CEOInterviews.AI
Start App
Get an email when Paul Kedrosky gives a new interview
One email per new interview. Unsubscribe in one click.
Paul Kedrosky
Managing Partner, SK Ventures

Nvidia's Data Center Threat, AI Stock Rebound, OpenAI's Codex Strategy

📅 Nov 06, 2026 The Information 64 MIN 939 VIEWS 149 SEGMENTS · 6 SPEAKERS
SK Ventures General Partner Paul Kedrosky joins TITV Host Akash Pasricha to explain why proposed data center construction pauses could threaten Nvidia's growth and reshape AI infrastructure spending. Plus, Lyft CEO David Risher discusses Waymo, Dropbox Co-CEO Ashraf Alkarmi explains the company's Tigris acquisition, and OpenAI's Andrew Ambrosino and Every CEO Dan Shipper discuss coding agents and AI model competition. Chapters: 00:00 Intro 01:13 AI stocks, data centers and Nvidia | Paul Kedrosky 12:49 Lyft earnings, subscriptions and Waymo | David Risher 20:16 Dropbox's Tigris acquisiti...

Questions asked in this interview

12
  1. 2:34I mean, like what's going on?
  2. 4:20... estimate that sort of suggested that based on the Texas data center halts, Nvidia's revenue, depending on your assumptions, I mean, it could be impacted somewhere between two and 15%, which is a big range. Depends on how many, right?
  3. 5:41Do you think Nvidia has anywhere else to look for growth in this world where maybe Texas, you know, the data centers in Texas are not buying as much as they would have once had?
  4. 7:37Do you think that that pressure, you know, decreases at all because even if the tools aren't getting more cutting edge, I mean you still have the story of transformation, right?
  5. 9:20Why do you think that is? What's the story here?
  6. 10:39Does that match what you're seeing or what are you expecting?
  7. 33:13And I mean, help me understand when you think about the technical challenges that Pierre solves for you, what were those challenges?
  8. 34:28So when people think of code storage, I mean, is this like a competitor to GitHub? Are you still using GitHub?
  9. 38:10All these different platforms use, you know, how do you think about the nuances that you want to be ahead in the tradeoffs on the product level?
  10. 40:04How has that changed today compared to even six months ago from your observing?
  11. 41:35I mean, what do you think when you see those headlines?
  12. 43:07So it doesn't concern you, is what you're saying?
Akash Pasicha 0:22 ↗
Heat. Heat.
Heat. Heat.
Welcome everyone to the Information TI.
My name is Akash Pasicha. It is Friday, November 6th. Today on the show, we are going to dig into the big selloff we saw yesterday on Wall Street for tech companies and whether or not the narrative around the AI buildout can recover. We're also talking to the CEOs of Lyft and Dropbox about each of their quarterly results. We then have a conversation for you with OpenAI's member of the technical staff, Andrew Ambrosino. We talked about OpenAI's latest acquisition and how the company is looking to stand out with its product strategy. I sat down with Andrew at the Thesis Conference, a big event here in New York City hosted by media and AI company. We're also going to play my conversation at the event with Dan Shipper, Every's founder and CEO. We talked about why one unexpected player in the model race right now could have the edge. It's going to be a great show, so let's get right on into it.
Tech stocks are looking to rebound after Thursday's steep sell-off which hit some of the biggest names in AI. Nvidia is back in the green this morning after falling 9% yesterday. It is a similar story with Micron, AMD and Marvell. To break it all down, I want to bring on Paul Kedrosky, general partner at SK Ventures. Paul, welcome back to the show. It's great to have you here.
Paul Kedrosky 2:33 ↗
Hey guys, good to see you.
Akash Pasicha 2:34 ↗
Okay, so what do you think here of the selloff yesterday? Were people just asleep? Did they not wake up the day earlier? I mean, like what's going on?
