← AI to ROI
AI to ROI
AI to ROI: OpenAI - The Most Important AI Company in the World, and the Most Fragile
00:38:48
Loading summary
Transcript73 lines
- [00:00]
A
Foreign.
- [00:08]
B
Hey, welcome to AI to roi, the Big Story edition. I am Peter Buchanan, the managing partner of New Plan, and I am joined by the maestro of metrics, Ray Wright, the CEO of Benchmarket. Ray. It's great to be with you, Peter.
- [00:23]
A
It's always great to do this. And I, I'm really interested and excited about today's story because we're going to dive deep into the open AI business and I think it's an important one. And it's important to an industry that's still in the early stages and probably a company that could have the most potential impact to our listeners over the next 12 to 24 months.
- [00:47]
B
Oh, no doubt. So they're the most important company in AI history by far. They're valued at over $850 billion. And they're also in serious trouble on multiple fronts. They had these wondrous things going on, including the model Release and the Image 2 release this week and the agents release this week. And yet they have lots of storm clouds too.
- [01:13]
A
Yeah, and I think this goes beyond noise for our listening audience, Peter, because the executives here are making AI use and deployment decisions that strategically critical to to their business. And I think it's important to understand OpenAI in context of the entire landscape. You know, Gemini with Google and Anthropic with Claude, etc, so they make the right decisions now that could have real impact in the future. Remember when cloud costs first started surprising people? Finance would open the AWS invoice. No one could explain what drove it. And Engineering would spend a week building a spreadsheet to explain it. And it was still wrong. AI is doing that again, except faster, more dynamic, and spread across more systems and departments. Today, a single Enterprise has multiple AI costs running across multiple vendors such as AWS, inference costs on Anthropic and OpenAI, GitHub, Copilot, Cursor, and a handful of AI agents that no one billing system sees the entire spend picture. So Finance asks, what did we spend on AI this month? And the answer takes three days to find out and it's still probably wrong. And what did we get for it? Most don't even try to answer that one. That is a problem Maverick was built for. Maverick gives Finance, IT and engineering a single source of truth for AI spend so you can allocate costs, enforce budgets, and connect investments to business outcomes. Learn more at Maverick AI. That's M A V V R I K AI now on to the show.
- [02:51]
B
Right, so what we're going to do today in our very ambitious podcast episode is we are going to go through and we're going to look at six documented challenges that OpenAI faces and we're going to look at their IPO prospects and then we have six recommended specific actions that you and I think could reach really improve their prospects and performance going forward. And so I think let's just dive in. Ray, get started.
- [03:21]
A
Okay, so the first thing we talked about discussing today is, you know, why should OpenAI be winning this race? And generative AI. So you know me, I'm the numbers guy. So let's put some numbers behind this. So they're annualized recurring revenue. Notice I say annualized because for all the listening audience out there, the little trick I've seen is every time you see, you know, 19 billion from anthropic and then 30 billion, it's always an annualized run rate where they take the last month and they multiply it by 12. Still impressive, but just so everyone knows that.
- [03:56]
B
But it's not a gap financial statement.
- [03:59]
A
It is not GAAP. It's more like funky ARR calculations. But you know, they stand about $25 billion run rate. So that's pretty amazing because if you think about it, most people are learned about OpenAI late November of 2022 when ChatGPT was launched. So that's a pretty incredible scaling and run rate. Next, they have almost 910 million weekly active users and over a million businesses run on OpenAI. So this is a consequential company in a very short time, Peter?
- [04:37]
B
Yeah, just three and a half years really. I mean they were sort of a quiet lab before that, before that release. And the other way to really think about them is Microsoft, Amazon, Nvidia, Oracle, all these run by these revered CEOs. They've all signed nine and 10 figure partnerships and invested tens of billions of dollars in OpenAI. So there's something to believe in here. And when OpenAI ships a major product, work across the industry stops. Like buyers, partners, competitors, they all dive in to see what they actually release. That's happened in the last couple days.
