This post was originally published on this site.
0:00 spk_0
Welcome to the Opening Bit Unfiltered podcast. I’m Yahoo Finance executive editor Brian Sazi. There’s been one company at the very center of the AI trade this year that is none other than OpenAI. OpenAI has won Yahoo Finance’s 11th annual Company of the Year award. Here’s why. First, it undeniably led the market narrative for investors in 2025.When it announced big deals, stocks popped. When it said something cautionary, Stocks got hit. Second, OpenAI secured scores of new transformational deals with large public companies such as Nvidia, AMD, and Oracle. And 3, it’s your deal making will likely see the company and the year as the world’s most valuable private business with.Estimated evaluation of $500 billion according to Yahoo Finance data. I sat down exclusively with OpenAI CFO Sarah Fryer to go inside the company’s next chapter.
0:51 spk_1
Sarah,always good to see you. Thank you for doing this. Congratulations on winning this award. Uh, I know it’s busy times for you, which we will get into, so thank you for doing this. Um, what has your year been like as CFO of OpenAI? Can you put it into one word?
1:09 spk_2
It has been busy.But thank you, Brian. So first of all, we’re energized to get a word like this. It, you know, really cements, I think what chat GPT has become, which is the world’s default intelligence layer. On top of that, compute has become the foundation of this next era of AI. And I think thirdly, it’s been a year where we have really hit.Kind of this day one of the next phase for OpenAI. We became a public benefit corp. We’ve raised a lot of capital with a really strong balance sheet, and we’ve seen, as you said, a lot of the ecosystem really start to move towards us to help us create thisfuture.
1:51 spk_1
Over the past, let’s say 10 years of my career, Sarah, I’ve watched.Your career from afar, you and I have, have talked, um, prior in your prior roles as a CEO of Nextdoor. You’ve grown businesses and worked with some really transformational leaders. I think about the business you helped scale at Square, help, uh, Nextdoor grow its business. Now you’re at OpenAI. How are each of these situations different and, and what similarities do you see?
2:14 spk_2
Yeah, so I think first and foremost, I do love growth. Um, growth is a moment where you get to really see impact to your customers. And probably the big thing for me this year has been the shift from not just 800 million weekly actives on our consumer platform, but we’ve gone from kind of single product on a single platform.Um, doing a single thing, uh, in terms of the modality of the, of the actual model itself, um, and also singular revenue streams. And what I’ve seen at those other companies, but now I’m seeing, you know, just at a much faster pace at OpenAI is the model itself has gone to multimodal, so you can now engage with voice, with vision, with auditory.Uh, we’ve seen our revenue stream go from a very simple single subscription, $20 a month, to being able to buy a high-end subscription for people who are power users on products like Codex, to being able to offer chat GPT into areas of the world where $20 a month is price.Prohibitive, and we’ve started to offer credits. We have, of course, now offered chat GPT in many different ways. So you can get it as an app on your phone, you can get it on your desktop, you can get it through our Atlas browser, and of course you can start to download our Sora app. So a whole other way to experience AI Nvidia.In video. And then finally we are multi-cloud. So in terms of risk mitigation on the back end with compute, we now have many, many more suppliers and partners in our ecosystem, and that helps risk mitigate, but it also introduces a little bit more healthy competition, and we think positions us to really build something incredible as we head into 2026. You’re,
4:02 spk_1
you’reno stranger to looking at financial statements, sir, to say the very least. How do you know?With the platform growing as fast as OpenAI, how do you know where to allocate capital?
