Transcript: Is OpenAI’s business model sustainable?

This is an audio transcript of the Behind the Money podcast episode: ‘Is OpenAI’s business model sustainable?’

[MUSIC PLAYING]

Michela Tindera
Silicon Valley’s hottest start-up — I’m talking about OpenAI — recently hit a milestone. In December, the company surpassed $2bn in annualised revenue, which means that if OpenAI earned as much money each month as it did in December, $2bn is what it would make over the course of a year. Now, this is a big deal for a company as young as OpenAI. The FT broke this news earlier this month, and George Hammond was one of the reporters on the story. 

George Hammond
OpenAI are on this very steep curve of growth. So they are talking about potentially doubling or more than doubling that revenue number over the next 12 months. So it’s an incredible rate of growth that they’re going at. 

Michela Tindera
A huge part of that lightning speed success is ChatGPT, the AI chatbot that OpenAI launched a little more than a year ago. 

George Hammond
$2bn in revenue puts OpenAI amongst the most successful companies ever to come out of Silicon Valley, in the likes of Google, Apple. 

Michela Tindera
Now, businesses want to use OpenAI’s generative AI technology to make their workplaces more efficient. But the CEO, Sam Altman, says that he has much bigger goals. 

George Hammond
I think there is no end to OpenAI’s ambition. Sam Altman, in particular, is looking to solve this problem of superintelligence. He’s looking to set up his own semiconductor manufacturing pipeline. He’s trying to lengthen human lifespan. The guy is extremely ambitious. His company is extremely ambitious, and that is part of their sell. 

Michela Tindera
But accomplishing things like creating superintelligence, well, that’s gonna cost you. 

George Hammond
I think that’s the main tension for OpenAI. They have what is a very grand mission at the centre of their business, which is gonna cost them huge amounts of money to pursue. And it could be that the pursuit of that mission is at the expense of them just creating, you know, quite neat but less ambitious tools for enterprises. There is a question for them about which track they want to pursue. There’s a question for investors about which track is likely to be more profitable. And it’s gonna be an interesting tension to watch play out for OpenAI, as they get increasingly sophisticated and they require increasing amounts of money. 

[MUSIC PLAYING]

Michela Tindera
I’m Michela Tindera from the Financial Times. Today on Behind the Money, OpenAI is one of the fastest-growing companies ever. But can the business model it has now sustain its massive long-term goals? I’m talking with venture capital correspondent George Hammond and the FT’s AI editor Madhumita Murgia about the challenges that lie ahead.

Hi, Madhu, George. welcome to the show. 

Madhumita Murgia
Thank you. Nice to be here. 

George Hammond
Thank you for having me. 

Michela Tindera
So hearing all these crazy growth figures, I mean, some people might forget that OpenAI got its start as not a commercial endeavour, but actually as a non-profit. Tell me more about that. 

George Hammond
So OpenAI is founded in 2015 by this very talented group of young researchers and engineers and venture capitalists, including Sam Altman. And it starts as a non-profit research lab with this very grand mission to create artificial general intelligence that benefits all of humanity. So they want this to be a very pure, focused research effort. 

Michela Tindera
And just to clarify, what exactly are we talking about here when we say artificial general intelligence benefiting all of humanity? 

Madhumita Murgia
AGI or artificial general intelligence has become this vague, amorphous term for something that’s superintelligent or more intelligent than humans. But really, the way that those in the industry kind of describe it, what they’re aiming for is a set of software, computer software that can make generalised decisions and understand data in a way similar to humans. So, yeah, the kind of goal when they were founded, it was born from a fear of, you know, superintelligent AI that could be turned rogue and do bad things. And they kind of funded this company as a non-profit to make sure that beneficial AI was created with guardrails in place and risks could be managed. 

Michela Tindera
Wow. So it went from non-profit research lab to a commercial venture. What happened? 

George Hammond
They quite quickly realised that it’s gonna be a prodigiously expensive task to develop increasingly intelligent artificial intelligence. And to remain ahead of the competition, it’s very demanding in terms of the amount of computing power you need and there are other resources they need to go out and buy. So they switched to a for-profit or a capped-profit model in 2019. And that basically allows them to start raising capital from more traditional investors who take a stake in the company on the basis that they will get a portion of the company’s profits. 

Michela Tindera
Right. And one of their biggest investors becomes Microsoft, which goes and puts about $13bn into the company. So what’s OpenAI’s business model today? I mean, of course, no AGI yet. And they’re banking on these enterprise customers to lift their revenues in the coming year. But more specifically, what are they selling? 