- [05:20]
A
Yeah. So today's discussion almost, you could say, stands in a stark contrast to this amazing out of the gate start they got like we talked about in three and a half years. But this is not a stretch to say OpenAI has been the defining company in the most technology market in history. This is about the serious questions that you and I have and I think a lot of people in the industry have because there's some real concerns. So we broke this down into six different challenges we see. So let's start with the first one, if that's okay.
- [05:56]
B
Sure. So, so we have these six, these six challenges. We're going to cover them all, but let's, let's get started with their technology moat.
- [06:07]
A
Okay, well, let's say maybe the absence of one, Peter, because I was actually at an event in San Francisco last week visiting back, visiting home for 30 years and one of my favorite analysts speaking and I had a chance to meet him and his name's Benedict Evans, everyone should follow him. And he recently published an analysis and the first thing he said was that the six companies that are really leading the large language models, they're really fairly equivalent capabilities today. Right. And what he pointed out was there was no mechanism or no, you know, like a network effect or proprietary data advantage that lets one company decisively pull ahead. You know, models continue to leapfrog each other every few weeks. You know, we get excited about the latest announcement and then six weeks later the competitive advantage is non existent.
- [07:06]
B
Right. The model's not the moat. So the single most important sentence in the competitive analysis in the market right now, you have to have a different moat.
- [07:14]
A
Yeah. And it shows up and I call this momentum. Right. And user numbers is a great sign of momentum. And I remember I started doing research on companies that were using large language models and two and a half years ago it was like 80 some percent were using chat GPT. Right. And then it dropped to 74% last year and right now it's, I think it's about 60% so significant decrease at the same time and accordingly. So Claude has grown their user base by over 14%, Gemini by 12% while chat GPT only grew by 4%. So even though the overall market continues to increase, which is a Good sign, chat GPT's market share continues to decrease and decrease fairly significantly from my perspective.
- [08:15]
B
Right. And it's not just customer loyalty, it's you can have weekly active users, but if 80% of your weekly active users sent less than a thousand messages per year, that means they're sending three messages, three prompts per day through OpenAI. So they're kind of light users. Like they, they use Google way more,
- [08:39]
A
you know, so, and let's be really candid, Chat GPT is or has been primarily a consumer application almost the way Google started for their first few years. Right?
- [08:53]
B
Sure.
- [08:53]
A
And for me, I'm a B2B enterprise guy and a lot of people, a lot of people a lot smarter than me are saying that's where the real stakes are. And you know, we've heard about Anthropic and their amazing growth and they anchored upon B2B and Enterprises almost three years ago. So at OpenAI they're trying to make the pivot to be considered more of an enterprise class player. I believe that their enterprise sales now account for about 40% of the revenue. And what's amazing about that is that's doubled from 20% a year ago. And their goal is to be a 50, 50 company. 50% enterprise, 50, 50% consumer and they have some real wins. But at the same time, when I think of enterprise AI, I think Anthropic.
- [09:44]
B
Oh absolutely. And data from Ramp shows that Anthropic's models are preferred by 73% of enterprise buyers as their first purchase. So that's a hard, that's a hard number for OpenAI's enterprise sales team to overcome. And so the good news is most of these companies, they buy multiple products and they're trying out multiple things at the same time. But that is a pretty big hurdle to overcome because Google and AI is kind of an enterprise company too. And there are a few others that kind of can get in the way.
- [10:23]
A
And Peter, I think Google just announced their new agents product which is being integrated into Google Enterprise. Right. So I think Google's taking the enterprise very seriously also.
- [10:34]
B
They OpenAI basically the same day announced a new agents product which confusingly are called workspace agents. So I think Google's lawyers would be
- [10:44]
A
calling now the other thing that I think is a challenge and I think part of this is my own bias. And we grew up in a company, Peter. Executive leadership was key to long term market success. And I'm talking about when Jack Welch was the model for CEOs right back in the late 80s and 90s. But I'm actually a little concerned about Sam Altman of the captain of this OpenAI ship. Over the next couple years and a couple great reports have come out the New Yorker, you know, they published this 16,000 word profile drawing from over 100 interviews and undisclosed information and documents and it really raised some serious questions about his integrity and management behavior. In fact, I didn't do a word count but the word lie or lies was pretty prominent there.
