The AI Race: Who's Ahead, Anthropic or OpenAI?

0:00 Scott: You need the marketplace to really drive the winners.

0:03 Based on what I read,

0:04 I think Anthropics is going to be a winner.

0:22 Hello, everybody,

0:23 and welcome back.

0:25 To another episode of How Many CTOs Does It Take?

0:28 My name is Scott Porad,

0:29 and I am here with my friendly co-host.

0:31 Brad: Brad Hefta-Gaub.

0:33 Scott: And Brad,

0:34 today we have a little AI business check-in.

0:40 Brad: Yeah.

0:41 Scott: I saw an article in the Wall Street Journal this week.

0:45 called Anthropik is on track to turn a profit much faster than open AI.

0:55 Financial documents from both companies show the different approaches they are taking to the AI boom.

1:02 And if I just look in here, I just read a few key facts.

1:08 Anthropic has really focused on business users because of the capabilities of their Claude chatbot encoding and actually expects to break even in 2028.

1:21 OpenAI,

1:22 goodness,

1:25 and this is just mind-blowing, they expect their losses to swell to $74 billion.

1:32 in 2028 and doesn't expect to be profitable until 2030 and expects to burn through 14 times as much cash.

1:46 Crazy sauce.

1:49 They're just taking different approaches.

1:51 AI expects thinner margins.

1:54 It is investing more in things like chips and technology.

1:59 It just has a more aggressive approach,

2:01 a more risky approach.

2:05 Let's see here, just scrolling through,

2:07 what does it say about Anthropic strategy?

2:10 Anthropic is taking a more cautious approach with costs growing more at a pace with revenue.

2:15 They're focused on increasing sales among corporate customers and avoiding,

2:19 this is key,

2:21 avoiding getting into image and video generation, which is much more expensive in terms of a cost.

2:26 Brad: Interesting.

2:27 Scott: So very interesting.

2:30 Brad: Yeah.

2:30 Scott: Any thoughts on that?

2:32 Brad: Well,

2:33 yeah,

2:34 I have a couple thoughts on that.

2:36 First of all, my first thought is that this is just a, I mean, the battle, the battle that's going on here for AI is really,

2:49 really interesting. And anthropic and open AI are the two small fish, right?

2:54 And,

2:54 Scott: Mm-hmm.

2:55 Brad: you know, you got the big players of Meta, Microsoft and Google that are really aggressively out there.

3:02 And it feels to me a little bit like anthropic is,

3:11 I think it's fair to say they appear to be taking the more conservative approach.

3:16 Scott: Mm-hmm.

3:17 Brad: And I think you're right that kind of open AI is kind of just trying to swing for the fences, get big fast,

3:24 you know, scare people by, you know, kind of trying to act like they're the biggest possible player in the space and try to cover as much flood the zone kind of thing.

3:37 And,

3:37 you know, the the stakes are high enough that.

3:42 And Sam Altman is a good enough salesperson that, you know, they're, you know, he's making the way of it. But from a sort of like the business, do the numbers pencil out like things have got to get gigantic. Right.

3:57 So a good example of this is and.

4:02 You know, we record this when we record it and it gets released a little bit later.

4:06 So any of the numbers that we're throwing out now are going to be, you know, could be old news by the time that this episode comes out. But at the same time, it's on trend, so it kind of doesn't matter. But, you know, the recent,

4:18 you know, Sam Altman's been, you know, tweeting that,

4:21 you know, they expect to hit a $20 billion forward annual run rate by the end of this year.

4:27 here but at the same time they've got 1.4 trillion dollars of you know capital commits right so it's just like you can't make 1.4 trillion dollars of capital commits pencil out on a you can't even make that pencil out on a hundred billion dollar a year business you gotta you gotta get well well north of that and so they're just it's huge numbers we're talking about and it's huge amounts of

4:52 you know, I'd say risk because the reward is so great.

4:57 So that's my first thought is that I100% agree that OpenAI appears to be taking a much more swings for the fences, everything,

5:06 everything AI approach versus Anthropic, which really seems to be focused on business users,

5:12 especially coding.

5:14 So that

5:14 Scott: Mm-hmm.

5:14 Brad: I think is accurate statement.

5:16 And I didn't know about the video and image generation,

5:18 but it's totally it totally syncs with what I happen to know as a user of the two.

5:23 But I hadn't even thought about,

5:24 oh, that's right there.

5:26 Anthropics not really doing image and video generation.

5:30 Here's another really interesting piece of news recently that is very related to what you just said.

5:39 Cursor, another independent company,

5:41 which embeds all the models,

5:44 right?

5:45 So you can hit Quad,

5:47 you can hit OpenAI,

5:50 Gemini.

5:52 They just released in the last week or so their own model,

5:57 right?

5:59 So they released a model.