Madhumita Murgia
So really, when they think they became a revenue-generating company is really post-ChatGPT, which launched in November 2022. Since then they’ve sold premium access to ChatGPT. So this is like a freemium model where you can use it for free up to a certain point, or you can pay to have a premium version that has extra features where you can use it more frequently, et cetera. 

Michela Tindera
Right. And of course, the uptake of ChatGPT has just been totally remarkable. Now those ChatGPT users are made up of the free users in addition to paying consumer and enterprise accounts. But how else is the company generating revenue? 

Madhumita Murgia
They also sell underlying access to the model that powers the chatbot, which is called GPT-4. And this is for companies who have software developers that can just directly plug in to the model and build whatever they want on top of it. So you could have a management consulting firm or a bank who wants to build their own very specific product, you know, looking through their past reports or interviews in order to generate new sort of solutions to specific business problems. So it’s very tailor-made for them rather than a more generic ChatGPT. 

[MUSIC PLAYING]

Michela Tindera
So to recap, the main areas that OpenAI is generating revenue from include both ChatGPT and the model that underlies ChatGPT that’s known as GPT-4. But the company isn’t profitable. In fact, it’s far from it. And that’s because it has these huge operational and research and development costs. The first is continuing to develop the generative AI technology that underpins its chatbots. That’s like GPT-4 that Madhu mentioned. At the moment, they’re developing the next version, GPT-5. And doing that is expensive. 

George Hammond
It requires a huge amount of computing power to train these models, the ChatGPT models, and it requires slightly less space than a considerable amount of computing power for them to do inferencing, which is basically to service requests that are run through them. And that costs a lot of money. It’s what the Microsoft partnership with OpenAI is all about, is providing some of that computing power. 

Michela Tindera
And then there’s the second thing that’s even bigger and even more costly. It’s its goal that hearkens back to the reason that OpenAI says it was founded in the first place, and that’s to develop artificial general intelligence or AGI. Pursuing this could cost billions, if not trillions of dollars more. But this also puts OpenAI at a crossroads. 

George Hammond
There is this very grand goal that OpenAI has to create artificial general intelligence, and that is partly what’s given us such sophisticated tools. And for consumers, that’s great. For an enterprise, it’s a little bit more complicated. And this is kind of the central tension for OpenAI. They want to create superintelligent tools. And it might be that the companies they’re selling into just want a bit of software that basically furthers their bottom line. And in the words of one investor who we spoke to, you know, he was like, not everyone needs a Ferrari. These enterprises don’t care about an all-knowing, all-seeing entity. They care about making money from this tool. 

Michela Tindera
After the break, we’re looking at the pressures that OpenAI faces and if they can be overcome in the long term. 

[LIFE AND ART FROM FT WEEKEND PODCAST TRAILER PLAYING]

Michela Tindera
So OpenAI’s at this crossroads. Madhu tells me she interviewed a professor named Ethan Mollick from the University of Pennsylvania’s Wharton School, who told her as much too. 

Madhumita Murgia
And his point was, there’s this kind of tension where, you know, they could easily build a profitable business, stop putting so much money into R&D and just improve their primary product and kind of make loads of money from it. Or they could continue to, you know, aim for creating this superintelligent or AGI system for which they’re gonna need billions, you know, if not trillion dollars. And so that’s kind of the evolution that the company is undergoing. 

Michela Tindera
For what it’s worth, Madhu also tells me she’s sceptical about how much creating AGI is really a part of OpenAI’s goals anymore. 

Madhumita Murgia
For me, it’s not quite clear whether the AGI thing is a little bit of like marketing speak or you know how they want their company to be perceived. Sam still talks a lot about wanting to do that, but everything they’re doing in reality doesn’t really support that they’re just focused on creating sort of some beneficial superintelligence. They clearly do care about building a commercial, you know, set of products, making money just like, you know, any other sort of Silicon Valley tech company. 

Michela Tindera
Still, whether it’s for AGI or continuing to develop their GPT models, the company has to figure out a way to keep funding itself. So what’s getting in the way of that? First, George tells me that in his and Madhu’s reporting, they’ve talked to some executives who’ve started using OpenAI’s tools in their businesses, and some have been really enthusiastic. 