- [11:41]
B
Yeah, it was all over the place.
- [11:42]
A
And then Wall Street Journal follow with an expose on Altman's personal finances kind of documenting how his private investments has created potential conflicts with OpenAI's business decisions. So I always try to take a pretty neutral, objective and quite frankly data driven perspective on things happening in the industry. But the data and reporting that I'm reading doesn't feel right here.
- [12:09]
B
Right. So he has over about 400 private investments in tech companies because before OpenAI he ran y Combinator, so he met all these companies. But he has almost no shares in OpenAI and he's the CEO of the most important AI company in the world. It's very odd. So let's move on here because the next thing you go to is the partner ecosystem. So in a February investor presentation, when they were trying to raise this Money at this $852 billion valuation, OpenAI stated that in the future they're going to create products that will replace software from Salesforce, Workday, Adobe, slack, Atlassian. And OpenAI has revenue generating partnerships with most of those companies, in fact all of those companies. So that's not a partnership strategy, that's like a warning shot that says maybe you can't trust this model partner and you ought to be looking for some different model partners if you're one of these SaaS companies pivoting to AI.
- [13:16]
A
Yeah. You know, Peter, I think this is a challenge right now for the entire large language model ecosystem, the top five players, because Claude also sometimes introduces new products that you could say, compete with their potential partners. But I hearken back to something else that Benedict Evans said in his February article, and he says that OpenAI specifically has not articulated product boundaries. That gives their partners a credible kind of durable business case. So there are a little concern. Now, meanwhile, even though I said anthropic, I think has some similar challenges. They have invested $100 million, which by the way, I say $100 million today, Peter, and it just doesn't seem like much.
- [14:05]
B
Right, I know, I know, I'd like to have a hundred million dollars, that'd be good.
- [14:09]
A
But they invested 100 million in their clan partner network to really attract product developers and integrators. And to me, just that announcement, backed up by a significant amount of money, says that partner ecosystem is important to OpenAI, that that middleware layer of software is important to them, and it hasn't clearly been staked out. So even though they may say or be thinking, should I compete here or not, simultaneously they're really investing in that partner ecosystem, especially with the software developers.
- [14:46]
B
Right. So every systems integrator and every enterprise software company, they're just sitting around in their boardrooms right now and they're trying to figure out which model providers to bet on or whether they should bet on all of them. And if you're telling those partners you're going to compete with Them, you're probably, those companies are probably going to make a different bet because unlike consumers switching chatbots, I mean, you and I both have three of them, we move around all the time. Enterprise partners can't switch back all that easily once they've made a commitment because the partners are implementing these things on behalf of their customers and they just can't switch them. So given all that, let's move on to the financial picture at OpenAI because, wow, there's, there's a lot to unpack there.
- [15:36]
A
Yeah. Right. Before we dig into the financial picture a little bit more, you know, for our enterprise listeners out there, if you're building your application to be very dependent on LLMs, I think the best strategy is not to only have one, have two or three, have a model orchestration capability. Often orchestration is used to get the right cost model or the right model for the right job. But honestly, I think it's a survival strategy also where if one of the model vendors ends up announcing a product to compete against me, I've got a second and even a tertiary choice. So just something I was thinking about, what you think about that, Peter?
- [16:22]
B
Well, companies are doing it and people build, you know, customers building complex use cases are using sometimes three and four models inside of their workloads, depending on the type of work that needs to get done. They might use CLAUDE for some heavy lifting data element, but they might use an open source model for a basic element with data that doesn't connect to the Internet. So I think that's happening in general so that they. The problem is really for the application vendors that run their applications, they have to run their applications on top of these models.