Paul Kedrosky 2:44 ↗
You know, the market's a really funny place. It came to a conclusion. So, let's just go backwards really quickly. So, 48 hours ago back to the midterms and the market came to a particular conclusion about what the midterms meant in terms of the AI buildout at large. And it was really funny. I was in New York in a bunch of meetings with hedge fund managers and asset managers and others. And the conclusion everyone came to was actually this is good. This is actually good for the hyperscalers because if you constrain the supply of data centers and you're a seller of data centers, it's no different than constraining the selling of oranges, it's good for the orange sellers, which is a really, I mean, glib and naive take on what constraining data center construction might mean. And by constraining, of course, I mean that something like, and I put this out the other day, but something like 41% of the data center construction queue in the United States, which is in the hundreds of billions heading to trillions, is in states that now have governors and governors-elect that have said in one form or another that they'd like to either pause or delay the data center buildout. So if you work backwards, I grant you that, you know, as an orange seller, this may be good for you. But as a data center supplier, this is very bad for you because you're seeing a significant fraction of the queue become at the very least pushed out, if not actually cancelled. And so if you work backwards and do fairly straightforward math, it turns out there's this company called Nvidia, for example, that's a very large supplier to many of these data centers in the buildout. And so of course you can map out exactly what the consequences might look like in terms of revenue shortfalls and delays. And that's finally the math — finally the markets did the math and it turned out the figures weren't so good even for the orange sellers.
Akash Pasicha 4:20 ↗
Yeah. So look, you put out a note, I can't remember if it was yesterday or today. Well, you published a lot of it. I mean, you flagged this Baker Institute estimate that sort of suggested that based on the Texas data center halts, Nvidia's revenue, depending on your assumptions, I mean, it could be impacted somewhere between two and 15%, which is a big range. Depends on how many, right? But I mean, I guess my question is, you know, it's not a small proportion. Even if it's 10%, you know, it's not a small number.
Paul Kedrosky 4:54 ↗
No, it's not a small number. And you know what? Just to really quickly and not to interrupt but to put it in another kind of context. It's not just that those data centers are in states where something like 40% of the queue is potentially delayed. It's also the global scale of Texas alone as a data center hub with respect to construction. The way I map it out globally that it goes like this: the US, Texas, China, and then Texas if it were a sovereign nation — which I'm sure would be delightful for it — if it were a sovereign nation would actually be the third largest hub with respect to global data center manufacturing. So this is consequential not just in terms of the US or regionally, on a global scale. And so this goes back to this point you're making that, you know, the maths matter because Texas in effect is a global player with respect to the size of its data center market for construction.
Akash Pasicha 5:41 ↗
Do you think Nvidia has anywhere else to look for growth in this world where maybe Texas, you know, the data centers in Texas are not buying as much as they would have once had?
Paul Kedrosky 5:56 ↗
No, because the permitting and buildup process is so slow. This just becomes this game of saying I now have a new project and it's we have it speced for, you know, Louisiana or somewhere else. And the problem of course is, and we've seen this in the recent construction data, that the queue itself is very, how shall I say, slippery, the data center construction queue. So whenever utilities ask for deposits — and these can be modest deposits, not on the order of $20 I'll grant you, but you know, in the order of low millions — suddenly the data center construction queue begins to shrink dramatically. A northeastern utility found that requiring deposit shrank it by almost 70%. And so the queue itself, it's very difficult to understand what's actually in it because it's so far out in terms of when actually things will get built. So yes, the answer is yes, you could play that game and no, it doesn't matter.
Akash Pasicha 6:42 ↗
And nothing overseas, you think?
Paul Kedrosky 6:45 ↗
Well, no, because again, think back to the size of Texas as a sovereign nation. Texas is a sovereign — to replace that, there is no place where you can replace that size of queue if you grant that something on the order of, you know, 25 or 30% of the projects within this so-called sovereign nation of Texas in the data center sense is going to see delays. So no, it's not straightforward to see how you do that. And I mean, obviously this has all kinds of other consequences which I've written about many other times. The data center is a de facto stimulus program in the United States. It was on the order of 25 to 40% of US GDP growth for 14 quarters in a row now, which is one of the reasons why the prior — while the current administration embraced it so strongly — that it was a de facto stimulus program that was in part bankrolled by the private sector and then steamrolled a lot of local interests in terms of the permitting and buildout. So this has consequences in terms of the health of the economy at large, not just, you know, Jensen Huang's bankroll.
Akash Pasicha 7:37 ↗
So let me ask you this. You know, the data center buildout is one half of the story. The other half is businesses trying to find efficiencies from the AI tools that we have as of now. So even if the models don't continue to get better, I mean we have these models and there's probably still some pressure from shareholders to use them. Do you think that that pressure, you know, decreases at all because even if the tools aren't getting more cutting edge, I mean you still have the story of transformation, right?
