← Cheeky Pint
Cheeky Pint
Meta CFO Susan Li on headcount vs. GPU allocation, “free cash flow” hats, and almost becoming a PM
00:29:43
Loading summary
Transcript90 lines
- [00:00]
A
The joke people have about Instagram is the ads are better than the content.
- [00:02]
B
As someone who bought like 25 umbrellas that change color in the rain off of an ad, they work. They do. They are a source of real delight. Headcount's really easy to account for because you have org charts. GPUs don't have that property. In fact, you often want to build out your infrastructure.
- [00:16]
A
You can have shenanigans for it to be very fungible. Susan Lee joined Facebook in 2008. She became CFO in 2022. It's a really interesting time for this discussion because Meta has a core business that's firing on all cylinders and Susan's had a front row seat for of the company. Cheers. You went to high school at 11, college at 15, Morgan Stanley at 19, and you're now the youngest CFO of Fortune 100 company. So just what's. What's got. Like, was this you? Is this your parents? What's going on there?
- [00:52]
B
Well, you know, some might say, because I started kindergarten when I was 4 and I graduated from college when I was 19, that having 15 years of formal education is, you know, I'm woven undereducated, as it were. So I'm really just having to make up for that rough start. But if I remember right, you are also done with formal education at the age of 19, 20.
- [01:17]
A
That's kind of dropped out. I wasn't early in progressing through the milestones. I just quit. Whereas you actually got your.
- [01:22]
B
Well, it seems like we shared the same disdain for sort of getting out of the schooling system as soon as possible.
- [01:27]
A
You were interviewing, not mine.
- [01:30]
B
I was in a school system that identified when kids were bored in school and then just gave you opportunities to keep moving ahead. And my parents always took them. When I showed up at Morgan Stanley for my first day, I was on the trading floor in the big Broadway headquarters at 1585. And the equivalent of an HRBP basically got the attention of everyone on the trading floor.
- [01:53]
A
This is investment banking, which is known for being an inclusive and nurturing culture.
- [01:58]
B
Very, very much so. And so she wanted everyone on the floor to stop and look at and know that no one was to serve me any alcohol at any company gathering. So it was exactly the way you think about sort of beginning your career on Wall street by being mortified.
- [02:13]
A
But it improved from there.
- [02:14]
B
Yes.
- [02:17]
A
You worked under Michael Grimes at Morgan Stanley. For people who don't know, he's been leading tech investment banking at Morgan Stanley for 20 years. And he's just a Phenom, like, I don't know how to describe him. He's just one of the most energetic people I've ever met. What did you learn from working with.
- [02:35]
B
Michael Grimes is extraordinarily sort of like you said, very high energy. Applies that to a whole host of things. You go talk to Michael about tech companies, about banking, about parenting, about why there should be more undergraduate sales programs and colleges in the country. He's got a point of view on everything and he's endlessly curious. He is going to outwork you and out learn you. And it's actually a pretty spectacular thing as a young person starting in your career to see what really excellence at this looks like.
- [03:13]
A
You've been at Meta for a very long time. You joined in 2008, so you joined in 2008. And one thing I've observed before is just the senior leadership in Meta are all very tenured and often have kind of done multiple things around the company or grown up around the company. What traits do the successful leaders of Meta have in common?
- [03:36]
B
Oh, that's a good question. Infinite patience. No, it's more than that. So I joined as IC4 in finance. So really actually pretty far away from the center of the.
- [03:51]
A
A few rungs down from the cfo.
- [03:52]
B
Yes. And also far away from the core of engineers building newsfeed. Right. I mean, when you kind of talk to some of the folks now, they've always been at the sort of very heart of what the product was building or doing. But what I think is unique about Meta is we have a pretty strong culture of internal succession planning and trying to identify people who are talented quite early in their careers actually, and think about a many year Runway in which you're going to grow and develop them.
- [04:25]
A
How did Meta succession plan you?