- [17:05]
A
You're right. Okay, let me answer though, the question you mentioned, and that's a little bit more about the financials. I think this is kind of an IPO problem because there's been talk about OpenAI and anthropic and SpaceX going public in the next few months or by the end of the year. So OpenAI's the revenue story is incredible. Right. We've talked about it, but you know, 13.1 billion last year, which was a 3.5x growth over 2024, you know, projected over 30 billion this year. I think they just announced that $25 billion run rate. So from a revenue perspective. Notice I didn't say financial perspective, Peter. It's revenue. That's pretty extraordinary growth.
- [17:48]
B
Oh, absolutely. But the cost structure is what's so scary here. OpenAI is committed to spending over $600 billion on infrastructure over the next five years, it used to be $1.4 trillion and somehow they saved $800 billion. But still huge number, their gross margins fell to 33% in 2025. They had a 46% forecast. That's basically because inference costs ate them alive. Like it ate everybody alive, frankly, last year. And then so they project burning $100 billion in cash before they reach cash flow positive, which isn't going to be till the end of the decade. So you got to have a lot of faith.
- [18:38]
A
Well, Peter, I don't even know if I am completely sold on that because they just raised 122 billion. Right. In private funding.
- [18:47]
B
Sure.
- [18:48]
A
It feels to me like OpenAI actually needs additional capital. Right. And that's one of the reasons they're considering going public maybe before they're ready. And then when I see their cfo, Sarah Fryer, who has had a couple interviews where she talked a little bit about the financials and the need to maybe have a little bit more of a profitable operating model. Right. She got demoted because I know she's also told colleagues privately that the company's just not ready to go and IPO. And there, I think, you know, $852 billion valuation, you know, that's about what, 25 to 26 times 20, 26 revenue? 30 billion. Right. So maybe it's like 28. With 33% gross margins and a projected 14 billion loss. That doesn't tell me it's a company ready to go public. And one of the outlets, I know you and I both follow the information, you know, they recently interviewed 11 institutional investors about a potential OpenAI IPO. And most of them, even though they don't own any open API, set out right that they would not buy the stock. In fact, they might short it.
- [20:04]
B
Right. That's not a vote of confidence. Plus, Sarah Fryer actually has an absolutely sterling resume as a CFO leading up to this. So you don't demote a star because they give you bad news. There are also legal risks that don't show up in the financial model. So next week is the beginning of the Elon Musk vs OpenAI trial. There's going to be a whole bunch of dirt dished during that trial that neither side wants that public, but it's going to happen anyway. You've got content and so elon suing for $109 billion. So if he wins that suit, that could be incredibly damaging. The second piece is the New York Times and a slew of authors also have content related lawsuits. Against them, which could create billions of dollars in damage. And we mentioned Sarah Fryer being demoted. Well, she reports now to the chief AI officer, in essence, the number two person in the company, Fiji Simo, who is currently on medical leave because she has an autoimmune disease and she's not feeling well. These are challenges, but there are things that OpenAI can do to really, I think, right this ship.
- [21:24]
A
In fact, Peter, if I'm not mistaken, and I know I'm not because I read them just an hour ago.
- [21:29]
B
Right.
- [21:30]
A
In the newsletter that we published on Monday, April 20th, kind of. There's six, at least six actions that they can take to write to ship. You want to start with the first one?
- [21:40]
B
Sure. So the first action is really the most. The most urgent, actually, and that's lead with enterprise customer evidence. So when you look at what OpenAI, the coverage of OpenAI from Axios, the Information, the Verge, CNBC, Reuters, Financial Times, all the places that are really big with wide circulation, the stories about them are about internal memos, lawsuits, leadership friction, and product announcements. But it's not about successful enterprise use cases. Anthropic is absolutely great at that. So you don't want to publish drama, you want to publish success.
