gold card, the platinum card, and then maybe some discretion that the Trump administration can grant to companies and just say, “Oh, you don't have to pay the $100,000 per employee.” So tell us about that.

One thing the proclamation does is create discretionary authority to grant exemptions for the fee at the individual level, at the company level or at the industry level, entirely at the discretion of the administration.

One way to think about this is that it's analogous to moves that the administration has made on trade policy or other areas where they want to be able to negotiate and they want to bring people to the table to make concessions. The promise of an H-1B exemption as the carrot and the stick of a $100,000 fee might be a way to get companies to try to apply for these exemptions and make promises. How all of that plays out—

We'll have to see.

Well, we'll have to see, but it seems like if the Trump administration wanted it to be, this could just be another avenue for influence peddling. “Oh, you did something we don't really like. Well, yeah, we're not going to grant you any exemptions to this onerous new policy we've created. But hey, if you do what we say, maybe suddenly you don't have to pay that $100,000 fee.”

So I think we do have to watch that, actually.

What is the long-term picture here? We have an immigration system that, for whatever problems people have with it, has resulted in a dynamic and very profitable tech industry in America. Many of our largest companies are run by immigrants. What is the long-term threat to American innovation if we start mucking around with this immigration system that has served our tech industry so well?

We run the risk of threatening the talent pipeline that has long been our major advantage in technological and scientific competition. One thing I think about is that the timing on all of this is somewhat ironic because next Wednesday, China is rolling out a new visa program called the K visa for high-skilled STEM talent, particularly aimed at East Asian talent.

Historically, China has not been a big destination for global talent. People haven't wanted to move there. But the central committee has said that they want to be competitive with the United States on global talent by 2035, and they're deliberately trying to play the best hand that they can. The United States, by contrast, seems to be walking away from the field, maybe ceding the biggest advantage we have in technological competition.

Even if these rules get struck down by the courts, it will likely sow a cloud of uncertainty that will deter people into the future. I think there are huge problems with the H-1B program. We've talked about some of them. We've talked about how much the outsourcers are bringing in relatively lower-skilled people, and I think the H-1B program could be a much better and much more highly skilled immigration program.

These particular proposals, though, don't really do that, and they do significantly threaten our ability to continue to benefit from the smartest minds in the world.

Hmm. Well, Kevin, H-1B or H-1 not to be.

That is the question. Jeremy, thanks so much for joining us. This was really helpful.

Thanks a lot.

7. TikTok Predicts the Rapture

Well, Kevin, praise the Lord and pass the smartphone because TikTok is talking about the Rapture.

Oh boy. Is this another segment where you just explain something from the internet that will disturb and puzzle me?

Don’t make it sound like you don’t love that.

No, it is. It is a time-honored Hard Fork tradition, where you dive into the bowels of the internet and bring me back something delightful.

I’ll be honest with you. We’re bringing you what the youth are talking about.

Yes.

Okay? This is not some arcane phenomenon, Kevin. This is what everybody was talking about on TikTok. And now that you’re a dad and you don’t have time to participate in youth culture, I try to reserve some time every week to smarten you up and make it so that you can actually participate in society.

That’s very nice of you.

Thank you. So let’s talk about God. Now, have you read the Bible?

I have, yes.

In fact, you went to Bible college for a year and wrote a book about it.

I did, yes. Back in my youth, I spent a semester undercover at a Christian college and spent a lot of that semester reading the Bible.

And what did you learn about the Rapture in particular?

Well, the Rapture is a foretold event in the Book of Revelation, which is the last book of the Bible—

Mm-hmm.

—where, at least this is the version I heard, one day all the people who are saved, who are born again, just disappear. They go to heaven and vanish. Their clothes and their earthly possessions stay behind, and they ascend. This was written into the plot of a very popular Christian series of novels called Left Behind.

Left Behind, of course.

One day, you’re just sitting there in the coffee shop minding your business, and then poof, all the Christians go to heaven.

Now, does the Bible actually specify that you don’t get to bring your clothes to heaven? Because that is what you just said on this podcast.

No, but that’s kind of the plot of the Left Behind series, that people are just—you know, their cars just drive off the road and their clothes get left behind with them.

Well, I think we can all agree. This would be a very newsworthy event if it happened.

Yes.

And perhaps that explains why, Kevin, over the past week, TikTok has been abuzz as person after person makes a video in anticipation of a Rapture that was supposed to happen this Tuesday last.

This Tuesday last?

