Hello, and welcome to Decoder. I'm Nilay Patel, editor-in-chief of The Verge, and Decoder is my show about big ideas and other problems.
Today, let's talk about the looming AI monetization cliff, and whether some of the biggest companies in the space can become real profitable businesses before they careen right off it. My guest today is Hayden Field, who's our senior AI reporter here at The Verge, and she's been keeping close tabs on both Anthropic and OpenAI, and how those two companies in particular tell us a whole lot about the AI industry as a whole in 2026.
You've certainly heard a version of the monetization cliff story before. Anthropic, OpenAI, and all the other big AI startups are built off the back of hundreds of billions in capital investment, and they're linked to even greater amounts of forward-looking investment in data center buildout, chips, and other infrastructure spent.
At some point, that return has to pay off. The profits have to materialize, or the bubble pops. Maybe AGI arrives, maybe the economy crashes, who knows? If you've been listening to Decoder, you've heard me talk about this with tons of CEOs right here on the show, and the majority of them have hinted towards the bubble popping. They think some companies will fail in spectacular fashion, and others will succeed, but that the opportunities, and especially the money, are simply too big to ignore.
The AI industry is going to do this, whether we want it to or not. The market depends on it. And so these last few weeks have felt like a very important inflection point as both Anthropic and OpenAI have started to react to the reality of needing to go public to make money.
The catalyst for all this change is the rise of AI agents, products like Claude Code and Co-work, the open-source Open Claw, and OpenAI's Codex. They've all radically changed how these companies are thinking about their resources, and that's starting to affect how they behave. The products they support or suddenly kill, the restrictions they impose on customers, and the money they're willing to burn on the way towards their next big milestone. That's because agents are valuable to customers right now, but agents also use far more compute. And so the way people are using agents is burning tokens at a rate way faster than these companies anticipated.
And that's causing them to make hard decisions. We saw this most evidently last month when OpenAI abruptly killed its video generation app, Sora, ditching a $1 billion Disney deal in the process. Why? Well, it cost too much to run, and OpenAI needs the compute for Codex. And we saw it again just last week when Anthropic decided it would no longer let Claude users burn through compute resources using the Open Claw agent framework through a standard subscription plan, instead forcing those users onto a pay-as-you-go plan, which costs substantially more. As you'll hear Hayden explain, these are glimmers of a make-or-break moment for the AI industry as both Anthropic and OpenAI barrel towards two of the biggest IPOs in history.
And the pressure on these companies to make money has never been this intense. The projections these companies have made, which just this week leaked to The Wall Street Journal, tell a story of mind-boggling growth to the tune of hundreds of billions in revenue and profitability by the end of the decade. But the most important questions now are, can these companies pull all this off? And what compromises will they make to reach that goal and avoid crashing and burning?
Before we start, a quick reminder that you can listen to this episode, or any episode of Decoder, completely ad-free by subscribing to The Verge. Just go to theverge.com/subscribe. Okay, Verge senior AI reporter Hayden Field on the AI monetization cliff and the race to profitability. Here we go.
Hayden Field, our senior AI reporter here at The Verge, welcome back to Decoder. I'm excited to talk to you. There's a lot going on in AI world recently. It feels like we are at a very important inflection point for this industry. What are you thinking right now?
Yeah, we absolutely are. It's kind of like time to pay the piper in a way, you know? They've been raising a ton of money, raising a ton of hype for years, and now, you know, as companies prepare to go public and the competition is heating up more than ever, and they're kind of entering all these different sectors and trying different side quests, it's finally time to really like face the music and see how much money they can really make, and there's never been more pressure on them also.
When you say them, I want to stay focused on OpenAI and Anthropic, which seem to be on different trajectories. Now, there are obviously other big AI companies in the mix. Google exists, it's going to do whatever Google does, but Google has a big business already. It can subsidize finding product-market fit with AI. It can subsidize making efficiency improvements on TPUs. It's just very different from, in particular, OpenAI and Anthropic, which have to become companies. Like they have to graduate into becoming companies, particularly if they're going to go public, and then they're going to have shareholders, they're going to have to show profit and loss and all this other stuff. Can you just describe how OpenAI and Anthropic are currently situated and where they might be going?