- [04:28]
B
So I started off my career doing mostly revenue forecasting, which was kind of like the mathiest part of finance. And at some point I had done that probably for about 5 ish years and I was trying to figure out what to do next. And the two kind of paths in front of me were I was talking to some folks actually in newsfeed about whether I should just go do something totally different and go be a PM in News Feed or the other option was to broaden my scope in finance to take on sort of more traditional finance responsibilities that I really had had not that much exposure to. And I remember sitting down with David Ebersman, who was our CFO at the time, and he, he looked at me and said, look, I know you're considering these options. And I can tell you I think doing that newsfeed PM job would be really fun and I think it'd be a great learning experience for you. I totally get it. But I also want you to know that I think you could be a CFO of this company someday. And to have someone who I admired as much as David Eversman say that about me was an extraordinarily confidence building thing. And I will remember that conversation forever, very viscerally so I've had managers who I think have really invested in me by pushing me to take on things that I wouldn't have said, oh, I'm, I wouldn't have said, hey, can I please go do this thing next? It wouldn't have made obvious intuitive sense to me. But I think they thought it would be a good opportunity and that I was ready for it. And you know, and I think, and I think they were right.
- [06:01]
A
That's really cool. In the 17 years you've been at Meta, how has Mark changed as a leader?
- [06:10]
B
You know, there are ways in which you clearly see someone evolve over 17 years. You know, Mark has done all hands for all of those 17 years and he clearly has become now a truly excellent public speaker. Mark is really good at giving feedback, like really world class at it. And maybe you should try to get yourself into a position where you can get some feedback.
- [06:32]
A
I'm sure Mark already has feedback for me. Yeah.
- [06:34]
B
So you can experience it. But it's very timely, it's very direct, it's very respectful. But the sort of direct and respectful, it's never mean, it's never like belaboring some point. But you cannot be mistaken after you have received the feedback. Yes, he's really good at it. He kind of walks that line between being direct but kind in an extremely good way. You know, one of the things people will often ask me is like, you know, what is, you know what, what kind of skills do you need to stay at a company for 17 years or whatever it is. And when I think about it, I go back to I, when I, when I was IC4 and I joined in 2008, I'm building these first revenue models and the, you know, I'd gone from banking, which is super organized, super structured. They don't even need to know your name. Like they just train you to immediately figure out how to find the backup to everything so that two years later someone else can do this and so on and so forth to. There was no infrastructure, right. So I'm like hunting down the exact engineer who has Built some ad server so that he can tell me what the parameters mean. And of course the next time he changes them, he's not going to tell me and I have to go find him again. It's like, oh, she's coming if I don't look her way. But a few months in, I got a meeting invite for power users of SQL. I thought, my gosh, I've been getting a good amount of feedback about how things could be better. So here was finally this moment of recognition that I didn't even know how to write queries in SQL when I started. I show up to this meeting and there are five other people and the meeting organizer tells us that we have been called because we are the five users of SQL who consume too much power and we have just been churning with our massive joins tables through the infrastructure. Basically, yes. But I often think back to this because this was a data analyst who didn't know any of us that well, but it just generated his reports of who's using the most infrastructure and looked at the top people on the list and thought, okay, this person in finance, it doesn't make sense why she's the third highest person on the list and called us in and then taught us to write better queries. And no one, I think specifically told him to do that. And I think it's a little awkward when you call people in to do this, but he did it because it would make us all better at our jobs. And I think for 17 years I have been the beneficiary of a lot of feedback that has made me better along the way. So when people ask me this question, I always say, just be a person who's good at receiving feedback.
- [09:06]
A
Yeah. You've mentioned your experience in forecasting. And what I think is the central challenge of CFO in a large tech company is it's so hard to put numbers around the core thing we do. And what I mean by that is if you're Boeing and you're producing the 787, you have a very clear model that we're going to spend this much manufacturing the 787 and then each one we're going to make this much gross profit on. And then at the component by component level, we're going to change from hydraulic brakes to electric brakes and it'll add this much cost, it will save this much fuel. It's all extremely quantified as a domain.
- [09:39]
B
You're driving me into the resource allocation questions. Yes. Okay, here we go. We really think about it as there's Stuff that we can rigorously measure. Right. So that's a lot of the core family of apps work in terms of the impact on engagement, the impact on monetization. There's a lot of that stuff that is really finely tuned where.