- [22:26]
A
Yeah. And, you know, most days I have the chance to speak with a lot of CFOs, including some enterprise CFOs and even CEOs, and they're facing increasing pressure from their boards to start showing the return on investment from their AI investments. So one of the things that might be really interesting is for OpenAI, we know they have a pretty good PR group. So let's put together kind of a customer evidence PR program with name deployments, quantifiable outcomes wherever possible, and even a spotlight on their integrator partner ecosystem who help those companies succeed and really start showing those use cases. Now, it's hard right now to get enterprises to really open to kimono and talk a lot about it because the benefits often aren't hitting the income statements yet. And because of that, it can make investors a little weary. But I think there needs to be a very structured PR program about their enterprise case studies and their wins.
- [23:31]
B
Right. So the second thing here is there's so much sniping, especially between OpenAI and Anthropic, and most of the sniping is actually from OpenAI at Anthropic, although not wholly, but every leaked attack memo, every public pronouncement. Sam made one about Mythos the other day. I just don't think customers want to hear it. I Mean, they really want to hear about product features, reliability, case studies, integrity. They just don't care about a corporate feud when it comes to making purchasing decisions. So they might want to read about it and eat popcorn and oh well, here's another adventure feud adventure. But this isn't reality tv. This is real things for real businesses.
- [24:19]
A
And you know, we, we touched upon this earlier, you know, with a couple of the articles that have come out. Sam Altman's integrity is already at least a caution, right, in being questioned. And as you know, Martin Piers at the Information said they need to stop issuing numbers that continue to undermine their own credibility when you miss them. And OpenAI does have a pattern of this, right, where they make projections that they later miss or they walk back. And a lot of people would say, yeah, Ray, come on. Elon Musk does the same thing. He always over commits, right? Like the Starship launch, which is five years late, or the, the magnitude of the Robotaxi rollout. Sam Altman is not Elon Musk. Elon Musk has made a lot of people a lot of real money. That's actually cash that can be invested in buy things versus making money on paper. So I think this credibility issue needs to lead to less discussions about financial and product projections.
- [25:21]
B
Right. They need to be quieter for sure on that front. And the third thing they need to do is they need to invest in safety as a business asset, not a PR exercise. So the New Yorker article documented a decade long problem of making public safety commitments and then quietly abandoning them. There's an organization called the Future of Life institute that gave OpenAI an F on existential safety. What a word. And that's the same grade they gave every other major lab except Anthropic, which got a D. So obviously they have very high standards. But safety is a competitive differentiator. I mean, the reason the Department of Defense chose Anthropic as their CLAUDE as their first model to deploy in the summer of 2025 is that they had a much more robust approach to safety.
- [26:17]
A
Yeah, let's be honest. Enterprise buyers, especially in heavily regulated industries like healthcare or finance and the government, they need that auditability and those governance, safety, best practices. And you know, I think the recommendation is pretty simple and specific. Maybe a well funded independent AI governance and safety board that has veto authority over novel risk product launches would be a really good idea with these quarterly reports on safety evaluations of the models and maybe even more money in research about how enterprises are rolling things out and how they're ensuring safety and Governance. So I think that could be a real powerful thing for OpenAI to do.
- [27:11]
B
I also think, I mean OpenAI has a powerful new cyber product that sort of in the same category as Claude Mythos, maybe not quite as scary, but customers are going to buy more than one of those. They're also going to buy the Google product from Wiz. I think treating it also more as an industry imperative that these products build on top of each other. While we think ours is better, we know there's going to be multiple ones. The other three actions that we recommend are, they're equally important, but let's go through them more quickly. So in communications they need really IPO grade discipline. They have to really stamp down on these leaks. They can't have their CRO's internal memo ending up on the front page of the Journal of the Information. They can't have premature product discussions. They can't have off the record access. So major leaks need to be treated as firing offenses.
- [28:19]
A
I agree. I still remember back in the 90s when I was at Netscape, I went through quite a bit of communications training and I thought it was going to be more about how to communicate things and how to position our product announcements and more of it was about what not to say and what not to do when you're talking to the press.