Sure. So, yeah, I mean, it's interesting because in some ways they're in the same position. They're both preparing to go public this year, reportedly, and kind of racing each other to do that, and they're both constantly raising a ton of money and hype. But where they differ is, of course, OpenAI has traditionally been really courting the consumer-facing stuff, and some enterprise and government, but they've really been focused equally, if not way more so, on consumer. Anthropic has always been pretty focused on enterprise, and they have remained pretty steady on that focus. So, you know, they're not really doing as many side quests, they're not, you know, rolling out as many other experiments or projects. They're just kind of laser-focused on their enterprise goals and their enterprise clients.
So, now, are they only doing that? No, sometimes they kind of, it seems like get FOMO and they're like going into, you know, Claude for healthcare and, you know, Claude for education, things like that, but it doesn't totally remove them from their goals of enterprise because like, you know, that's pretty focused on healthcare organizations or, you know, education systems are enterprises too, as we know. So, you know, they're really laser-focused on this. They kind of have the reputation of being the adult in the room in some ways because they're not as perceived as like going wherever the wind blows them. They're kind of on one trajectory and they're staying really steady, it seems like.
Whereas OpenAI kind of has the reputation of, you know, changing their focus a bunch, internally and externally. It's like, you know, going on a ton of side quests, you know, trying things, you know, throwing a ton of spaghetti at the wall, like consumer, enterprise, government, everything, just seeing what works. And even Sam Altman himself has described OpenAI as kind of like betting on a ton of startups internally and just kind of seeing which one pulls ahead. But now they're finally having to realize, 'Hey, maybe it's time to focus on the most money-making endeavors here and deprioritize some of these other projects, kill them off so we can just kind of compete with Anthropic and focus on coding and enterprise.'
Yeah, I think that brings me to the news this week that really made me feel like, 'Oh, we're at an inflection point.' And that is Anthropic started raising its rates for people using tools like Open Claw. They really want you in their system, on their subscription plans, using the tools and their pricing their way. And if you want to use Claude to power other systems like Open Claw, you're going to have to pay in a different rate structure that to me feels like they don't want you to do it at all. And then next to that, OpenAI killed Sora, which was their very buzzy video generation product that was basically a deepfake nightmare, but they also had a deal with Disney for a billion dollars, which always seemed confusing, but they canceled that deal, too. Let's start with OpenAI. You're saying they're killing off all these side projects, they're trying to focus on Codex, which is fundamentally enterprise software, right? It is a tool for software developers to make software. Why did they kill Sora, and where did this sense of focus come from?
I think the sense of focus, honestly, just comes from the competition and the fact that pressure is building on them to generate more revenue than ever. They've never had more eyes on them and their balance sheet in their whole company history because they're preparing to go public, and because they just raised so many billions of dollars, their post-money valuation right now is $852 billion. So, yeah, I mean, investors are saying, 'Okay, like what's the plan here? What's the plan for returning our money?' So, in order to deliver on those promises, they are having to not only devote their time and like money and staff to the projects that are going to make the most money, but also their compute.
So, that's something that we saw execs at OpenAI talk about when they killed Sora. We saw a couple internal memos go out. One of them was from Fei-Fei Li, who's the CEO of AGI deployment, and she said that basically the company needed to stop focusing on side quests and just really dive fully into enterprise and coding. And yeah, I mean, compute is super limited. OpenAI is always talking about how they don't have enough compute to fulfill what they want to do or to scale appropriately. Sam Altman at DevDay in SF in October said to reporters, 'I've just never seen him so stressed when he was talking about this' about how the compute constraints were stopping them from scaling appropriately and how they couldn't really deliver what clients wanted unless they just could somehow get their hands on more compute. I've never seen him more stressed out. And so, yeah, that's, you know, playing out now months later.
Sora took up a bunch of compute, and there wasn't really a big return there. And so they abruptly decided to cancel it, apparently 30 minutes after working with Disney on a related project, and then just suddenly pulled the plug with no notice. So, you know, things over there, it seem like are in kind of a tailspin. It's like if you're pulling the plug on a project with a huge company like Disney 30 minutes after talking to them about like how it was going great, there's some real issues there.
I want to come back to OpenAI, its management, which is all but turned over since the last time you were on Decoder, and its sort of strategic focus. But what really strikes me about the need for compute is, I don't know, when you were on the show a year ago or a year and a half ago, all of that compute was pointed at training. Right? We got to make bigger models, they're going to be more capable. GPT-95 will come out and it'll be digital Jesus or like whatever it was. And the idea was that we needed bigger models with more data and the compute, the scale of compute necessary, was going to get us to the capable models and AGI in some way.