Paul Kedrosky 8:10 ↗
Oh, of course. Listen, and this is the thing that people get wrong all the time is this whole idea that you can hold two contradictory ideas in your head at the same time. That the data center buildout is on this precarious point doesn't mean that large language models don't work. That would make all of this much easier, right? The problem is they actually work remarkably well even in their current incarnation. So as you say, without having to go to building out even more advanced models, corporations under huge pressure to not only find efficiencies, but more specifically, and you saw this with Revolut the other day, was to say, 'Wait a minute, this is costing me a fortune, and I grant you it works, but I need to find a way to make it less costly.' So what the big pressure you're going to see going forward is increasing on-premises use of large language models both in terms of sovereignty of data and cost reduction, which is bad for cloud-based large language models, frontier model companies, but tremendous in terms of the likely productivity consequences. And you saw that yesterday or the day before, I'm losing track, in the BLS data with respect to the latest productivity gains. And the only caution I'd make about that was that most of the productivity gains came from capital not labor, so the returns are accruing to capital not to labor, which is a very, very sort of dangerous moment in terms of where the gains are coming from from AI. It's coming, it's just not accruing to labor.
Akash Pasicha 9:20 ↗
So you also put out this other note talking about how S&P 500's earnings growth ex-Mag 7 for the first time since 2022 outpaced the Mag 7. And this of course keeps in mind that Nvidia hasn't yet reported. Why do you think that is? What's the story here?
Paul Kedrosky 9:43 ↗
Well, it's the intense cyclicality of the buildout that in the early years was relatively straightforward to fund it all from cash flow. And so funding it from cash flow was tremendous because I didn't have this overhang of debt obligations which in turn was eating incrementally into my cash flows alongside of this what I was spending on data centers. So now I have an extra — I'm feeding an extra hungry animal and the extra hungry animal is external financing, of which is now something like 50 plus percent, 55% of the funding of data centers. So that's the problem. So increasingly we have a lot of mouths to feed with respect to cash flows that are flowing out of these large and heretofore very profitable hyperscalers. And so that's beginning to cut into — and we're seeing this with negative cash flows almost across the board in the hyperscalers. And so it's really remarkable to see for the first time since the buildout really began that not only do many of them have negative cash flows, but for the first time the rest of the S&P is outperforming them in terms of earnings growth.
Akash Pasicha 10:39 ↗
Okay. And because you follow the data so closely, I mean you mentioned the estimates around Nvidia sales, how that could be affected. Have you seen anything around how capex growth could be affected? We had somebody on the show earlier this week talking about, I mean, the next year's capex probably unaffected, maybe the year after something like maybe 10, 15% at risk here. Does that match what you're seeing or what are you expecting?
Paul Kedrosky 11:06 ↗
No, it's on that order of things. And I think, you know, Bloomberg Intelligence had a report out, I think it was Goldman had a report out yesterday. And again, all of this is belated because this was all kind of baked in the cake after the midterms, but nevertheless, at least they showed up to the party and began doing the numbers. And so, yeah, on the order of something probably not — obviously this year, but looking out next year into 2028, on the order of sort of 12 to 16% would be my estimate, which is obviously, I mean, a titanic number given that we're dealing in trillions.
Akash Pasicha 11:33 ↗
Okay. And last question for you, Paul. I mean if you just look at the story that these big tech companies tell broadly, I mean they have been able to get away with this narrative that we will lose, you know, we'll burn cash or we'll invest for the long term, this is the game we're playing and, you know, we'll see the results three years out let's say. I mean in a world where they cannot tell that story anymore because like you're saying the estimates might come down, what do you think is like the new — is this like a year of efficiency story that we start to hear again come through in the next year or what do you think is the broader narrative they start to all adopt?
Paul Kedrosky 12:13 ↗
On-premises. That everything gets pushed on-premises because they're getting huge pressure from their customers to do AI on-premises. So you start seeing these hybridized cloud models of saying, you know, it doesn't have to happen just in our cloud, it can happen on your premises. So there'll be lots of talk about, you know, confidential data and secure computing and blah blah blah all this geekery, but the bottom line is it's a response to this pressure that they need to find — corporations need to find efficiency gains and large language models will try and do it by saying, hey, we can do that hybrid or on-premises. And so that's going to be like the litany for the next 12 months.