- [09:59]
A
And that really does seem extremely finely tuned. Like I was looking at the numbers and you doubled ARPU between 2015 and 2020 and then you doubled it again between 2020 and 2025. But like Meta wasn't bad at monetization in 2020 and it's doubled over that five year period.
- [10:16]
B
No. And you know what? I just did earnings two, three weeks ago now and was doing all my investor callbacks. And one of our largest investors on the call, one of the portfolio managers said, feeling pretty good. He goes, you know, the ads are so good. And you know what? Five years ago I would have told you that the ads were really good and that there was not really room for the ads to get better. But here we are five years later and the ads are even better.
- [10:49]
A
The joke people have about Instagram is the ads are better than the content.
- [10:53]
B
Well, I have to tell you, as someone who bought 25 umbrellas that changed color in the rain off of an ad, that was not something I knew that I, not that I needed, but that my children and all their friends needed.
- [11:04]
A
Do they work?
- [11:05]
B
They do. They are a source of real delight. So, you know, when the ads can be that good, that is a extraordinary thing. But getting back to your question, so there's this very sort of measurable part of the company and we generally try to trade those things off against each other. You know, when we are thinking, when we're evaluating things within that bucket and we generally try to fund the things that are positive roi. And I'm usually the person who's trying to just make sure we understand, like, yes, for every individual experiment, the expected return is something, but that's where we are in the curve today. But what about 50 experiments later, does the curve still have the same slope? And then there's a set of things, right, which we constrain more in terms of, you know, the, you know, there's some envelope of, of investment that we're willing to make that's not in this really ROI driven bucket. It is very difficult to pencil out what the annual revenue forecast for Reality Labs is going to look like over the next 20 years. And so for bets like that, we sort of invert the problem. But when we talk about the return on the investment, the question that we pose as a finance organization to Mark is. And make sure that Mark and the board understand is what does this have to be worth to pencil out at the end? Right. And does that pass sort of the sanity check, the intuition about what building, about what the size of these markets can be based on, maybe some comparisons to, you know, markets that exist today. But of course, you know, in another 10, 20 years, you expect that the world will look different and maybe those markets should be bigger or smaller for, you know, whatever reason. And. And that's kind of the guide which is like, hey, for this thing to succeed at the rate at which we're investing, it needs to be worth this at the end. And, you know, does that make sense?
- [12:55]
A
So, in a way, investors may underestimate your ambition in some of these new areas where it's like, this is not a hobby. This is us investing in markets that are worth a huge amount of money if we create a new platform here. But the thing people may miss is that the upside case you're considering is really serious.
- [13:15]
B
Yes. And we're only building because we think that that sort of. It not only exists, but it's compelling. And it's compelling for financial reasons, but also strategic reasons why we want that version of the world, you know, to exist. And this is a place where, I've got to be honest with you, like, I was one of the last people at the company to hand my BlackBerry over for an iPhone.
- [13:38]
A
So you're maybe not the.
- [13:39]
B
I am not a tech visionary there. You know, there are many things I'm good at, but sort of envisioning the future of the world and what I want it to be like is not one of them. I'm a very happy beneficiary of the technology built by the world around me. But Mark very much has a vision for what he wants that world to be. Right. And so I think. And for him, I think the sort of strategic imperative is that we have to be building these sort of next states in the world for us to again, be a good business, but also just be a compelling company that builds technology and puts it out in the world and, you know, builds incredible experiences for people. I remind people in the finance organization all the time is like, you know, we are very good at skeptically evaluating each bet. Right. But the point is not that we have to look at every bet and be like, this bet is going to work. The point is there is a portfolio of bets, right? And some of them are going to pay off massively. Beyond, in fact, what sort of the case on paper looks like when you make the bet and many of them are going to not work out, but the ones that pay off are going to more than sort of justify the overall investment strategy or the overall sort of roadmap that you're building toward. And if we just allowed ourselves to nix everything that sort of, you know, the paper case didn't seem high confidence, then we would never make a lot of the important bets that are, I think that have been really important over the history of the company.
- [15:05]
A
When did you take over?