6:00 But you know what?

6:01 It's not a foundation model that they built.

6:05 It's a repackaging of a Chinese open source model and they packaged it as their default model and they believe that it's outperforms Claude and Claude was a far, far,

6:18 it was far outperforming on coding tasks.

6:22 This, I think, is another interesting play where it's like,

6:24 OK,

6:25 so cursor is very much a targeted application for software developers.

6:30 That's what their focus is.

6:31 They're not a foundation model.

6:32 They're not trying to build models.

6:33 They're just trying to build this this world class application for developers to use.

6:38 And even they are picking and choosing which of these foundation models to use and going with an open source model.

6:48 uh was a bold and interesting move on their part um so there's just a lot of churn happening in these uh in this space and i i think that i think that that move by cursor is very telling about anthropic right like if

7:09 Cursors not like if cursor can can use this open source model and outperform Claude that puts that puts anthropics business significantly at risk.

7:20 Scott: Mm hmm. They're commoditizing those things and it's going to be a race to the bottom unless some model,

7:28 foundation model can have a demonstrable.

7:33 improvement quality,

7:35 right?

7:35 I mean, if you're a cursor and you can use a model that's 20% less effective,

7:43 but it saves you 50% of your cost,

7:47 you're going to do it.

7:48 Brad: Of course.

7:52 Scott: There's something else I saw in this article about open AI and anthropic that was the first thing that stuck out to me, honestly.

8:02 So I've been in a number of startups and I've been at startups that have been disciplined and I've been at startups that have gotten out over their skis.

8:09 And what I mean out over their skis is they're spending far ahead of the revenue.

8:13 Brad: Yep.

8:15 Scott: And typically, it is the case people spend ahead of the revenue because they think, well,

8:20 we're going to grow into the cost, right?

8:23 We're investing and we're going to grow.

8:25 And I've been part of a few situations where that has not happened.

8:30 You get the then you sort of head into the proverbial downward spiral of layoffs and then can't deliver and then more layoffs and then you're just put,

8:41 right?

8:43 And I've also been part of companies that have been disciplined and not gotten out over their skis or when they've gotten out over them, they haven't got too far as they've pulled it back a little bit.

8:53 I don't know. When I saw that, I just saw Anthropic being disciplined.

8:59 Brad: Yeah

9:00 Scott: And I believe that a key for success,

9:05 at least as a startup, but probably overall in the business world, is discipline.

9:10 Brad: Yep.

9:11 Scott: There are always going to be pressures pulling you and pushing you in different directions. There's going to be shiny new objects.

9:18 And the companies that I've been a part of that were able to stay focused and stick to their knitting and say,

9:25 this is what we do and we do it well, tended to be more effective and more successful.

9:31 And so when I see a story like this and I think to myself.

9:35 Well,

9:35 I would prefer to invest in Anthropik than invest in OpenAI because I bet what's going to happen with OpenAI is at some point in the next few years they're going to flame out or if not flame out have a big dumpster fire.

9:46 Brad: Well,

9:47 OK,

9:47 so another thing that happened again, this will be, you know, a little bit of history when people hear this, but, you know, there was some kerfuffle on the Internet this week where.

9:57 open ai which is a private company not a public company

10:00 Scott: Mm

10:01 Brad: so it's

10:01 Scott: -hmm.

10:01 Brad: not like we could get to see their financials but what they've shared and what their cfo shared um on twitter was something the effect of yeah we've got this 1.4 trillion dollars of of capital spend and you know it'd be really good if and they use the term it'd be great if you know the you know trump administration with you know the commitment to make sure that u.s wins the ai race you know if they would backstop

10:22 Stop R.

10:24 uh spend and the word backstop specifically was used and of course people turned that into wait a second are you asking for a bailout here and um you know david sacks the ai czar on trump's administration had to kind of come out and like no this is not we're not doing backups we're not doing backstops for this company oh that wasn't the word we meant to use right but you know i i heard it um on the on the um all in podcast i heard um david sacks talk about this like

10:50 look man this is one of the most vibrant marketplaces we've seen in a long time like this is cutting edge stuff but there are so many players in this space right you got the mag 7 all playing in the space we got these two other like really made a big big small startups small relative to the mag 7 but they're huge and and they're all fighting it out and we got that doesn't even include xai so we got xai we got open ai we got anthropic we got all the mag 7 and so

11:16 This is a very, very active and, you know, competitive marketplace.

11:23 And so, yeah,

11:26 so what if OpenAI gets over their skis and blows up?

11:30 That's not going to that that that'll be one out of half a dozen companies in a space that doesn't make it. And, you know, they're private companies. So the investors in them are, you know, hopefully they're all big boys and they knew what they were getting into when they invested in a risky company.

11:52 Scott: I agree with you 100%. I don't have any problem with OpenAI blowing up.

