SPEAKER_00 0:09

Thank you for joining us on the OneSearch podcast. On this episode, we're going to meet Sarah Freyer. Sarah Freyer is the CFO of OpenAI, presented with the OBE in uh 2019 by Queen Elizabeth II in 2026, named on uh the Times list of 100 most influential people in the world. It's a real pleasure to invite Sarah onto the podcast.

SPEAKER_01 0:35

Absolutely my pleasure, Dan. Great to see you.

SPEAKER_00 0:37

We haven't known each other all that long, but when we did get to know each other earlier this year, um, something that a lot of people may not know unless they've really done their homework is you're my fellow Brit.

SPEAKER_01 0:48

I know. I'll have to bust out my Northern Irish vial sign so people know that I'm actually from there.

SPEAKER_00 0:54

It's been a couple of weeks since we've spoken, and hopefully we've both gotten over the disappointment of the England World Cup exit.

SPEAKER_01 1:01

Yes, I was rough. My grandfather was English, so at least for uh one month every three years and 11 months I become violently English.

SPEAKER_00 1:09

Just briefly.

SPEAKER_01 1:11

Otherwise, I'm a staunch Northern Irish person.

SPEAKER_00 1:15

So look on that theme of you um growing up in a small town in Northern Ireland during the troubles, um, and then building this extraordinary career. I'm really gonna try not to be a sycophant, but it is I've got to tell you, I mean, um, of all the people I've had the pleasure of working with, you've got one of the most genuinely exciting CVs from engineering at Oxford, going into, I believe you won a competition and uh got a scholarship with um Arthur Anderson, you're in consulting banking, CFO of Square, CEO of Nextdoor, and now, of course, for the last couple of years, the CFO of OpenAI. And along the way, I was reading up on this. I saw you were in a gold mining Ghana. Just too many things to mention. That's before we get into board seats and non-profits. When you look back on this journey, how do you think about key moments that helped you get where you are? And and I guess a um a secondary question would be key people, mentors perhaps, that helped you get to where you are today.

SPEAKER_01 2:19

Yeah, um, I mean, it's a little easier sometimes with hindsight. As you can see, I'm a big believer in zigzagging a little and definitely taking risk. Um, I my brother's a doctor, and he's the doctor you meet if things are not going well. And he always talks about how like adrenaline is the thing they shoot into you at the end if they really can't bring you back. And so I used to have a thing in my head about like getting an adrenaline rush every single day. Because I think a lot of people get very kind of comfortable in their career, and and almost the adrenaline comes from not wanting to take the risk. And I feel like I've always been willing to lean in and take the risk. I think that's number one. Um, I think number two is like remembering taking a risk doesn't mean that you just have to change jobs. I've mentored many people younger in their career and said, hey, like when I went to McKinsey, the risk was going to work in South Africa. It wasn't deciding to go join a different company. When I was at Goldman, it was going from investment banking into equity research. When I was in equity research, it was about taking on, being willing to go the whole way to become the business unit leader. I could have stayed narrow. Um, and then I think though, there were also moments in my career, I'll I'll give you asked about who has um influenced it. Um, number one, definitely my husband. In fact, I have a kind of a two by two of mentors. I always say there's the people you work with who see you every day, they're great mentors. Um, there's the people you used to work with, and they take a little bit of a true serum, and you wish they told you some of the things when they did work with you. Um, there's the people who um love you so much they can give you the cold, hard truth. And I've put my husband very definitively in that square. And then there's the kind of the big reach stretch one. And for me, one of those people is Mark Benioff. Like back when I was at Goldman, trying to make this decision about what I fit in corporate. I felt like I wanted to build. I had kind of fallen out of love a little with the banking industry. Um, I called Mark, and I wasn't sure he'd even return my call as kind of a mentor. And he did, and he said, you know, the the line of I actually, he first of all, I was asking him about a different role. He gave me some advice. I hung up the phone. My phone called me back. I thought it was kind of pranking me, and I pack up and it was Mark, and he's like, Why don't you just come work with me? And so began a transition actually into corporate life. So um, I think a couple of other folks, Jack Dorsey, uh, would have to say, he always had this line about, don't tell me no, tell me how to say yes. And sometimes that can be really unpalatable, but it's still a better framing than just a dead no. Um, and then maybe finally, I would say, you know, uh Ruth Nora Johnson, who sorry, Susanna Nora Johnson, who was this amazing woman in the banking industry. She was the vice chair of Goldman. Um, what she taught me was to be very persistent when you want that coaching and mentorship. So I had tried for a while to get a meeting with her, and her assistant was like the block. She was a wall. I could not get through this woman. And then one day she kind of took pity and she said, Suzanne will be in a cab going from the tip of Manhattan up to Midtown. You better pray for traffic, but you'll get at least 20 minutes, but you need to be down here right now. So I ran out of my office, downstairs, got in a cab with Suzanne. She gave me that first 20 minutes of like tough talk, actually, frankly. But it was a good learning for me of like mentoring isn't always perfect. You don't perfectly set up meetings. You have to be really resilient and persistent, but those mentors can really turn your career around. And that conversation with Suzanne definitely did.