George Hammond
Some have found very obvious applications for it. And one person we spoke to was DoNotPay founder Josh Browder. His company provides online legal services using a chatbot. That chatbot is built with the help of OpenAI’s tools. It uses the conversational abilities of ChatGPT to go and communicate with his customers. It’s very, very central to his business, and he says those tools have been transformative. 

Michela Tindera
But that hasn’t been the experience across the board yet. 

George Hammond
From our reporting, it was pretty clear that businesses are still figuring out how to use this technology. There is this figure that has been trotted out that OpenAI’s tools are used by more than 90 per cent of Fortune 500 companies. But from our reporting, it would appear that often the use case is fairly peripheral to the business, and it may become more central. But at the moment there are a lot of chief executives who are kind of trying this before really, really committing to it in use across the organisation. 

Michela Tindera
Madhu says that it’s important that OpenAI does figure this out, though. 

Madhumita Murgia
So they have to hope that people do figure out why this is so useful and that businesses do see a significant impact on their bottom line or new business models or new products that they can build on top of these models. And I think they’re gonna have to bet on the fact that it’s not just them selling these products. They have to create a wide ecosystem, a sort of whole industry around this. The people who build on their products should make money as well, and only then will this kind of start to pay for itself. 

Michela Tindera
So businesses are using OpenAI, but they don’t see it as vital to their operations yet, which is a problem. Another challenge that OpenAI is facing is related to raising funding from investors, as the company angles for many more millions and billions in funding. The usual pool of investors for tech start-ups is dwindling. 

George Hammond
OpenAI, so they have this partnership with Microsoft. That’s their cornerstone investor. Then there’s the question of how much more does Microsoft want to plough into OpenAI, either in terms of cash or computing power, but that will be central to its ability to develop its next generation of tools. But they’ve really run through traditional venture funding unbelievably quickly, because venture investors take bets on early-stage companies where they might make a return. If all goes well of 10 to 100 times their capital invested. Very hard to make 10 to 100 times the capital invested on a company that’s worth almost 100bn. You need an enormous outcome for that to be the case. And so traditional venture investors are priced out of this, by and large. 

Michela Tindera
I mean, George, couldn’t another fundraising option be for OpenAI to try and go to the public markets? 

George Hammond
That’s a tough one. So OpenAI could go public. But this is a company that is not making a profit. And that might be a challenge to any attempt to go public. At the moment, as a private company, if it wants to fundraise, it has to disclose various financial details, but it can operate with a degree of secrecy, which might be quite helpful. And if it can continue to tap resources from Microsoft and others, then I suspect it will stay private. But yeah, going public is an option at some point. 

Michela Tindera
Yeah. So what’s OpenAI doing instead in the meantime? 

George Hammond
So OpenAI is having to look up the chain at wealthier and wealthier investors. And this is really kind of state-level actors at this point, sovereign wealth funds in the Middle East or the Far East. And there’s really a vanishingly few number of investors who can make a serious contribution to OpenAI. 

Michela Tindera
Now, the other risk to all this, of course, is that OpenAI isn’t operating in a vacuum here. So what’s the competitive landscape looking like? 

George Hammond
So if OpenAI can cut its costs and raise the money required to train its new models, then obviously it risks losing ground to its rivals and those at some start-ups. So Anthropic is the biggest of those, but also the big tech companies, including Microsoft, which is developing its own suite of tools, and obviously Google, which is trying very hard to catch up. 

Michela Tindera
So what do you think you’ll be watching for as the story continues to play out? 

Madhumita Murgia
So I think it really depends on what happens over the next two years in how AI is implemented outside of OpenAI in the rest of the economy. And yeah, maybe they will, you know, have a period where they run out of money briefly or, you know, they show such value with their next model that people continue to believe and back this. So I think this is a key year for them. 

Michela Tindera
Well, thank you both for being here. 

Madhumita Murgia
No, thank you for having me. 

George Hammond
My pleasure. Thank you for having me. 

[MUSIC PLAYING]

Michela Tindera
Thanks for listening. Don’t forget, we’re still collecting submissions of questions you have about markets, finance or economics. If you send in a question, we might answer it on a joint show we’ll be putting together with the FT’s markets podcast, Unhedged. There’s details on how to submit a question in our show notes.

Behind the Money is hosted by me, Michela Tindera. Saffeya Ahmed is our producer. Topher Forhecz and Manuela Saragosa are our executive producers. Sound design and mixing by Breen Turner. Special thanks to Marc Filippino. Cheryl Brumley is the global head of audio. See you next week.