Akash Pasicha 12:41 ↗
Right. Well Paul, I want to thank you for coming on. That is Paul Kedrosky, general partner at SK Ventures here on TI TV.
Lyft reported quarterly results this week. Revenue growth decelerated to 8%, down from 16% last quarter, but the company's free cash flow notched up 6%. I want to bring on CEO David Risher for a conversation. David, welcome to the show. It's great to have you here.
David Risher 13:03 ↗
It's great to be here.
Akash Pasicha 13:06 ↗
Okay, David, sorry. I got to ask you that one of the first lines in your release from last night was 'neither rain nor heat nor the gloom of night can slow business', which I didn't know that it seems to be a nod to the USPS, correct, postal system. The only thing was that the business did slow and the growth rate was about half of what it was last quarter. So, I mean, help me understand how you see this.
David Risher 13:30 ↗
Yeah. So we measure our business and run our business based on really two key metrics. The bookings growth and then EBITDA growth, and bookings actually were up 17% versus last year. So we actually...
Akash Pasicha 13:42 ↗
That was a deceleration as well, right?
David Risher 13:44 ↗
Well no, rides are accelerating quarter after quarter and bookings are accelerating quarter after quarter there. I don't actually think so. No, I think we've got an acceleration on the top and bottom line. But anyway, the real point is that like we've had all-time high riders, all-time active rides high. So, you know, quarter by quarter stuff kind of bounces around and there was a big one-time settlement that we had this last quarter which showed up in revenue to your point, but if you look at kind of the trend of our business, particularly where it was a couple of years ago, it is a completely different company, which is kind of an exciting place to be.
Akash Pasicha 14:15 ↗
Okay. So, look, I want to talk about some of the strategic initiatives that you announced. So, the most interesting was the Lyft Go subscription. And this is 10 bucks. You get DashPass, 8% cash back, kind of interesting. I mean, look, I think the broader question, we can dig into sort of the economics of this, but I was sort of thinking about how many subscriptions people already pay for. Why do you think they're going to pay for another one for Lyft here?
David Risher 14:46 ↗
Yeah. So, we designed, you know, the thing that our kind of focus has been over the last three and a half years or so since I've been here is customer obsession drives profitable growth. And you can see it in, again, how fast we're growing. You can see it in all-time high active driver hours and so forth, and so for riders and drivers really trying to focus on what they need. Okay. So, one of the things we've heard over the years is there's really only one subscription service out there for food delivery and ride share. And it's by the other guys, and DoorDash is a — and DashPass is a great product and Lyft is a great product. So, we decided it was the old — this is before your time, but there was an old commercial, 'two great tastes that go great together', Reese's Peanut Butter Cups. Super old school to be clear. Look it up, it's still — the ads still around. When we decided to sort of bring these two companies together, even though two separate companies, DoorDash and Lyft, and try to create something that we think is better than what's out there. And you're right, people have a lot of subscriptions already. You know, my hope — I think we've had about 500,000 people sign up just in the first 24 hours, which is pretty cool. So, what I hope is that people drop their other subscription to the other guys and pick this one up instead and get better value for it.
Akash Pasicha 15:51 ↗
Tell me, I mean, what gives you confidence that in this $10 — and there's a cheaper one I know, but let's stick with the $10 one for a second here. What sort of math have you done with respect to how much they are going to use your offering and your service compared to DoorDash's? And how does the incremental revenue from the membership then split between the two of you?
David Risher 16:20 ↗
Yeah. So, yeah, I probably won't go into the sort of the business, you know, kind of behind the scenes, who splits what revenue, but what I'll say is, you know, Lyft, if you look at our kind of, you know, what's powered our growth over the last three years, a lot of it has been partnerships, right? We have a partnership with United Airlines where you can earn miles on Lyft and burn miles on Lyft. We have a partnership with Alaska Airlines. We have a partnership with DoorDash that pre-existed this which gave you discounts on both platforms. And now this kind of evolves it to you can earn on Lyft and then spend on DoorDash or Lyft. 30% of our rides are tagged to partnership rides. So we've gotten really good over the years at partnering with other, in some cases very, very large companies much larger than we are, in the case of say United Airlines, to create things that benefit both companies and benefit riders or drivers depending. And I think that's a really sort of interesting almost an ethos that is going to work really well for us, for example, in the world of AVs where again we're going to be partnered with other companies much larger than we are in some cases but still trying to create, you know, economic benefit for both. And so far we've been pretty successful at that.