11:57 In fact, I think when you see a market like this, you need what they call it creative destruction.

12:03 You need the marketplace to really drive the winners,

12:07 to drive, you know, to flush out the losers and drive the winners. And I don't know, based on this article,

12:14 if I thought based on what I read, I think.

12:16 Anthropics is going to be a winner.

12:19 Brad: Well, or at least not the first loser.

12:22 Scott: He's not the first loser because they're actually trying to build,

12:27 and I could be wrong about this,

12:29 but it feels to me they're more trying to build a business than trying to just do a, you know, we saw this during the first dot-com boom, the eyeballs land grab.

12:39 Brad: Yeah,

12:40 so a couple of other things that also I think are interesting that have come out this week that I think are kind of worth noting.

12:48 A couple weeks ago,

12:49 you and I were talking about this maybe on an episode of the pod that hasn't been released yet,

12:52 but you were asking like,

12:53 hey, how come Siri sucks so bad?

12:55 Why isn't Siri at least as good as any of the other LLM models out there?

13:02 And what happened last week?

13:04 Apple announced a deal with Google Gemini to power Siri.

13:09 So Apple's going to be using Gemini to power Siri now.

13:12 And so to me, that's another similar example of your those that keep, you know,

13:19 to their own knitting.

13:20 Apple has been very like,

13:22 you know.

13:23 cautious slash deliberate about what they're spending in AI.

13:28 And I think it's really interesting that they, you know, partnered with Google instead of just throwing all of their own resources at this. They're the one of the Mag7 that aren't burning their free cash flow on infrastructure deployment.

13:45 Scott: Right.

13:47 That's very interesting. And that's the same play that Apple did, for instance,

13:50 with search, right? They did deals with Yahoo and Bing and Google,

13:53 Brad: I

13:54 Scott: etc.

13:54 That's kind of their playbook.

13:56 Who do you think pays in that deal?

13:57 Do you think Google is paying Apple or do you think Apple is paying Google?

14:03 Brad: think it tends to be like a kind of a little bit of both,

14:07 you know, I think in that Siri deal,

14:10 I have to imagine that Apple's paying Google,

14:13 but.

14:14 Um,

14:14 there must be like some,

14:16 you know, law back clawbacks where they don't have to pay if, you know, for certain types of things,

14:24 you know, exactly.

14:25 Scott: You know, I don't know, in the search world,

14:27 right,

14:27 Google monetizes those searches, so they pay Apple.

14:31 Brad: They paid Apple for that.

14:32 Scott: And so,

14:34 I

14:35 Brad: But there's

14:36 Scott: mean, there's...

14:36 Brad: not as much, there's not as much direct in.

14:40 You know, that's what that's what I'm saying. Like, I don't

14:42 Scott: Yeah,

14:42 Brad: think there's

14:43 Scott: it's

14:43 Brad: as much direct monetization from the LLM's voice recognition feature,

14:50 you know.

14:52 Scott: fascinating.

14:54 Well,

14:55 I'm excited to get to use Gemini on my phone.

14:59 Brad: yeah last time we were talking about this every time i said the word siri it lit my phone up and i had to tell it stop no i'm not talking to you siri um so that's one thing that's happened i think it's really i mean this is such a it's such an interesting space and i um you know again like it's it's fascinating times we live in right now

15:22 Scott: Yeah, that's one way of putting it for sure.

15:28 Uh,

15:29 do you remember Yahoo Pipes?

15:35 Brad: Maybe.

15:37 Scott: Well,

15:38 then you might remember a series of tools came out like

15:41 I-F-T-T-T and Zapier and

15:43 Brad: Yeah,

15:43 Scott: make,

15:44 Brad: yeah,

15:44 Scott: those

15:44 Brad: yeah, yeah, yeah.

15:45 If this, then that.

15:46 Scott: are just,

15:46 yep.

15:47 Brad: Zapier is still around.

15:48 Zapier, hey,

15:50 talk about people sticking to their knitting.

15:51 Zapier is a fantastic example of a bootstrapped company. Never took any venture funding and just like crushing it.

16:01 Scott: Well,

16:02 Yahoo Pipes was sort of the precursor to

16:06 zapier yahoo pipes so there's a thing where you could like visually drag and drop pipes each each little i mean the pipes didn't do anything but the little you know the y's and t's and things you know the pipe shapes did and you could like create these little workflows and it was all drag and drop and the workflows could do things like read in an rss feed and then send emails based on it or whatever all sorts of stuff like that and then

16:30 If this than that and Zapier those things came along it was the same kind of thing you have these connectors you connect them up and it could based on this do that but I just saw today that Google released Google flows you want to know what Google flows can do they're

16:44 Brad: send stuff to an LLM.

16:48 Scott: now using LLMs to do stuff like this I mean it was really I mean they're look they're just doing great work over at Google.