- [15:07]
B
November 1, 2022.
- [15:08]
A
Okay, yeah, so the, I think the day you took over the market cap troughed at $230,000,000,000.
- [15:18]
B
A real sign of market confidence in me, as you can tell.
- [15:21]
A
You probably remember what the number was. I think it was around $230 billion. And so that means the day that you took over as cfo, one could have bought Facebook or, sorry, Meta, excuse me, as an investor for three times 2025 net income. And that's like coal plant territory. I mean it's a very easy way to make money is to buy good and growing businesses for three times net income.
- [15:46]
B
Well, I hope you did.
- [15:48]
A
I did not. And this is why I'm not in the investing business. There was something that people deeply misunderstood at that point about Meta. What did they misunderstand so much?
- [16:02]
B
Well, there's a bit here, by the way. Someday I'm going to ask you how you feel about having public market investors someday and when will that day be?
- [16:11]
A
In my interview.
- [16:13]
B
But more to the point, you know, that sort of October 22nd moment happened at a like there were multiple things going on. If you kind of rewind the clock. There were sort of two big revenue headwinds. One was that the sort of platform changes with ATT had kind of rolled through from 2021, which is when apps.
- [16:33]
A
People changing their policies around what tracking was permissible inside of apps.
- [16:37]
B
Yes, exactly. So that was one thing. And then the second thing was just this sort of COVID fueled e commerce avalanche was pulling and both of those things very.
- [16:48]
A
Or buying fewer color, changing umbrellas, sadly.
- [16:51]
B
For the children of the world. Yes. And so both of those things had the effect of unfortunately happening for us at the same time. So we really like, you know, went from this e commerce fueled heyday in 2021 to now like negative year over year growth for the first time, which is obviously very alarming. And so, you know, those stars kind of aligned in that stock price low kind of way in October 2022. And I think what you since then is a Few things. One is that yes, there are these two exogenous factors that happened that were bad for revenue at the time. But the fundamental sort of underlying like business, which is can we show the best possible ads to the right people at the right time across, you know, the, the surface of consumer experiences that we are building that continued to be very strong. Right. And then the second thing is I think we demonstrated as a company that we are in fact able to turn the ship on costs in a very, very meaningful and very quick way.
- [17:50]
A
Speaking of that, you have to explain the free cash flow hat. And thank you for the hat by the way.
- [17:54]
B
Oh yes, well you're welcome. Everyone really should have one. I think they are underworn out in the world. The joke is of the story is that Mark at one point gave me an Ebitda hat which was a very kind gift from him to help me.
- [18:11]
A
Really send a message like I hope you went and you know, prominently wore it or in many of the budgetary review meetings that you were in.
- [18:17]
B
I did.
- [18:17]
A
Ye, yeah, I did and I. This is the hat that Mark gave me.
- [18:20]
B
Yes. And I had it in my background, you know, my zoom background for a long time. But I realized pretty quickly that we actually as a company should be wearing free cash flow hats instead. Because of course the D of EBITDA is a number of growing importance, you know, through our financials. And so and I didn't want Mark to misinterpret and feel like, you know, EBITDA was going to be the end all, be all financial metric for us. So I'm now there's only one EBITDA hat. There are many free cash flow hats. I give them out like candy and try to make sure that people really understand that this is the hat that matters.
- [18:59]
A
You know, Charlie Munger had the joke that anytime you hear EBITDA you should substitute with bullshit earnings. And so you similarly for a capex intensive business, you want to make sure people are not forgetting about the CapEx.
- [19:08]
B
Yes, exactly.
- [19:09]
A
Where does CapEx go for not just Meta but the tech industry broadly, because all of Microsoft, Google and Meta have gotten more capex intensive over the past few years compared to their prior steady states. Do we continue spending this fraction of revenues on capex over a 5 or 10 year period? Do we somehow get some kind of amazing compute gains? Are we ultimately like we're bottlenecked on power and so you just can't keep growing capex at this rate because you can't plug the data centers into anything. But where does Capex go at an industry wide level.