Akash Pasicha 17:27 ↗
You talked about partnerships so let's talk about Waymo which I saw it looks like you guys are leaning into your partnership with Waymo in Nashville. Yeah. This comes at a time where Uber — I mean there was an interview with the CEO of Waymo. Looks like Waymo is sort of winding down their relationship with Uber. Yeah. And so this is sort of an interesting tale of two ride-hail, ride-sharing companies here in terms of that. I mean, the issues around, you know, it being a bit of a scrappy or messy experience, you know, not knowing what vehicle you're going to get. That was something that they highlighted with the Uber-Waymo partnership. Why are you confident that you are not going to have those same issues?
David Risher 18:14 ↗
We're better partners. I mean it's just — and I do think this is it kind of...
Akash Pasicha 18:19 ↗
Why are you better partners?
David Risher 18:21 ↗
Sure. So I think a couple things. I think first of all we're very focused on the customer experience and, you know, I don't want to comment on those other guys but my general — so for example, I'll give you some stats and then kind of go back to the core of your question. Right now there are markets where we pick you up four to even six minutes faster than the other guys. We price very competitively against the other guys. We have almost a 50% preference. It's like 28% versus 56% driver preference. People who drive on both platforms, they prefer driving on us. Okay. Why would that be the case? It's because we start over and over again with what does the customer want and work backwards from that. Now, so for example, you talk about the relationship and some of the reasons that Waymo has given for that breakup, and part of it is they couldn't figure out a customer experience that really worked well. It was a little bit unpredictable and so forth. We've been in very, very deep conversations with Waymo for a long time on exactly these same issues and we think we've got a great path forward with them. Let me give you the evidence there. We run a big 80,000-square-foot depot in Nashville where we're actually servicing Waymo's cars and doing a really good job there. But we're also providing demand to them and it's working. And the reason I can say it's working is we're expanding the fleet by, you know, multiples. So, it's kind of one of those things where I think companies just have different ways of doing business and ours is kind of collaborative and I think generative, right?
Akash Pasicha 19:40 ↗
David, let me ask you one last question before you go. We saw the DoorDash-Instacart deal. Yeah. Has DoorDash come knocking on your door at all?
David Risher 19:50 ↗
I think they're pretty focused on delivery. I think both of us have a kind of similar philosophy, which is let's really do one thing really well and then partner with others to kind of create additional value, and I think that's what they're doing with Instacart, really trying to lean in the delivery side. You know, we've had acquisitions in Europe for ride share. We've had acquisitions around the world actually for ride share, and so I think both of us are going to kind of pursue those types of strategies and then work together to kind of create these collaborations like Lyft Go.
Akash Pasicha 20:15 ↗
Great. Well, David, I want to thank you for coming on. That is David Risher, the CEO of Lyft here on TI TV.
Dropbox announced it is acquiring cloud storage company Tigris Data for $350 million. The company also announced its third quarter results. Another roughly flat quarter with solid free cash flow margins around 40%. For more on all this, I want to bring on Akash Garg, co-CEO of Dropbox for more color. Akash, welcome to the show. It's great to have you back.
Akash Garg 20:45 ↗
Hello. Hello. Thank you for having me again.
Akash Pasicha 20:47 ↗
Great to have you back. So, you bought this company, Tigris, and you know, I was looking at the descriptions of the company. It's more of a developer storage company and you obviously are more consumer storage. The first thing that came to mind is you couldn't have just built this in-house. Why did you have to buy it?
Akash Garg 21:07 ↗
So, let's start with a little bit of context. So, Dropbox is almost celebrating our 20 year anniversary in a couple months. And that's a huge milestone. For the last 20 years, we've built a lot of incredible infrastructure. And that happened to be an amazing product for people. So basically, you have the ability to have a trusted content foundation. You're able to not just store things but securely share them, apply permissions. And what we found as we started rolling out our own AI capabilities that we have a structural advantage. People don't want to just use AI. They want to use AI where their content is. They want to be able to use AI in a trusted safe way. You want to be able to apply permissions and sharing. You don't typically work in a siloed environment. You work with people. And so what we solved for people over the last two decades, we realized there's an opportunity to solve it for agents as well. And as we started thinking more about it, we realized that what we've built over the last 20 years has an insane potential to enable other developers and AI companies to build on top of. And so as we started looking at this, we found Tigris to be an incredible match for us. And I'll explain why. Number one, they have actual thousands of customers who are developers, both AI builders and large known AI companies like Fal AI, for example. And so, not only do they have traction, they have a lot of experience building for AI developers. And this is a company that built their product from day one for agent use cases.