4:13 spk_2
Yeah, so that is probably one of the the most important jobs, I think of a CFO. It’s also one of the most complex, and of course it changes, right? You just asked what it’s like to be in a growth company. A lot of what changes is you get new information, not just every 90 days, but as often as maybe every week, and you have to be able to stay nimble and flexible, but also have a through line so that youare always investing both for the really long term, as well as making short term tactical moves. So I’ll give you some examples of that. Sometimes it’s a new product or a new feature just hits and hits big. So in 25, we said this would be the year of agents, and that’s absolutely what we’ve seen begin to happen. But there was that viral moment with ImageGen back in kind of the May time.Frame where we just saw our app go exponential on exponential. And that was a moment where we had to pivot and say, OK, now we need more compute to put against this, and we’re really going to double down. Clearly on the creative, creative side, consumers are finding real value here. So that’s a moment where you make a tactical shift. Um, another good example has been our shift towards more enterprise investment.So when we entered this year, we looked at our revenue and it was about 70% consumer, 30% enterprise, and I think actually we did an interview where I said it’s hard for the enterprise to catch up because the consumer is growing so fast. But we said that doesn’t matter. Enterprises need to be able to get access to intelligence, so we’re going to go.Really hard after enterprise customers and get them there. And so we’ll actually exit this year with our revenue now tipped 60 to 40. So that’s a big, big uplift on the enterprise side. So that was a decision too about capital allocation towards the enterprise balancing with consumer.And then I think the hardest decision is actually around these really longer term duration compute contracts because of course the compute that we’ve used in 2025, we didn’t just find in 25, we actually had to make a decision probably in 2023.And what we have seen is that as we invest and compute, our revenue grows almost at the same pace. So if you look in 2023, the total compute we exited that year was only a mere 200 megawatts. We’ll exit this year at 2 gigawatts, so 10x. But our revenue, similarly, or ARR went from about 2 billion exit in 2023 to almost 20 billion now exiting 2025, actually a little bit more than 20 billion.So that decisioning about what compute we will need, not even in 26 or 27, but actually in 28, 29, 30, those are the sorts of long term capital allocation decisions we’re having to make right now as well. And so it’s a, it’s a very interesting balance, and I will tell you I use our own product for a lot of that too. Chat GPT often helps me make tough decisions or at least shows me what might be the trade-offs.
7:17 spk_1
Hold on, you’re, wait, wait, hold on, you’re forecasting your business using your own product.
7:21 spk_2
In finance, we have been massive adopters of chat GPT. Um, it’s something I talked to CFOs about all the time.It can start in really small ways. Like, for example, we have agents that help us with onboarding suppliers in our procurement supply chain. We have agents that help us look at our monthly close to say, you know, what is this flux? Like why did this happen? We use connectors to actually connect back into other systems like Slack or into our Google Docs so that we can actually explain why did performance marketing double or triple in a particular period.And then often I will use it when I’m in the midst of like a tough problem, like during planning, for example, to say, OK, this team just gave me this planning memo. It looks excellent. But knowing what you know about the over the other areas where we’re investing, what might be the best question for me to ask right now or what might be the trade off that I haven’t considered.And it is fascinating and fabulous to feel accelerated and amplified by intelligence, right? Remember chat GPT is like having a PhD in math, physics, chemistry, biology, sitting right beside you, helping you with your work.
8:35 spk_1
I wish this is one of those moments where I had like, I always wish this was one of those 6-hour podcasts, but you and I don’t have that time, especially you. Um, all right, let me just, uh, let’s get some numbers out of the way. What, how many users are on the platform? Uh, it sounds like you just told me you’ll hit what, 20 billion in sales this year. Um, what does that look like? What is chat GPT telling you it might look like next year? Does it, does it double? And, and how many people are paying for the platform?
8:59 spk_2
Yeah, so today we have 800 million consumer users, so weekly active users of the product. On the enterprise side, we have 1 million enterprises using the product. And remember, our enterprise business is only a couple of years old. It’s kind of mind blowing to me. From a developer standpoint, about 4 million developers are developing on the platform.And in terms of growth, uh, we just put out a report. Our amazing econ team put out a report yesterday called the State of Enterprise AI adoption. It has a ton of fascinating facts and figures about how much adoption is going up. One thing I would call out to your audience is what we see is frontier users in the enterprise are using messages about 6X more than even just the.And so I think one thing we all need to be careful of, particularly CEOs and executive teams listening in, is that that widening gap, you can’t let it get too wide because you’re going to get behind your competition. And our job is to help you figure out how to deploy it appropriately. As we go into next year, you mentioned that 20 billion, it’s 20 billion of ARR. This year we’ll actually do about $13 billion of in period gap revenue.And our growth rate just continues to, you know, climb. So I’m not giving forward forecasts, but to say we’re very upbeat, very optimistic about what the business looks like in 2026.
10:28 spk_1
You have been in a period of, uh, what your CEO Sam Altman has called Code Red. Now, full disclosure, before he came out with that memo, I never heard of Code Red or any of those like moments or calling it that moment. Uh, during this moment, what is it like inside of the company?
