Earlier this year, Anthropic edged to the front of the AI race thanks to its focus on business and coding products. But now, its chief rival OpenAI is starting to catch up. WSJ finance and technology reporter Angel Au-Yeung joins to explain how the battle for enterprise customers is playing out, and how it might affect the startups efforts to go public. Danny Lewis hosts.
Sign up for the WSJ's free Technology newsletter.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
[00:00:04] Welcome to Tech News Briefing. It's Friday, October 9th. I'm Danny Lewis for The Wall Street Journal. Anthropic edged to the front of the AI race earlier this year thanks to its focus on business products. Its cutting-edge models and clawed code tool rapidly became favorites in corporate America. But lately, Anthropic's biggest rival, OpenAI, is gaining ground. WSJ finance and technology reporter Angel Aoyoung is here to break down the battle for business customers and why it matters as the startups look to go public.
[00:00:33] Angel, Anthropic was so popular with enterprise customers this spring that clawed code frequently went down and it was worried about running out of computing power. What changed to give OpenAI a chance to get back on top? There are a number of factors. I mean, the pace of this industry is so crazy that all the changes that have happened in the last couple months, it really feels like it could have taken years. First, OpenAI, over the summer, came out with the GPT 5.6 models.
[00:01:00] And these models were really interesting because they were advertised as being not necessarily the most intelligent models, but that just made them a lot cheaper to run. And the release happened at a time when corporate America sort of woke up to this idea that maybe they didn't need the most intelligent models to do the majority of their tasks. A lot of companies were also starting to complain about the high AI bills that they were getting.
[00:01:30] Also over the summer, Anthropic came out with Fable, which was its most frontier model at the time. And this model, the company said, was so powerful that they were going to change their data policy. And the main change was that they basically told customers that it was possible for them to store data from customers for up to 30 days.
[00:01:53] And there are a number of customers that are in industries that handle sensitive data, healthcare, financial services, that just didn't like this change. It was really those two main incidences that contributed to OpenAI gaining ground on Anthropic. Yeah, what does this say about how enterprise customers are changing their approach to the AI tools that they're using?
[00:02:16] At the start of the year, the data that we've seen shows that Anthropic had a very strong lead among enterprise customers. But as the year progressed, companies started model mixing more, which means combining different models for tasks for two reasons. The first is they want to bring costs down. And a way to do that is maybe you use Anthropic's most intelligent model for a very complicated task.
[00:02:44] But for something that is more administrative, maybe you use a lower intelligence model that is cheaper. And it could be one from OpenAI. It could be an older model from Anthropic. It could also be an open weight model, many of which are coming from China. And the second change in how companies are approaching AI models is there seems to be a desire to not be 100% reliant on one model maker,
[00:03:09] especially when these model companies have shown in the last year that they are also launching products and apps themselves, many of which are the bread and butter of some of these startups. And so to just gain a little more independence from the AI giants, you see a lot of these companies just wanting to experiment and, again, just not be completely dependent on one company.
[00:03:32] It is kind of a funny thing as these AI companies are encroaching on the enterprise software world because, I mean, traditionally, companies are locked into a vendor's ecosystem for whatever the length of the contract is. Why isn't that the case with AI models? It's very easy to model switch. It's not like a cloud contract where it's more difficult just from like an infrastructure perspective to switch cloud providers.
[00:03:57] With models, there is a surprising amount of power from the customers where they can choose whatever models they want for a variety of tasks. And so it definitely makes it more competitive and dynamic for the AI giants, right? Because, yes, companies are signing contracts with the AI giants, but these contracts don't preclude them from signing contracts with other AI companies. And it's also very easy to switch from one model to another.
[00:04:26] Because you're in this moment in time where all the AI model makers are just trying to get market share, they are subsidizing pricing just so they can get the customers. It's a good time to be a customer right now. Could we start seeing some of these big AI companies lock that down a little bit more and be like, nope, you've got a contract with us. You've got to use our models. Possibly, but not right now. It still feels like early days in the AI industry.
[00:04:54] And so many analysts tell me that when it comes to AI adoption in the American economy, we're just at the beginning. So there's still a lot more customers for OpenAI, Anthropic, Google to get. And right now is probably not the time to have restrictive contracts. Right now they just want to grow. They just want to capture customers.
[00:05:19] So we could see that in the near future, but I don't think that this is the time for them to do that. If you're an AI user, what tasks do you use different models for? If you're a listener on Spotify, let us know in the comments. Coming up, how might Anthropic and OpenAI's efforts to win over customers affect their efforts to go public? That's after the break.
[00:05:53] Welcome back. We've been talking about how OpenAI and Anthropics pushes for enterprise customers are affecting the AI race. But Angel, both companies are eyeing going public in the not-too-distant future. How might this back-and-forth for first place impact their respective IPOs? Right now is such a crucial time for both companies. Both Anthropic and OpenAI are on the precipice of planned IPOs. We've had reporting showing that Anthropic is likely to go public this November.
[00:06:22] OpenAI is likely to go public next year. And so now is the time for both businesses to show Wall Street and show investors that they have sustainable businesses. For the last year and a half, there's been a lot of attention on both companies' skyrocketing capital expenditures. They're spending a ton of money on infrastructure. They're spending a ton of money on chips. Now is the time for them to show that, yes, the demand is there. They've got the infrastructure.
[00:06:50] But now they need to show that that is actually a potentially profitable business. Over the summer, you had OpenAI slash prices for access to their models. And while a pricing war is great for customers and customer capture, it's not necessarily great for two companies that are just on the eve of their planned IPOs. Because when you slash prices, there's obviously potentially going to be less revenue coming in. They want to get more customers. They're trying to go IPO.
[00:07:19] They can't make prices too expensive. They also have a lot of debt on their balance sheets. There's definitely a lot of pressure on both of these companies. So what do Anthropic and OpenAI need to do to prove to investors that betting on them in an IPO will pay off? They need to show that they have a good number of customers who are willing to pay, that their products are sticky, and that customers will continually pay for their products for a foreseeable amount of time.
[00:07:45] And that's tricky for this industry because the models themselves are starting to become commodities. About a year ago, they were releasing models probably every three to six months. Now it's one to two months, which is causing customers to just switch more often from model to model. So it's not an easy task what they have right now to do, but we will see who wins first. That was WSJ Finance and technology reporter Angel Au-Young. And that's it for Tech News Briefing.
[00:08:15] If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by me, Dani Lewis. Jessica Fenton and Michael LaValle wrote our theme music. Our supervising producer is Katie Ferguson. Our development producer is Aisha Al-Muslim. Chris Zinsley is the deputy editor. Lital Mollad is our senior director of shows. And Samantha Hennig is the Wall Street Journal's head of multimedia. We'll be back Monday morning with a new episode. Thanks for listening.