- [28:40]
B
Do you remember our GE rule?
- [28:42]
A
No, what was that?
- [28:43]
B
Our GE rule was I never want to end up on the front page of the Wall Street Journal or New York Times with a bad article and I didn't know about it in advance. That was the Welch rule. And when I would, because I worked in strategy and I worked a lot with the PR people and boy, that was just enforced on us all over the place. And it's a really good role.
- [29:08]
A
One of the other things kind of as we go through these six and you know, we touched upon earlier also is the importance of that partner ecosystem. And I still remember, I think there's a quote that's associated with Bill Gates where he defined a platform as something that creates more value for their partners than for itself. Well, I look at OpenAI's investor presentation recently and it said it could replace or maybe even would replace Salesforce, Workday and Slack. Right. These aren't really compatible positions. Building a great partner network and saying you're going to replace companies that could be leveraging your product. So I think every Systems integrator and AI native software company is evaluating whether partnering with OpenAI is a safe long term bet. And if they don't have billions of Dollars to hedge their risk. I think this could say maybe I pull back from OpenAI.
- [30:05]
B
They should listen to Bill Gates because what's the most successful partner program in the history of the world? It's Microsoft with their 650,000 partners. Finally, let's look at how they need to operate as a company. Because you always hear about startups, startups, this is a hugely successful startup. But when you're 30 billion in revenue, you're not a startup anymore. Anymore. You need to be a mature growth company. That means that customers, partners, regulators, investors, they all want the same thing. They want high integrity leadership, they want predictable growth, they want laser like focus on customer success. They want strong governance and financial controls. And OpenAI has some of these things, but the things they don't have, they're far away from. So if you're an executive sitting in a room who's already bet their AI strategy on OpenAI, Ray, what do you want to see?
- [31:07]
A
Well, Peter, I think that partner ecosystem is so critical and the Microsoft relationship is probably the thing I would tell people to watch the most. Do you agree with that?
- [31:20]
B
I do. Although there's a lot of high stakes, fuzzy relationships between all these enormous companies. But the Microsoft one is probably the top one. The Microsoft, it's a three way thing. It's Microsoft versus OpenAI and it's OpenAI and Amazon because Amazon got some rights that probably should have also been available to Microsoft in their last agreement with OpenAI and that's causing huge contention and it really affects the way customers are going to build their applications in the future. It needs to be resolved.
- [31:57]
A
Yeah. I also think watching the IPO process is going to be extremely instructive. Right. If Sarah Fryer exits, that's going to tell you something really important. If the OpenAI IPO chatter says I'm pushing it to 2027, I think that's going to be extremely instructive. That tells me the institutional investors don't think they're ready. And then I would start watching like their Frontier alliance partners and how many enterprise case studies I really see coming out. Because those three things are going to tell us a lot about the company's strength of positioning and their financial readiness to go public.
- [32:32]
B
Right. And as an enterprise, as an enterprise chief AI officer, you don't want single model dependencies as we talked about before. So no enterprise strategy should rest on one vendor. So anthropic Google others, they're all shipping capable models. You want to think about API portability and you want to think about good architecture. And you want to think about how easy it is to move from one to the other if you have to.
- [33:02]
A
Another thing, I thought a lot about this and I think about all the noise around Elon Musk and how many executives he gets rid of and leaves Tesla. But I still think senior leadership in this AI category is, is still critical. So I think, you know, keeping a close eye on what happens to senior leadership and some of the top AI development talent over the next six to 12 months, you know, whether it's leaving in OpenAI or coming in open AI, I think that's going to be a good tell.
- [33:33]
B
It is. And you know, we have a special situation in the United States where we've got about 20 to 25,000 elite AI engineers. It's not a big number. They're concentrated in a relatively low number of model and AI application companies and in China they have a million. So that talent is really important. The bottom line for enterprise leaders, Ray, OpenAI is a bet worth making, but it's not a bet worth making exclusively. So you should require the same evidence from them as you would from any other technology vendor. And that's the case studies, verifiable outcomes, clear product roadmap, accountability. So don't let the scale and the brand overwhelm allow you to shortcut your work. Really do your diligence for what you want to accomplish.