10:46 spk_2
Yeah, so first of all, I’ll tell you I actually personally love a code red moment. It’s, it is a bit of a technic, a tech term like I’ve seen it used across almost every company I’ve ever worked in. You can’t do it too frequently because it exhausts people and you start to get a little bit the, you know, the boy who cried wolf sort of thing can kick in where people just hear it and then they go about their business.Code Red is really just saying to the company, the main thing needs to be the main thing. And so if you are working on many things, but you’re being told you have to get this done, in particular, a big focus on chat GPT and how we show up for our users, and it might be consumers all the way through into enterprise users, you need to prioritize.That first and foremost, it starts with our models and making sure that our models are on a faster kind of speed of velocity of iteration. As I said earlier, we do put new updates into the model all the time. So even though we might not name it a new model, the Cha GBT, the current, is 5.1. We may have a new model coming soon.Um, but that 5.1 is not the same 5.1 that you got a couple of weeks back. But we want to speed up that velocity and really take what we’re learning both in pre-training, post-training, and then even, um, during consumer usage, there’s a moment where if we offer more compute, and this goes back to why we need more compute.Uh, we can actually create a better experience and more learning about what is the highest, best answer we can give a user. So code red is about focus. Main thing is the main thing, and I think it’s really good for companies. Um, sometimes we don’t always call it, by the way, a code red. Like this time last year we were doing the 12 days of shipments. The feeling was equally intense, just sounded better. Maybe our branding was better back then, but I think it’s about bringing intense.And by the way, I think it’s only something that startups do well. Like big companies find code reds hard because how do you tell, you know, 100,000 people they all need to work on something with intensity. It just depletes itself as you kind of ripple out versus in a company at our scale, truly everyone can know this is the most important thing to get done between here and, you know, we usually have exit criteria.And again, we don’t want it to go on for too long because we also want people to have more rhythm and not get burned out. But opening it is an intense place because we have a big mission and we need to do it as fast as possible.
13:18 spk_1
When doyou anticipate being out of code red and what are those triggers?
13:23 spk_2
Yeah, so as I said, there’s always a very well, clearly defined exit criteria when you go into a code red. Like when I’ve seen them not work in prior lives, different companies, it’s usually because the exit criteria was either too nebulous, like no one really agreed on it as you set out to do the code red, or it was almost like so aspirational that it became like, you know, a year plus in the making. That makes no sense. I think what we’ve done well here is you have an exit criteria.Really around, um, what, what is the feel of, of the next kind of model for chat GPT? How’s it going to present to consumers? How do we think about balancing things like personality, use case, and so on. What I get excited about in chat GPT right in in where we’re headed into 26 is a couple of primitives. So number one is that we’re making sure that you’re making more use of personalization. So that memory makes chat GPT work for me, Sarah.Brian, um, but not generic. I think the second thing we’re really focused on is how do we become a little bit less insular. So it’s not just chat GPT by itself, but how do we offer more apps within Chat GPT. So for example, if I want to book travel, I can now call upon Expedia Booking.com to book travel inside of Chat GPT. If I want to do more creative design work, I can call on Canva, and you’ll see us do that with more companies. And then conversely,If we’re using an app and you want to pull that up in your Atlas browser, you can now pull on intelligence in the sidebar to help work with the app that you’re in. So that’s a way to go from being more insular to much more of a platform.And then the final piece, which I think probably resonates most for the consumer in a way, is Cha GBT becoming that super agent. So it might be my concierge doctor, it might be the tutor that I use with my kids or even for me, so I can learn more, or it might be my travel agent or my stylist. You can see it working today in study mode where chat GPT doesn’t just manifest by giving you the answers. It actually works in this more Socratic way as a teacher, um, and I love that.It keeps testing as it goes to see if you’re learning. And that might be because you’re a university student. I’ve got one of them right now, two of them right now doing finals, um, or it might be as a later in life learner, for example, that’s decided to switch careers. Um, but that’s that super agent that I think we’ll see a lot more of in 2026.
15:53 spk_1
Clearly, uh, you’re, you’re putting a lot of capital, a lot of focus during this code run moment on chat GPT. Does that meanWe will not, and you and I talked about this, I think at the Goldman conference. Does that mean we will not see that hardware device from an open AI over the next year? Like, is that pushed further out?
16:12 spk_2
Brian, I told you you’re not allowed to ask me about the hardware device. Well,
16:15 spk_1
I didn’t ask you what it might or may not be. I just asked if it is coming, if it is coming.