SPEAKER_00 5:41

Amazing. The taxi ride reminds me of the Will Smith movie in The Pursuit of Happiness, where he's in the cab with the he jumps in the taxi with the guy. Um, Sarah, how do you feel now you've become that you had this accomplished career? Obviously, still a lot to do in the future, but you are that person a lot of these young people look up to, and I'm sure they're hopefully persistently chasing you for your mentorship. How does it feel with the boot on the other foot?

SPEAKER_01 6:06

Yeah, I mean, it's always a good reminder. I'm a huge believer in trying to pay it forward. I actually have, I believe you have a nonprofit, but I have a nonprofit called Ladies Who Launch for women starting running and growing their own businesses. And it was a way to try to scale up in some ways. You know, it's hard to do one-to-one mentorship just given everything else on my plate. But is there a way to at least be able to do one-to-many? So that's kind of one thing I look at. Um, I do though get, you know, my assistant doesn't love this about me, but I do find it hard to say a complete no. So I often, um many people have gone on what I call walk and talks with me. So if you really want a little mentorship, I love to hike. I don't get a lot of time to myself to get out into the wilds. And so often on a Saturday or Sunday morning, if someone's willing to show up, I will take them out on a walk and talk. But you have to walk up like big mountains with me. And so it's kind of learning to fit people somewhat into my schedule, which feels sometimes a little lacking in humility, I guess, or being a little selfish. But on the other side, recognizing they want some of my time. I think the final piece there for me is recognizing too that sometimes if I get too far away from people and gets too abstracted, I lose a lot of what keeps me emotionally healthy and really helps me enjoy what I do because I get a huge sense of achievement out of giving back. And so creating some space for that is important, even when sometimes you look at a calendar where everything else seems so important. But if I don't create that space for my own emotional well-being, that can be a real miss as well.

SPEAKER_00 7:39

I did want to touch on the on ladies who launch and perhaps in the future find out more about that. Our nonprofit is young women into finance, and people in our community who are going to be watching this podcast are well aware of that uh project, which is really about underprivileged, very smart, underprivileged young women who don't have any role model in their network or a way in. And we find that the internships are really special ways to get in and get a foothold and realize that you can do this. And sadly, those internships we find oftentimes get passed around a community of people that are interlinked. So that's what our non-profit is about. And I will, in the future, I will pursue you on that front to see if we can get OpenAI involved. Um, moving this on, we obviously worked together uh recently. I was honored when you retained one search on a uh a search process. I don't want to leave the witness, but I do recall you saying the words better than great. I'll let you confirm that. Um but um on a serious note, how did you find that search uh process? But also, I guess more pertinently, in your experience, both sides of the fence, perhaps as someone who's been moved, as you said, zigzagged from one opportunity to another, but also as a senior executive looking to augment your team, how important is a good search film and how can they really make that difference?