- [19:43]
B
That is the question that I assume that all of my counterparts at these companies and I are all thinking about. For us, there are the drivers of kind of the way we're investing in capex today. Of course we have first of all just a massively scaled consumer business and core AI infrastructure that powers all the ranking and recommendations work and so on and so forth. So that's always been a reasonably big number for us, but also one because it was getting more mature that we were sort of driving to be more efficient over time. And then now you have among many of our peers and ourselves this big investment to train what we all aspire to be frontier models. And then if you use those models to build great and scaled consumer experiences, then how much inference computer are going to need on top of that, if just compute required continues to scale up in this way forever, then you're going to run into some true problems of physics. But hopefully there will be different kinds of research innovations along the way that will unlock things like being able to distribute the training. So you don't need sort of one extremely large cluster somewhere and that will help with a lot of the energy and other challenges. So there's some question about just what that looks like over time. And then there's this question about, you know, great, you can build all this capacity and what do you do with them if it turns out you don't need as much compute for either training or inference as you thought? And I think a lot of us have different backup use cases, right? So up to some point we would use a lot of compute very happily still, you know, in the core business and what we expect the core business to be three years from today, but frankly we'd use more compute in the core business. Now that doesn't scale forever. Right. So the real question is what happens in like two years if you've built so much compute that you cannot envision a reasonable ROI on the backup use case if what you're building doesn't come to fruition. And that's something I think we're all going to learn in the next few years.
- [21:36]
A
And sorry, when you say the primary versus backup use cases, the primary use case is new products like Llama and stuff and the backup use case is ADS optimization.
- [21:45]
B
Yes, exactly.
- [21:46]
A
You mentioned just doing earnings. Is there a specific anecdote that you can or want to Discuss.
- [21:54]
B
In the October 22period? So we had an earnings call the end of October and as usual I'm doing investor callbacks and you know it was a pretty, you know, the investors were not shy about their feedback.
- [22:08]
A
Yes.
- [22:08]
B
And in fact, one of the calls.
- [22:09]
A
You know, investor callbacks. I don't know what this is. This is where you called. This is like one on one, basically.
- [22:15]
B
Yes. It's pretty standard after earnings calls where you touch base with like some number of your largest investors. Sadly, it is not one on one. It's, you know, one of you and many, many people from their teams and most of the time they just ask you to, you know, clarify things. Obviously everything is, you know, reg FD compliant, but it mostly takes the form of questions. And, you know, in October 2022, for the first time, there were sometimes no questions. I mean, there was a call where basically one of the portfolio managers said, we actually don't have any questions for you today. We just want you to hear feedback from us.
- [22:49]
A
Wow. More of a comment than a question.
- [22:51]
B
Yes, it was actually very memorable. And so one of the things.
- [22:54]
A
And it was blunt feedback, I presume.
- [22:55]
B
Yes. And one of the things that really stuck with me from one of those conversations is someone said, look, I get that you're building the next, you know, the future of computing and the next mobile platform and all that, and that is great and I am glad someone wants to do it and I am rooting for you, but why should I invest in your stock today? Like, why don't I just wait for your, you know, your phone equivalent, you know, your scaled consumer product to come out, you know, and invest in you then? And you tell me that that's going to be like years away. And the way that question was framed actually really stuck with me and is the way that frankly, now Mark and I think about this, which is like, great, we've got a lot of these bets and the bets are technologically exciting. People can get excited about them and the vision of the world, but as investors, they're like, cool, why don't I just wait for your bets to be ready, be ready to succeed before I come? We need people to invest with us along the way. And when we think about the financial outlook of the company, you know, a large part of it is not just, okay, cool, you're building the next, you know, massive platform out here in some decades. It's, why would you hold our shares until then and what do we need to keep delivering in terms of consolidated results?
- [24:17]
A
I find it really interesting how when the AI revolution started really ramping up, people realized, oh, we need a ton of GPUs to train leading edge foundation models. You guys had Done a huge GPU scale up because you're just doing a lot of AI in the core feed. And so I think there's some interesting optionality in being a scaled infra and AI player where we are very good at putting GPUs towards their highest and best use. And you have seen that we're very good at allocating compute and that is why you should invest. And that's quite different from the pitch maybe 10 years ago where we're good at scaling social products.