16:21 spk_2
Uh, no, the hardware is still very much coming. Um, hardware has a longer.gestation period is usually how I think about it, right? It was one of my learnings at Square. Software, um, you know, has this very fast paced, you know, literally you can turn out software in days, weeks, months, and in fact now with our Codex product you can really do it in hours and it’ll do it autonomously for you. Hardware, you have to make decisions over quarters, years.Even and so that device is still very much on track, but will be, you know, a year plus in the making. So no changes right now, no announcements to be made, or Johnny Ive will come and get me. But we are quite excited about what that substrate can mean for how people experience AI. Like I think as we talked about at Goldman.The era of the internet and then mobile really had us think about, you know, a flat device where we kind of talk with our thumbs. Um, AI is multimodal. We can experience it in this much more human intuitive way. We can talk to it, we can listen, we can, it can see.Um, we just don’t do that enough. It’s not yet in our vernacular as, as individuals, and I’m pretty excited for what a new hardware substrate can bring. Fair
17:39 spk_1
enough. I wanna, I wanna see if you can, uh, help me settle some of these debates because I get these questions. I’m like, all right, well, I’m gonna put these to Sarah and very simply see what she has to say. Does the next update on chat GPT put you ahead of Gemini 3?
17:52 spk_2
You know, I think this model kind of chicken egg, who’s ahead or some chicken eggs, not quite the right phrase, but like who’s ahead in some ways I think it’s just going to keep happening. Look, our models are incredible. Um, I am pretty excited for what the reasoning part of this next model is going to bring, particularly in an enterprise setting, actually.Um, I think actually one of the biggest unlocks is less about who’s got the best model and the most intelligence, but how do we actually help people use it in a much more intelligent way, right? Our model, I already said it’s like PhD level in everything, and yet people still largely use it for things like what’s the weather in San Francisco today, or, you know, important things in life, but like what could I buy my 18-year-old son.Um, it’s not quite what I’d probably ask, you know, someone who sat down next to me and is a physics expert in black hole theory, and yet that’s kind of what we’re using it for.Um, beyond the model layer, as I’ve talked about, we have become much more sophisticated. So it’s about the APIs, how do people access our products. API growth has been probably one of the biggest surprises for me this year in terms of our overall P&L. Then it’s the application layer, how do people feel and touch the app? And then finally that kind of substrate layer that we talked about with consumer hardware. And so I think we’re much more multidimensional.Today and so getting into like, who’s got the best model is probably something that keeps changing. But I do think it’s incredibly important that we remain state of the art frontier. And of course, I always think we have the best models.
19:26 spk_1
Uh,let me, next one for you, um, and you’ve heard this before, uh, OpenAI has $1.4 trillion in various commitments, and if it does not meet expectations, it’s going to bring down the whole tech sector. What say you?
19:38 spk_2
So I think first of all people get way caught up in the fact they think that’s all happening today. As I’ve talked about, our compute strategy has to have a multi-year lens to it. So it’s about compute we will buy over 5 years, 7 years, even 10 years. So when you hear a lot of these big numbers, think about them as much more spread out over time. Second, many of those numbers inIn some ways there’s a double counting going on in the press. So when we talk about taking down, say, a large data center site with our partner Oracle, we have to be able to put chips into those data centers. And so often we will then go do, um, a deal with say an AMD or an Nvidia or even our own chip hardware, by the way, because we know that’s now going to go into that data center. The press often hears.Both numbers and kind of says 2 + 2 is 4, and I’m saying, no, no, it’s 2 and 2 because it’s the same fully loaded data center with chips in it that’s going to be our compute future. And then beyond that, of course, we need to keep building our business model to meet our commitments, but thus far we are the fastest growing consumer platform the world has ever seen, the fastest growing enterprise platform the world has ever seen.Just even in my time of looking at the company, in 23, we did a billion dollars of revenue in one year. In 24, we did a billion dollars a quarter. We just a couple of months back hit over a billion dollars a month, and the model and the P&L has continued to build since then. So we’re rising to meet that challenge.And I think in a world where we’re headed towards AGI, AI, mass intelligence out there in the world, we just see no reason to stop right now in terms of the requirements and the need for compute.
21:24 spk_1
Philosophically, Sarah, inside the company, do you, do you need to be a profitable company to become a public company?