SPEAKER_01 9:04

Yeah, I mean, if I pull back for a second on your question, first of all, just as a leader, actually, my North Star of leadership is always people first. I have a just a mantra that if you get your people right, everything else just works. Like great people will figure the A, the A team, even on the B idea, will figure it out much better than the B team on the A idea. So I love to get my people right. Um, of course, it starts with recruiting, getting the right people in the door. Um, and that can take all, you know, that it's a reason why I keep my own network quite warm. Um, you know, to the point about like have a hard time saying no sometimes. I do keep, you know, as a research analyst, I was taught the skill and thrived having a skill of always being able to network strongly into various communities. It was how I learned a lot of what was going on. And I think just keeping those networks warm personally, right? The the best way to go hire is to hire people that are in your network that you're going to get strong references from. Of course, then there's the balance of like inside a growing company, you have to build that muscle too. But I do like to augment it with specialists, and that's where a great search firm comes along. Someone like you, Dan, that's steeped in the area that I was particularly hiring in, which was around more infrastructure debt finance. It's not our core competence at OpenAI. That's why we needed an expert. And so to be able to get in and harvest out of your network was super important to me. And it sped it up. I think the thing that I often think about in hiring, I could probably spend six to nine months and maybe get to a great outcome as well, but you're going to speed me up. And that's why I often will use a search firm. But again, it's a balance because I want my company to also learn the scale of recruiting, right? Otherwise, it's just not super scalable. Um, what I like, you know, with a great search firm, um, that you take the time to understand the problem, um, that you listen carefully and kind of calibrate constantly. Um, in my case, like you manage to swoop in and find me someone that was already in motion. But often why I like a search form is I think about it as much about finding the candidate as it is about selling the candidate. So people forget that it's not just finding the person, but you kind of convince them to say yes and come on board. And I think a strong search firm feels, you know, to the candidate like they're their mentor, their friend, their their colleague, too. So in my own journey, right, I was found by a search firm when I was put into Square, and I really trusted them that they were finding a good slot for me. So sometimes you forget about it's not just finding the candidate, it's getting the candidate to convert. And I thought you all did that, you know, the whole thing was great candidate, great outcome, but at a really fast pace, which I valued a lot.

SPEAKER_00 11:53

Thank you for those kind words. And and and I will say, I mean, we've we've I've don't think I've I've ever seen a client move so quickly. And I do remember that first call I had with you, I think it was 7 a.m. And you said, this is my priority right now, and I'm gonna focus on this until we get this done.

SPEAKER_01 12:09

I also want to like make that point for people listening in. So you, if you say your core value is people first, you have to live it. And for me, that means knowing myself. So I am best early in the morning. When you said 7 a.m., that's what kind of tweaked for me. I was a rower in college. I'm used to getting up at 5 a.m. I worked Wall Street hours on the West Coast. It so I'm still the five, five thirty person, and my brain is flying from about like five through nine, 10 o'clock in the morning. Great prime time for me. And that's when I try to do the things that I think are the most valuable for accelerating the business. And so back to people first. It's often when I do recruiting, it's often when I do interviews, it's often when I do when I'm writing performance reviews. I try to do it when my brain is max fresh because it is my most important value. And I think a lot of people put their people things last. Like one of my tells on a new manager that's maybe gonna have a hard time making it, or I need to heavily coach them, is that they can't hire their team. And the reason is always I don't have time. I'm too busy doing the thing. And I'm like, you're never gonna get out of doing the thing if you don't start kind of getting the army on board. And so teaching people that skill of doing the thing that's most important, like building their team when they're fresh, is a really big deal.

SPEAKER_00 13:26

It's such a great point. And it's so easy to push calls or meetings with your team to the end of the day, by which time you're not fresh and things can go wrong.

SPEAKER_01 13:36

Absolutely. I mean, you're selling when you're interviewing someone, whether you like it or not, they're interviewing you too. And if you're not coming across as someone that's inspiring to them, that's going to help them do all the things they want to do, you're kind of making it so hard for yourself to convert once you decide they're the person.

SPEAKER_00 13:52

Exactly. Um, so switching this into AI and this, you know, uh I guess this next industrial revolution that we're involved in globally, and that you are uh, you know, one of the key protagonists leading. And OpenAI is in the middle of one of the most capital-intensive build-outs in in tech history. Media estimates keep rising in terms of the amount of money that you're going to be plowing into this this infrastructure, this compute infrastructure. Um, from your seat as CFO, why are you thinking so big and why does it have to happen so quickly?