Yes. I think that’s a very elegant way of saying it, don’t you? Now—

Ugh.

—on Tuesday, according to many folks on TikTok, it was going to be the Rapture, and we were going to be living in the scenario described by the Left Behind series and also the Bible.

Now, what was their theory of why it was going to happen this Tuesday?

I’m so glad you asked. It all started when a South African man named Joshua Malakela appeared on a podcast back in June, in which he said he had received a DM from old JC himself saying that the end of the world would occur on September 23 or 24. I’m going to quote him because I think you’ll find this interesting. According to this man:

“He,” meaning Jesus, “was telling me that by June 2026, the world is gearing up towards the World Cup. But after the Rapture of September 2025, the chaos that will be in the world, the destruction and the devastation that will occur with the Rapture, there will be no World Cup 2026.”

I know what you’re thinking. I didn’t expect there to be a soccer angle with this story, and I didn’t expect it either. But that is the nature of the Rapture. Now, should we listen to a clip where folks start to discuss this on TikTok?

Well, but—

Oh, yeah.

Yes, but why the specific date, before we do? He said September there, but he’s very specific: September 23. So where did that date come from?

That date came from Jesus Christ.

Okay.

Any questions?

Okay.

All right. So let’s just listen to a clip.

Jesus is coming back on the Feast of Trumpets 2025. Here’s why: When the Rapture does happen—when it does happen in the next 24 to 48 hours, September 23 to 24—it’ll happen in a single moment. It’ll happen because the world is split into a 24-hour time zone. For most of the world, it’ll be on the 23rd. For me, it’ll happen on the 24th. But it’ll happen in a single moment.

We’re going home. I already was convinced, but the Lord just keeps giving little—I don’t even want to call them little. These are huge. He keeps giving confirmations every single day. So get ready, guys. We’re about to see Jesus face to face.

So this starts a trend that becomes so prevalent on the app that it comes to be known as RaptureTok, right? You open up the search bar on TikTok, and all sorts of Rapture questions come up. People get into the act in big ways. People start alleging that they have sold their cars or given them away. One person posted the lock-screen message that she shared. Can we hear the lock-screen-message poster?

So I’ve been doing all this prepping for the people that are left behind, and I want them not to be fooled by what they’re going to be told. We all know that there has to be an explanation for billions of people just vanishing in the middle of the day or the middle of the night, depending on your time zone, I guess. So this is what we came up with. We did something—he did it, I didn’t do any of it. Let me just show you.

So now she shows a lock-screen message that’s basically like, “Hey, in case you were wondering, we’ve all been raptured.”

So the thought is we put this on our home screen, on our lock screen, and then when someone picks up your phone, they’ll see that. It’s just more prep.

Oh, yeah, that’s how they’ll know. Not by—

The fact that billions of people have disappeared.

Yeah.

Yeah, so this goes nuts on TikTok. And while there are a lot of Christians who I think sincerely believed that Tuesday was going to be the big day, they were going to catch a flight to heaven, there were also people who had a lot of fun with it. For example, there was this person who may have hoped that the Rapture was going to result in a bonanza of vintage shopping. If we can play Lil Rowtini.

Obviously, I believe in the Rapture, and the Rapture will be occurring tomorrow when all Christians ascend into heaven, and God bless everybody. But more importantly, we’re all too worried about the things that we’re leaving behind. We’re worried about leaving out scriptures for demons that enter our homes and wonder, “Where did everybody go?” We’re not worried enough about the demonic clothing collections that we’re holding onto. Jesus doesn’t have access to eBay. You need to leave demonic clothing outside by midnight if you want to ascend into heaven. You will not be getting into heaven if you still are in ownership of a fall/winter 1996 Alexander McQueen piece. That is for the devil. It is.

All right. We can stop that one there.

Okay.

Yeah.

So I have a question, Casey. I have many questions.

Yeah.

What is the ratio of people who are sincerely predicting the Rapture would happen on September 23 to the people who are clowning on those people and making jokes about it?

I haven’t done a regression analysis on the TikTok results.

How dare you.

But I would estimate that more people were sort of laughing about the Rapture—

I see.

—than they were believing that it was really going to happen. But then we get past Tuesday, we get well into Wednesday, and no Rapture has taken place.

And so people are beginning to reckon with the fact that, in fact, it’s not the end of the world, Kevin, and we have to keep doing this.

Yeah.

Yeah.