So, they have capabilities like forking and snapshotting. I'll give you what it translates to in simple terms. If you build agentic or AI workflows, you're going to get better speed, better latency, and you're going to be able to do things much faster and closer to your GPUs. So, essentially, they've built their product from the ground up in a way that is really compelling. And what we could do with them, the match here is we can help with taking this to more customers and we can also help bring the exabyte scale efficient infrastructure we built in our own data centers to that company. So the match was really great and it accelerates significantly the opportunity.
Akash Pasicha 23:30 ↗
So are you talking to like the Replit and the Lovables of the world about trying to make Tigris or I guess, you know, whatever you call it, DBX as it — DB Dropbox Cloud, is that what you're calling... yeah, so are you talking about them making it the de facto developer storage tool? I mean, is that the distribution strategy here?
Akash Garg 23:55 ↗
Yeah, the strategy is twofold. One is to enable a developer motion that people can come in, pick up an object storage API designed for agentic use cases. We're starting with object storage, but we're not stopping there. The things we built over the last 20 years will also be added as APIs. So, I'm going to give you examples. When you store content, you need to apply typically if you're building a product capabilities on top. So, who has access to that content? Can you share it? Is it secure? So there's a lot of governance on top of that content that we have today in Dropbox that will add as a set of APIs. We also have capabilities to edit and transform over 300 file types because over the last 20 years we built these capabilities. So how to process the content. So now your agents not only can just drop in stuff, it can actually govern it, it can process it. And so these are capabilities that we want to add. So part one of the motion is to enable developers in a self-served fashion to come in and pick up these APIs and use them, and part two is going after companies that are building at scale AI solutions and need to store things. You and I both know that more agents needs more content. And so we have a lot...
Akash Pasicha 25:12 ↗
So you are talking to all these AI coding companies about trying to make this the de facto tool. Sounds like...
Akash Garg 25:22 ↗
Yeah, as an API. Yes. And not just AI coding. We're talking about companies that, you know, you think of HeyGen, you think of Hex. There's a lot of companies that store a lot of content, need to process it.
Akash Pasicha 25:37 ↗
Right.
Akash Garg 25:37 ↗
So there's a lot of different audience for this. Again, from self-serve AI developers all the way to companies that have established products. Yes.
Akash Pasicha 25:46 ↗
Right. Okay. So, last time we had you on, we were talking about Dash, which is another AI tool that you guys are developing. This is, of course, sort of the tool that will help people make sense of all of the context, I guess, that lives within their gigabytes, terabytes of Dropbox files.
Akash Garg 26:08 ↗
Sometimes petabytes. Yes.
Akash Pasicha 26:10 ↗
Yeah. Yeah. Petabytes. I didn't know what came after terabytes, to be honest. So I asked you the question last time. I said, 'How are you thinking about monetizing it?' You said, 'Well, we're going to test a few different ways in the quarter.' And we are going to start monetizing it. And then they asked you on the earnings call, they said, 'What's the monetization strategy, where it's at?' And you didn't really answer the question. So, I mean, where is that at? You know, how much money is Dash making you, if any at all? When are we going to start charging for it?
Akash Garg 26:39 ↗
Okay. So, let me give you also a little bit of context so that because it's important. Dash was rolled into the main Dropbox experience. So, before it used to be a standalone experience that customer had to go to and interact with. What we did based on customer feedback, they want AI in context. Customers — I'll quote one customer that said something that I really loved and appreciated. They said, 'This is incredible to see AI coming to my content versus my content going to AI.' And so that's literally one unprompted customer feedback. So what we've done is we've put AI where people actually do work. So we've upgraded the actual Dropbox experience and it's the largest, I would say, most significant update we've done over the last many years. So it becomes very AI native. It actually understands your content. And you can action it. So, that's the biggest difference that we've done. It's not just about asking stuff about your files or finding them, although that's very important. You can actually action it. You can say, 'Prep me a brief on this campaign.' Not even describe what the campaign is because we have your files. It can take steps forward, put in briefs. It can edit PDFs, create a PDF presentation.