SPEAKER_01 14:28

I mean, AI has come at us so fast, and yet I would say we're still in this early inning, right? The first phase of it. Um, what's changing is it's transforming the unit of work, right? It's going to help people beyond just individual tasks, but into really complex long-term workflows, things that will change the world, whether it's in areas like education, right? We see the whole way that teachers teach being profoundly shifted. Life sciences, like how people do drug discovery, how they do disease diagnostic. We just did something with our most recent model where we looked at the long tail of child diseases, pediatric diseases with Boston Children's Hospital that have not yet been diagnosed. There's just a bunch of symptoms, people don't know what it is. And out of, I think, don't totally quote me, about 360 as yet undiagnosable diseases, we found a path to 19 of them. That's kind of mind-blowing. Um, you get into worlds like finance, manufacturing, supply chains, and so on. There's almost nothing that AI won't touch. So as we think about where models are going, right, originally they used to think for seconds or minutes, but now we're talking about hours, weeks, months, maybe years. We're talking about persistent agents that are going to goal seek on things, right? That's really gonna shift the whole curve out on both what is needed today in terms of compute. Um, I'm gonna quote Jensen actually from yesterday. Jensen Huang said in his own podcast, we've lived in a world where we're lucky today, billions of people use compute for hours per day, to a world where maybe you have trillions of agents using compute 24-7. And so when I think about some of the building blocks, I need to help OpenAI get really max optionality on its compute. Because if that's going to be the world we're going to live in, we need to be able to build the compute that we're going to need for that future state. Um, and of course, we also view it as competitiveness with the rest of the world, US, Western economies, Western democracies. China is putting a lot of emphasis on here, a lot of investment. And I just don't think we can afford to lose competitively.

SPEAKER_00 16:39

Yeah, I mean, on that competitive point, is the biggest competition here domestically in the States, or is it coming from China?

SPEAKER_01 16:46

Sitting where I am today, I mean, of course we have competitors here in the US that we admire, take very seriously, and want to beat. Um and they're amazing companies in their own right. But I think broad-based, if you look at what's going on in China, there's definitely government sponsorship of both the labs themselves sharing info across labs, but also how they're ultimately going to build and finance that whole AI move. We're doing it here in the private sector, which I think is a massive fan of capitalism. Um, but we have to make sure that we're staying ahead. Thankfully, today we're at the frontier. Um, the model we put out last week, a SOL, is the frontier model. It's on the Pareto efficient line in terms of the cost for intelligence. You can go down into our more distilled models off of that one, Terra and Luna, but again, they stay on this Pareto efficient line. So you can ultimately find the best cost to get a task completed. I just put out a blog on this. I think matching models to tasks and so on, it's all a little bit, you know, not the right way to think about it. In the end, a frontier model getting what you might consider a more low-end task done, one shot, might be less expensive than that same task being done by a cheaper model, but taking five or six turns to get to the right answer. So again, I think we need to do a better job for clients to really understand how it what is my intelligence cost overall, and am I getting real ROI for the AI that I'm putting to work in my company?

SPEAKER_00 18:18

I'm gonna move on to the back to the point around the infrastructure, which is the world I've sat within for 20 years, of it's actually the 20th birthday of OneSearch next week, would you believe? Um and the definition of infrastructure has changed so much over those 20 years, and infrastructure has continued to develop into other areas. If you'd have said to me a few years ago that data centers would be, I mean, no one in infrastructure had 10 years ago even heard of a data center. Um so that whole definition of change, and you now find yourself financing the biggest ticket infrastructure projects of all time. You're squarely in the infrastructure investment world. How does that change the way you think about building your team, the types of skills that you're gonna need to take you and to take OpenAI where they're going?

SPEAKER_01 19:04

Yeah, I mean, I think you're nailing it first that we're building the physical backbone for a once-in-a-generation technology shift. It's bigger than electricity, it's bigger than the railroads. Um, unlike software, capacity can't simply show up when demand appears, right? That was the amazing thing about the internet generation, is you could almost build it mostly in real time as you needed it. Not quite, but it was not mismatched. In our case, you're building data centers, chips, networking equipment, power. All of these have to be planned years in advance. And in many cases, there's real scarcity at the moment. Um, so that means that capital is not just about funding growth, it's actually about securing capacity that could be make or break in terms of allowing us to stay on the edge of being a frontier lab and then being able to supply our customers. And so when I think about the team, we just actually did a little on-site, off-site, which are my favorite because they're fairly inexpensive to run, um, where we had my team, so fine from a finance perspective, the compute strategy team and the compute, the operational team under our CTO. And it's really a trifecta in this case. So you can't just be a purist about this is the best way to finance a deal, because actually the best might have a different definition. You might be willing to give up some economics in order to get some flexibility on something else that you're doing. You might be willing to make a trade with a partner for something they need in order to get access to a site that no one else is getting access to, right? So it's it's very much like an iterative process across those three areas. And that means when I'm hiring people, I need expertise, like the person that we hired together, but I also need someone who can think broadly, um, is willing to get down into the details, because this is not something you can fly at 50,000 feet on. And one kind of likes the, you know, just the grime that you get. You mean there's a reason I worked on a gold mine at one stage. Like I kind of like being right at the, you know, the face of where you're doing the digging. And it can get a little mucky and dirty, but you have to understand the details because that's where you'll find the opportunity. Um, and maybe just the one other thing I would say is that's that's a very kind of operator's perspective on what we're looking for. I think in all of this, we also can't forget about trust. Um, and how do we earn trust as we go into communities? We just announced a big power deal yesterday with Georgia Power in Effingham County in Georgia, it's called Project Camellia. Um, and really, it's about building a data center there, potentially up to 3.2 gigawatts of power. But it was important to also say we're not going to raise the price of your electricity, we're not gonna use your water. These liquid cooled, um, totally centralized cooling systems are very efficient. But also, we're gonna put $80 million to work. Community benefits. We're going to give you 70 million bucks of codex credits for all Georgia students. There's going to be hundreds of millions in tax revenues because you got to meet the community where they are, but also not tell them what they need. Like this is a thing I learned as a CEO of next door. You also need to recognize communities will tell you, but you've got to give them something to work with. So a lot going on from you know the guts of operating it, but through to like how do we earn trust and do it the right way? Because again, we're putting these in place, you know, all around the US and across the globe.