So, some other questions I have. One: Did this person who originally predicted that the rapture would happen on September 23rd post anything since Tuesday?

What’s up?

The person who originally predicted that the rapture would happen on September 23rd—have they posted since Tuesday?

The person who started all of this, Joshua Makela, did appear on a livestream in the middle of the week and was asked, “Hey, it seems like the rapture isn’t happening.” He essentially stuck to his guns and said, “Hey, look, it’s still going to happen.” So that’s basically what we always see in these cases.

Yes.

This is not the first time a false rapture has been predicted. You may remember in 2012, when there was an Aztec prediction that we were about to reach the end of the world. That didn’t happen.

Yes.

Yeah.

I mean, usually what’s so interesting about this to me is that when you have these very specific predictions of the apocalypse or the rapture or what have you, they’re based on elaborate numerology. You know, this ancient prophecy foretold that on the 4th day of the 10th month, Moloch would appear, and we’ve found this buried papyrus script that says it all. This was just literally a guy saying, “Jesus told me the rapture was coming.”

Yes. But interestingly, on TikTok, people did try to apply numerology to the date. People were saying, “Well, that date is actually this ancient holiday known as the Feast of the Trumpet, and this is the year 2025, and the Feast of Trumpets takes place exactly 25 days from this.” But, of course, in the end, it didn’t really work out.

I see.

I’ll tell you my hot take about this, Kevin, which is that rapture talk is just a platform phenomenon, right? Platforms are always looking to amplify the craziest, wildest stories and rumors, and what could be wilder than the idea that billions of people were about to leave the Earth simultaneously, right? And so, through the sheer force of recommendation algorithms, a bunch of people got sent into a tizzy, possibly sold their cars, and once again, we are confronted with the crazy systems that we have built for getting the news.

Yeah.

I’ll throw out an additional hot take on this—

Yeah.

—which is that I think we are experiencing something of a religious revival in this country.

Mm.

Obviously, I’m joking. If you look at the overall trends of church attendance or whatever, we are still rapidly secularizing as a nation. But I am starting to hear more people than usual in and around the tech world talking about God and church and faith. Peter Thiel gave a lecture the other night about the Antichrist, and it just feels like we are seeing people take some of these ideas more seriously. I have no idea whether that’s connected to rapture talk, but it does feel like there is maybe some little counterculture starting to bubble up, where people are getting interested in religion and some of these older religious ideas.

Now, Kevin, of course, one explanation for why the rapture didn’t happen is that this is essentially just mythology, and there is no truth to the idea that one day billions of people are going to ascend into the sky. But one TikTok user did have an alternate theory that I would like to present in the name of airing all views.

Please.

Let’s just air this one. This one comes to us from TikToker a_hardcore_angel.

Okay, guys, rapture update: September 23rd at 4:53 p.m. Eastern Time. Jesus did come back, but was very quickly intercepted by ICE.

So that would obviously be very unfortunate, if true. And that’s what’s going on on TikTok, Kevin. Jesus H. Christ.

I have no notes. Seems like our information economy is working great.

Every day we stray further from God’s light.

It’s true. Now, Casey, in the biblical version of the rapture, there are signs. There are things like wars and rumors of wars.

Mm-hmm.

We’ve had some of those. Natural disasters.

Mm.

We had an earthquake in the Bay Area the other day.

True.

Plagues and pestilence. You were sick last week.

Yeah, I had an ear infection.

So I’m starting to think maybe I belong on rapture talk.

I think, with how elegantly you presented that, you would crush it on rapture talk. Because a lot of these folks—I’ll say it—they’re not putting together quite that level of rationale to make their cases.

All right. Well, I’m going to gather up my vintage clothing and change my lock screen. Get ready to ascend.

See you later, loser.

By the way, great song about the rapture: “Airline to Heaven” by Billy Bragg and Wilco, based on a song by Woody Guthrie. So if I was getting raptured, that’s what I’d be playing on Spotify.

All this is reminding me that when I was at Christian college many years ago, there was this viral prank series that would go big on Christian YouTube, where people would trick their friends into thinking the rapture had happened. Their friends would go to the bathroom in a coffee shop, and while they were in the bathroom, people would lay out clothes on the chairs and leave the coffee shop. Everyone would be in on it, and their friend would come out of the bathroom and see all these clothes lying there, and they would start melting down because the rapture had happened and they were not in on it. So as bad as rapture talk is, I think we may see a revival of even worse kinds of rapture-related internet content.

Mm-hmm.