Akash Pasicha 27:56 ↗
But Akash, I want to know when it's going to raise overall revenue growth because this is — I mean this is the...
Akash Garg 28:03 ↗
I'll get to your part of the question, trust me.
So basically we did this over the last few months. We announced this as part of our strategy just last quarter and this quarter it's already in the hands of over 70,000 teams, and teams think of them as the small businesses. The goal is to roll it up to all English-speaking teams by end of year and roll it to all customers including people individual plans next year. The idea is we gave people a bunch of free credits in their accounts. And the goal of this is to have people use, learn, discover, and we actually have in the rollout experience — sorry, we actually have in the rolled out experience the ability to add and buy additional AI packs. So, you get a bunch of free credits. We believe they're enough for you to experience the value of bringing AI to your content and save you a significant amount of time for the mundane tasks. And the goal is over time what we want to see is these AI packs getting sold. And you know, it's too early — the reason why I didn't talk about it in earnings specifically is because we don't talk about speculative numbers. We want to be concrete about what's happening. And this is just a brand new experience that we launched. We do see people and teams buying AI packs. Until we are able to see like a few months of that trend, we talked about, you know, next quarter we'll talk more about the monetization. But this — yes, there is a monetization plan. Everybody gets a bunch of free AI credits and as they see the value they can buy more. That's what we have.
Akash Pasicha 29:48 ↗
Last question for you. So, I mean, look, you guys operate or collocate data center facilities around the world as well. How are your data center colleagues thinking about the new era for data center development that we might be in as a result of the elections?
Akash Garg 30:09 ↗
So I would say again we don't tend to speculate over what may or may not change, but what I can say is we have a really incredible set of people who work on our data centers. We have incredible supply chain relationships. And as of today, we actually have available capacity to continue to grow and we have the ability to continue to grow that within our existing footprints. And so we believe that we're actually in this regard really well positioned for the next couple of years to continue to grow without having these potential headwinds.
Akash Pasicha 30:49 ↗
Great. Well, Akash, I want to thank you for coming on. That is Akash Garg, the co-CEO of Dropbox here on TI TV.
This week, media and software company Every hosted its Thesis Conference with AI executives, founders, and researchers on stage. TI TV was on the ground at the event. We got to speak with executives from OpenAI, Microsoft, Runway, and Every's own Dan Shipper. We're going to play a couple of conversations that we had at the event for you. Now, we're going to have more coverage for you across our social media platforms. Let's take a look.
I'm here with Andrew Ambrosino, member of the technical staff at OpenAI. Andrew, welcome to the show. It's great to have you here.
Andrew Ambrosino 31:47 ↗
Thank you for having me.
Akash Pasicha 31:48 ↗
Okay, so you got to tell us, I mean, member of the technical staff. What is your purview of work inside this giant member of the technical stack?
Andrew Ambrosino 31:56 ↗
Yes, we've all got kind of the same title here. So I lead our workspace org, which is ChatGPT work. It is the Codex...
app and tools for development, and it is ChatGPT space as well. So that kind of group of how you use our application, our products, our models at work is what I mostly focus on. When I look, when I open ChatGPT, Codex, I've got work, I've got sites, I've got docs.
Paul Kedrosky 32:21 ↗
I mean, that's all under your purview?
Andrew Ambrosino 32:23 ↗
Yes, it is. Okay, yes, it is. So let's talk about how you're building Codex in a second, but I want to start with the news, which is that you guys made an acquisition, the Pierre Computer Company. Tell me a little bit about what it is that Pierre does and why you decided to buy it.
Paul Kedrosky 32:39 ↗
Yeah, we've known Pierre for a long time. We've worked with Pierre. The Codex app was one of the first to use some of Pierre's new technology on the front end. One of the most talented groups of people I've ever known in tech. And so they work on both code storage and libraries for code review. They're big in the open source community, and we're bringing them in to help us out with some code storage, some end-to-end SDLC of building software and what that means. And I couldn't be more excited.