Now it seems like the compute's all for inference. Right? There's people that want to do things with these tools particularly in software development. And if we don't scale up the compute to meet the demand, we'll get left out in the cold because our big rival is sitting right there waiting to scoop up all those customers. Has anyone pointed that shift out explicitly that we've gone from all of the focus on training and capability in the model to the models are pretty good, everyone wants to use them, we need massive amounts of compute for inference?
So, in the numbers that leaked from investors in OpenAI and Anthropic this week, you know, there are a ton of crazy charts showing just how much of their money and profit, or lack of profit, is going towards inference, training. The way they break out these categories is just really interesting. It's not always apples to apples, but it's like Anthropic is going to spend 1/3 to 1/4 on model training compared to OpenAI, but OpenAI's revenue is expected to hit like 275 billion in 2030, but Anthropic says its revenue will hit 150 billion in 2029. So, it's kind of like Anthropic seems to be spending a lot less on the same categories as OpenAI and growing slower, but OpenAI is spending a ton and then just hoping that's going to lead to a return on investment, which kind of tracks with the way the companies have been operating for the past couple years.
Yeah, Anthropic seems to be much slower and more focused and constantly worried that it's going to kill everyone in the world with every successive model. OpenAI is just pedal to the metal all the time.
Yeah, but Anthropic isn't worried enough that it'll stop, because they just took out their frontier safety pledge and said, 'Oh, actually, we're going to stay competitive even if we think it's a little dangerous. Sorry.'
The race to an IPO makes a lot of people rethink a lot of their values, apparently.
We need to take a quick break. We'll be right back.
We're back with The Verge's senior AI reporter Hayden Field talking about Anthropic and AI's race to profitability. The other piece of news, right? This is again pricing compute usage. Anthropic changed its pricing structure to make using Claude with Open Claw much more expensive in different ways. So, if you have a Claude Pro or Max subscription, you can't just hook it up to Open Claw and go. You've got to buy tokens on top of that subscription. That in some ways felt inevitable, in other ways obviously made a lot of people very upset. It made the developers of Open Claw upset. They said they could only delay the decision by a week. What's going on there?
So, when I talked to an economist about this this morning, he said that agents have just changed everything and I talked to another couple of tech leaders about this and they said agents are consuming hundreds of thousands more tokens than basic chat models have been. So, it does make sense even if it's frustrating because it seems like, you know, the way Anthropic put it was, 'Hey, you know, our infrastructure isn't built for this. Like we didn't plan for this.' And yeah, I mean, if you're using Open Claw with Claude, obviously like, you know, you basically have an agent on your behalf like prompting Claude for you and delivering those prompts back to Claude and saying, 'No, do this. No, do this.' So, it's like way more than a human would be able to do. And Anthropic's point was, 'Hey, like we only really built this for humans to be able to prompt Claude right now unless it's for our own products.' Obviously, it's a money grab. Like of course they don't want a third-party tool doing the same stuff that they want their own.
Right. They want people using Co-work.
Yeah, of course. So, it's like basically they just want to keep it like a walled garden. They want it to be a moat. That's kind of the only advantage that AI companies have right now is just trying to keep their users engaged and keep them on their own platform, build a moat of some sort because otherwise like, you know, everyone loves to switch between whatever model is the best that day that week. So, yeah, this is just Anthropic looking to like deepen its moat and also, you know, compute is constrained everywhere. So, they don't really want a third-party tool prompting Claude way way way way way more than a human would be able to.
They don't seem to mind when it's Co-work doing the prompting because you're inside of Claude and then they can monetize you in whatever ways that they come up with to monetize you.
Exactly. They want to monetize it if they can.
Is there a path either in enterprise for Anthropic or consumer for OpenAI to actually make a dollar in profit? That seems to be the big question everybody has.
Yeah, that's a really good question and it's something that when I was chatting with economists, they didn't feel that it would be that easy or likely, but they think that one or two LLM providers will come out on top and the rest will have to consolidate. So, yeah, there's a chance for sure. Anthropic and OpenAI are both projecting some form of profitability in 2029, 2030. Anthropic said maybe this year it'll also like be slightly in the green and then go back in the red and then go back in the green for 2028, 2029, but yeah, I mean, I think that they've all realized that if that is ever going to happen, it's going to be via the boring, unglamorous, back of office stuff, enterprise, military contracts, government contracts, all that stuff because consumers just honestly, yeah, they'll maybe pay for a $200 a month subscription if they're a power user, but there is no way that stuff is ever going to add up to the amount of money that's involved in these enterprise or government contracts.