- [25:04]
B
Yes, I think there's definitely an interesting point there. You know, as part of not wanting to miss the boat, you know, we built out enough capacity for reels, but also for like future things, and we found that we were in fact able to put that capacity towards very good use. Exactly as you said. So I do think an interesting question in the future will be, I think allocating compute as a resource, that's something we. It's a muscle we've built later as a company. Right. Because we had gotten very good at allocating headcount as a resource. And headcount's really easy to account for because you have org charts. So you know. Exactly. This person reports to this person, to this person, to this person is incontrovertibly working on Facebook marketplace, for example. GPUs don't have that property. In fact, you often want to build out your infrastructure.
- [25:50]
A
You can have shenanigans for it to.
- [25:51]
B
Be very fungible because you want it to. You need to divert capacity to where suddenly something has happened in India and you want a lot of compute to be available to be used there. So it's not all like this GPU is labeled for Facebook Marketplace and this is labeled for. And so it's actually quite a bit more difficult to account for. You know, where the capacity is being used at any given point in time. And that means it's harder to manage and it's harder to create the incentives around. Like, are you using GPU effect GPUs efficiently?
- [26:18]
A
You allow people to trade between people and GPUs. Right.
- [26:23]
B
In the budgeting process, we have allowed people to trade. And not too surprisingly, even though you'll find that groups are often asking for compute when that particular trade is on offer. People almost never trade for compute for exactly the reason I described, which is that if they get allocated 100 new headcount, there is no chance that 26 of those headcount will accidentally be working for something else.
- [26:53]
A
Yes, I see. So again, it's harder to account for, but you could Joke that AI has shown up everywhere except in the large company hiring plans. And when I talk to startups, sometimes they are actually delivering, they're having a huge amount of impact with a very small number of people, and they plan to grow headcounts slower than maybe the generation of startups that came before them. How do you think AI productivity actually shows up at more established companies like a Stripe or like Imeda that just have a larger installed base?
- [27:26]
B
Yes. When we think about AI for productivity at Meta, I think there are two dimensions. So one is how do you make the most operational parts of people's jobs less so and more interesting? And I say that as a person who is like a very expensive machine learning model for approving expenses. Right. I'm not certain that when I approve expenses, I'm really adding a lot of deep human intelligence to this process. I'm scanning for a fairly checklistable set of things and yet I get multiple expenses every day. And so how do you take that part?
- [27:56]
A
Really funny ones.
- [27:58]
B
Those are concerning. Yes, some of them have taken me down some really interesting rat holes. But so how do you basically make those parts of people's jobs automated so they can do more interesting things? And the second thing is there are actually things we don't do enough of today because right now they're pretty low ROI to do. And so like the kind of canonical example is everyone knows someone who has gotten locked out of their Facebook or Instagram account. It is a pain to get back in. We know it is a pain to get back in, but it's super laborious, the process of like verifying that you're a real person, you have real friends on the platform. It's a hard problem. It's a hard problem. If we could actually make that more efficient and more productive and enable a currently sort of a human reviewer or customer service agent to go from reviewing. I'm making up these numbers, but five a day to 50 a day, unlocking 50 accounts a day. You can actually make this a pretty high ROI thing to do that you would invest in on an ROI basis alone. So I think there is a bit where I think everyone is sort of worried about the world where the machines have come for all of our jobs. Definitely my expense approval job and maybe more. But I think there's actually a window before that where I think it's really about making humans substantially more productive than.
- [29:18]
A
They are today and makes new kinds of things possible that weren't economic or just kind of possible. Yeah, yeah, I've kept you for way too long. Thank you.
- [29:26]
B
Thank you so much for having me. I really look forward to seeing that Tesla hat everywhere in the wild. It is the perfect photo accessory.
- [29:33]
A
There we go.
- [29:34]
B
Yes.
- [29:34]
A
It's a good look.
- [29:35]
B
And it's green.
- [29:36]
A
And it's green. Exactly. Thank you. It's very culturally unbound. All righty.
- [29:40]
B
Thank you.