Andrew Ambrosino 33:13 ↗
And I mean, help me understand when you think about the technical challenges that Pierre solves for you, what were those challenges?
Paul Kedrosky 33:24 ↗
Yeah. I think with these agents, more code is being written ever at speeds that we're not used to. And some fundamental assumptions about how you build these stacks for code storage do have to change, and code review as well. And I think we've seen some needs internally, both with Panda, which is our in-house source control, and...
Andrew Ambrosino 33:50 ↗
Source control means...?
Paul Kedrosky 33:51 ↗
Code storage, okay, very similar. And you know, what we're bringing on with Pierre, that in this agentic era, I mean, everybody's kind of rethinking a lot of these base pieces of the stack that we all build on, right? Even things that weren't software before are sort of becoming software. Like our finance team uses a lot of sites, for example, where in a previous era they might use Excel for something, and so that site has code that is running that site that they are using and sharing around, and people are doing this many, many times a day or an hour, right? And so the way that we store code performantly and scalably is more important than it really has ever been.
Andrew Ambrosino 34:28 ↗
So when people think of code storage, I mean, is this like a competitor to GitHub? Are you still using GitHub? Just help us understand that a little bit.
Paul Kedrosky 34:38 ↗
Yeah, GitHub is the biggest name in code storage. We've been a customer for a long time. We still are a customer. We have our open source repos on GitHub. We interact with the community there, Codex being one of those. GitHub is a great partner. I think what we see internally is just needs that we want to kind of play with, experiment with, and see what we can do here.
Andrew Ambrosino 35:02 ↗
Right. So the Pierre acquisition, again, the conversation around technical challenges, there's so much that's difficult here in what you're building. I wonder if you can talk about some of the other technical challenges that you're seeing, not because things are going wrong, just things that are really difficult to solve right now. I mean, the conversation that I've had with Dan Shipper many times on my show is what's difficult for the models to do right now. And so in your capacity, it's kind of interesting. You're building this program that the front end is changing, the back end is changing. Tell me about what's hard about it right now.
Paul Kedrosky 35:41 ↗
Yeah, the job has changed a lot, right? And I think as we've looked at the, you know, we started about a year ago, we started building the Codex app, the original version of this, and the models were very, very different at the time. And so we've kind of seen some of these cycles of how software gets developed and how the models are, you know, bad at something until they're not. And I think, you know, if you look at things like plan mode and goal mode and these kind of featurettes that we've launched for software developers, some of them are really meant to be temporary. Some of them are meant to say, hey, you know, in April, the model is good at a lot of this stuff, but it doesn't keep itself going for a long time, right? So we introduce goal mode, and goal mode is kind of a more deterministic, more structured feature that says, hey, you know, we're going to prompt the model to keep running until it hits this goal, right? With newer models, like with Astra, goal mode's almost not needed anymore, right? Because it can keep going. But it's kind of always these arcs of things where we're like, hey, we're going to introduce this like little helper here in the software, and at some point we'll be able to retire it because the model will be able to do that kind of natively, right? I think where we still see the models struggle overall across the industry is just like what does it mean to maintain a product, a large product, and to take in feedback from everybody and to make decisions about what features to prioritize and what not to, and how to design them. And even though the models are writing most of our code, still, you know, this aspect of things is kind of the challenging part.

67 more exchanges in this transcript

Sign in free to read the rest of this interview. No card required.

Sign in to read the full transcript

Cite this transcript

APA, MLA, BibTeX
APA

Kedrosky, P. (2026, November 6). Nvidia's Data Center Threat, AI Stock Rebound, OpenAI's Codex Strategy [Interview transcript]. The Information. CEOInterviews.AI. https://ceointerviews.ai/interview/3208130/

MLA

Paul Kedrosky. "Nvidia's Data Center Threat, AI Stock Rebound, OpenAI's Codex Strategy." The Information, 6 Nov. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/3208130/.

BibTeX
@misc{kedrosky2026_3208130,
  author       = {Paul Kedrosky},
  title        = {Nvidia's Data Center Threat, AI Stock Rebound, OpenAI's Codex Strategy},
  howpublished = {Interview transcript, The Information. CEOInterviews.AI},
  year         = {2026},
  month        = {nov},
  url          = {https://ceointerviews.ai/interview/3208130/},
  note         = {Speaker-attributed transcript with timestamps}
}