21:33 spk_2
So philosophically, I think you have to be able to show investors that there is a sustainable financial model in front of them.And the best investors in the world have looked at our financials and have made the investment decision to become part of our cap table, and I’m really proud of those investors, folks like T. Rowe Price, Fidelity, Dragoner, SoftBank, right? There’s many ultimateter. There’s folks that we have announced, and then of course folks thatWant to make sure that they stay somewhat confidential because that’s how they run their business. So that I think is the most important thing for investors is a belief in a strong, sustainable business that’s unlike anything that they have experienced before and that there is a path to profitability over time.And often what we can show them is a lot of our cohort behavior. So for example, what we have seen with chat GPT, it’s one of my favorite things. I call it the smile curve. So in terms of usage in consumer platforms, usually what happens is people join and then over time the cohorts naturally degrade because some people join, they don’t like the products, some people join, they forget to use the product, and so on.In Chat GPT’s case, what we have seen is as we launch more and more features, and it could be something fun and creative like ImageGen, or it could be something very deep and intellectually stimulating like deep research, more and more people who have joined come back and come back more frequently, and that is an incredible underpinning to a business model. And then of course this lift in enterprise, penetration, adoption, and ultimately revenue streams. Investors love that.Too, because that tends to be a very sticky revenue stream, um, where, where businesses see through things like connectors, a way to help their data come to life, and that makes, you know, us just an indispensable layer in their tech stack.
23:34 spk_1
Do youfeel the pressure from your investor base to go public next year, or they’re still saying, look, stay private, continue building what you’re building, come talk to us in, I don’t know, 2 years.
23:47 spk_2
I think our investors trust us to really look hard at our business and come up with the right strategy and the right path for open AI.That said, you know, if you think about a lot of investors, often they’re looking for an exit because they’ve been in a venture capital investment, for example, for 5678 years before that that investment really can mature. Again, what’s incredible about OpenAI is we really did our first fundraise last September. Up to that point.Right. Remember, we had really been funded largely by Microsoft, but we didn’t have a lot, really any external investors. And so none of our investors are, you know, none of their investments have been in there for a long period of time. We always keep an eye on liquidity because of course it’s not just about institutional investors, but our employees are.Also investors in the company. And so we think about things like liquidity and how we balance that. But right now you want to make sure you have a strong business model. So first of all, a strong product, and in our case it goes back one more step, incredible models, building strong products across consumers and enterprises that build a strong business model output.And if you do all of that, then you always have the optionality to decide what’s the best way to keep raising capital and to show up, um, in theworld.
25:08 spk_1
So, I wish you could see my paper because I have like 20 short rapid fire ones I have to ask you, but let me just pick three. These are short, just before I let you go. Short. Um, it’s 2035. What is the biggest impact OpenAI has had?
25:22 spk_2
I mean, you’ve brought mass intelligence to the world. I mean, we live in just a totally different world. Almost like imagining what it’d be like to live in a world without electricity or live in a world without the internet.
25:33 spk_1
Sam Almond in one word.
25:39 spk_2
Curious.
25:41 spk_1
OK. Best career advice you ever received.
25:46 spk_2
To not fixate on perfection. All right,
25:49 spk_1
we’re rolling here, Sarah. What’s your superpower?
25:52 spk_2
My superpower. I think I’m probably never, I’m tireless, so I will never give up. And I think we joked about it at the beginning. My husband often says I have only an on or an off switch. Uh, but I think I too like Sam, I’m very curious, uh, and I also will not stop until we get the thing done that we need to get done.
26:17 spk_1
Last one, I promise, the quality you value most in leaders.
26:23 spk_2
Uh, steadfastness, that what they tell you is what they do.
26:28 spk_1
Sarah Friar, thank you. Uh, congratulations. 11th annual Company of the Year Award here at Yahoo Finance. OpenAI. Uh, thank you for always making time with me, uh, for, uh, numerous years. I appreciate it. Thank you.
26:41 spk_2
Yeah, my pleasure and thank you. Um, it’s a total honor and as I said in the top, we’re really energized when we get recognized this way. So thank you. Thank
26:50 spk_1
you, I appreciate it and good luck, uh, in 2026. That is it for the, uh, Yahoo Finance annual Company of the Year award. Be
26:57 spk_0
sure to like and subscribe to Opening Bid Unfiltered on all major podcast platforms and YouTube.