- [34:34]
A
Let's wrap up here. Peter. You know, at the end of the day, OpenAI is an extraordinary story, it really is, but it's fragile at the same time. And I think back to the history of technology in my own experience, you know, I was in the middle of a category creator called Netscape, the Internet browser. And I was in, right in at the executive table that was talking about the strategic direction of Netscape and, and specifically around their browser. And as we all know, Netscape didn't win. We had a competitor that had unfair distribution advantage. Right, Microsoft and operating systems, but they're not alone. Lotus 1, 2, 3. Right. They were an early winner, they lost to Excel and then BlackBerry was killed by Apple on the iPhone. So. And by the way, I just watched a movie the other night called BlackBerry and, and it was based upon a book called Losing a Signal. The Unlocked Story behind the Extraordinary Rise and fall of BlackBerry. It's a must watch because what it does, Peter, is it shows just because you're an early innovator and category leader, it doesn't mean you're the long term winner.
- [35:48]
B
Right? We had that phrase the other week. Pioneers get the arrows and Settlers get the land and everybody has to figure out how to be a settler here. So OpenAI could really be a settler. If they figure out leadership, communications, partner, ecosystem, financial discipline, they could be. Right now, they are the defining technology company of the AI generation, but there's no guarantee that they're going to stay there. So enterprise leaders just really need to pay close attention. That's all I can say.
- [36:23]
A
Well, I think we're going to wrap up. I was going to ask you a really controversial question, but it was unfair.
- [36:28]
B
So let's just say now you've got my interest peaked.
- [36:31]
A
Okay, so if someone gave you a million dollars and said you can only invest in one large language model company, where would you invest? We are not an investment advice people. This is just to solidify our perspective on what's going on and specifically around OpenAI versus the competitive landscape.
- [36:51]
B
Google, because they do everything. So what I am noticing from Google is they had this big sprint to get to Gemini 3, which was absolutely transformative. And then they went, and then they went quiet like they were tired for the last few months. And so they fell behind in coding. They just did their big agentic announcement. And so Sundar needs to keep the, the pedal to the metal with that DeepMind group. But they have the most advantages and also they control their own destiny because they have the best access to cash and a very durable stock.
- [37:37]
A
So Google, by the way, it's unfair if I don't give my pick. My pick is Tropic. And, and the reason I pick Anthropic is not because necessarily they're better than Google. They're starting at a much lower base of revenue. I think the investment upside is bigger with an Anthropic than it is with Google. So it's purely my greed that would say Anthropic, right?
- [38:00]
B
Yeah, you like to be an investor. So what I would say. The other thing that's interesting about Anthropic is everybody seems to like them. So Microsoft uses them in Copilot and hosts them. Amazon's put their. They've put $13 billion into them so far in investment. Could put another 20. They're building specialized data centers just for Anthropic. Google has built an inference environment for them that's very high performance. And so they're the only company that they don't create among other companies.
- [38:38]
A
Peter, Peter, Peter. You know, we should end the episode because we usually agree. I have to disagree with that because I think they pissed off one of the most important entities in the world and that is the United States dod. Right. And I think that's a pretty big cautionary tale.
- [39:00]
B
It is a cautionary tale, but there's people on both sides of the table there, and that's a whole episode.
- [39:07]
A
Okay. We're not going to get into politics here. No investment advice. No political opinions.
- [39:12]
B
Minimal politics. Yes.
- [39:14]
A
So, hey, thank you so much for our listeners. If you want to read the details of this that we published on Monday, April 20th, go to AI to ROI. That's AI, the number two, roi.substack.com Go ahead and read it and let us know what you think. Either make a comment on substack or, hey, if you're listening to this podcast, make a comment right here on your favorite podcast app. Thank you, everyone. Thanks, Peter.