SPEAKER_00 22:34

Amazing. And and Sarah, final question. I know we're running out of time, but just for our audience who everyone these days, even small search firms like uh OneSearch, all of our clients, everyone's talking about how they're using AI. I've noticed some search firms have started to call themselves AI native. Um, you know, in other words, we live in AI. Not sure if that's really true in all cases, but everyone's using these AI tools. I presume if anyone's using them, effectively, it's you. Is there anything? Could you give us a sneak preview or a tease on tools that you use for forecasting, maybe investor materials or decision support that we don't know about, that maybe we will know about in the future? But how are you using AI?

SPEAKER_01 23:19

One of my objectives on my team, after people first, is to build the company of the future starting with finance. So our finance team is very AI-pilled. Um, today, if you were to look, our economic research team just put out a piece a few weeks back showing if you look at the total tokens used on my team as a measure of intelligence used. A year ago, 95% of that would have been in chat. So, in other words, people were summarizing a document, they were writing emails, they were maybe building a business plan, but it was very much language-oriented and very kind of static and very much ask. Like now we're shifting to do. And so today, actually, about 95% of those tokens are in today what you all will start to call Chat GPT work, so agentic behavior. And that means you're moving from asking to doing. So don't ask what our gross margins are going to be right now, but can you do build a dashboard for me that will show gross margins today, but also gross margins historically and in the future? So much more of a shift from asking to doing. We use it throughout, whether it's in procurement, to do things like credit checking, tax to pull tax forms for all our entities, fill them out, ensure they're accurate, to building dashboards for forecasting and so on. Maybe a fun one from just last night. So I at our local high school have been sponsoring a way for them to get their hands dirty on AI tools through the principal, so done through the school administration. She sent me two documents. Um, hopefully, she wouldn't mind me telling you the story. Both were kind of very wordy documents. Um, I love our principal, but that's how she sent them to me. I put them into my ChatGPT work and said, please create for me a site, which is effectively code for an actual website that I'm gonna be able to share. I want you to highly um, you know, really take these documents, make it very visual, make it seem very much the future, and create a slide deck for me that she could use as a pitch because she's gonna take this to other parents to effectively pitch them because we we're gonna have to fund it if we do it. And then I kind of let it go. Frankly, I was working on a bunch of other things. The amazing thing about agentic behavior is you do not watch it. It is doing it. It's like giving it to you know a member of your team. And very quickly, it came back with the most brilliantly um, you know, super designed pitch deck for her, which I sent her last night. I'm talking to her on Saturday, and she just sent me back mind blown. And I'm like, awesome. I feel like so good because I have no time and I figured out how to do this, and I we're just scratching the surface of how amazing this transition is gonna be for the world.

SPEAKER_00 26:06

Well, look, that's a great note on which to end. We're all looking forward to getting our hands on this technology in the coming weeks and months and the change that you and your colleagues are gonna make to the world over the next couple of years. We thank you for it. Sarah, you've been very generous with your time, and we're super grateful for it. And good luck with everything.

SPEAKER_01 26:23

Thanks, Dan. Great to see you. Good luck with for you too.

SPEAKER_00 26:26

And I'll I'll send you the uh the info on young women into finance. Please have a look.

SPEAKER_01 26:30

Please do.

SPEAKER_00 26:30

Thanks so much, Sarah. So grateful.

SPEAKER_01 26:32

Good to see you. Take care. Bye bye.