16:58 Let's see, they can notify me about urgent emails,

17:01 can catch up on news,

17:03 I can get a pre-meeting brief,

17:04 I can get a daily summary of unread emails,

17:08 can notify me about emails from key people.

17:11 When they're using these expensive AI to do stuff that Yahoo could do in 1997.

17:16 Brad: Yeah, yeah, yeah, yeah, yeah.

17:18 Yeah.

17:19 Scott: I don't know when Yahoo pipes really came out, but it was probably about 97.

17:23 Let's actually, let's just Google that real quick.

17:26 I should really go to Yahoo.

17:27 Can you even go, does Yahoo even exist anymore?

17:30 Indeed,

17:31 it does.

17:32 Yahoo pipes, just searching for it.

17:35 See what comes up here.

17:36 Brad: Oh, Wikipedia's first article, of course.

17:39 Scott: Yeah,

17:40 it was launched in 2007 and it was shut down in 2015.

17:44 Brad: There you go.

17:46 Hey,

17:46 speaking of Wikipedia,

17:48 have you heard of Grokopedia?

17:52 Scott: Is Grokopedia Elon Musk's version of a...

17:58 Yeah. Elan's view of the world.

18:00 Brad: Well,

18:01 it's it's, you know, I mean, it's grok based Wikipedia and,

18:06 you

18:06 Scott: What's that mean,

18:07 grok

18:07 Brad: know, I mean,

18:07 Scott: -based Wikipedia?

18:08 I don't even know what that means.

18:09 Brad: they're basically, you know, building out.

18:13 I mean, I think that your your characterization of it is Elon Musk's version of the truth of the world. Well, that's certainly not the way that he would describe it,

18:21 Scott: Oh,

18:21 Brad: you

18:22 Scott: isn't

18:22 Brad: know.

18:22 Scott: grok

18:23 Brad: -

18:23 Scott: So I'm reading on Wikipedia what Grokopedia is.

18:26 It is an AI-generated online encyclopedia. That's actually kind of interesting.

18:30 Brad: Yeah, right

18:32 Scott: When it's just a search box.

18:36 Or should I type in that search box?

18:38 I'm going to type in my own name.

18:39 Of course, that's what you always do.

18:41 Now it's given me 85 pages of names.

18:44 My name has yielded 1,010 results, but none of them are my name.

18:50 They're other people's names.

18:52 That's not what I wanted.

18:53 But this is actually kind of interesting.

18:55 So what this is telling me is there are no articles in this thing.

18:58 It is just AI generated like real time.

19:05 Oh, this is interesting.

19:06 Up at the top of this page, I clicked on one of these guys, Scott Wolf,

19:10 who apparently is an American actor.

19:11 He played Bailey Salinger in Party of Five.

19:15 It says fact-checked by

19:17 Brad: Yeah.

19:17 Scott: Grok two weeks ago.

19:19 Brad: Yeah.

19:20 Scott: That's interesting.

19:21 Brad: Yeah.

19:22 Scott: Very

19:22 Brad: Yeah.

19:23 Scott: curious.

19:23 Brad: So there you go.

19:24 So there's more more news in AI is AI generated Wikipedia.

19:31 Scott: Fascinating.

19:32 Brad: Yeah.

19:33 Scott: Fascinating.

19:34 You know, it's a fun time to have a bucket of popcorn and be watching it all unfold.

19:41 Brad: Right. I mean, unless,

19:42 say, you're worried about white collar job attrition and you've got you're a young person or you're closely aligned with a young person struggling to find work in this in this new economy or all these different things.

19:59 Scott: if you care about climate change.

20:01 Brad: Or if you care about climate change.

20:03 Scott: Yeah.

20:05 Brad: Well, there you go. Did we got

20:06 Scott: Yeah,

20:07 if we solve nothing today.

20:09 Brad: well,

20:10 so what did we what did we learn? We learned that anthropic is being more business focused in building their business.

20:18 Open AI is swinging for the fences. There's massive amounts of churn going on.

20:23 We we've got our big giant bags of popcorn to watch it all and it's fine.

20:29 It's fine.

20:30 As flames are around us.

20:33 Scott: This is fine.

20:35 Yeah, we learned, look, I think we also learned something else. The big players,

20:39 we talked about Apple and Gemini.

20:41 Ultimately, the big players have such, it's not like monopoly power,

20:46 but they've got such power in the marketplace that they're going to get a huge chunk of share,

20:51 right?

20:53 And I mean, just look, I'm going to just use a concrete example from where I'm at at TaskRabbit.

20:57 We didn't buy everybody an OpenAI license for ChatGPT.

21:01 We bought everybody a Gemini license.

21:04 Because

21:04 Brad: Oh.

21:04 Scott: why?

21:05 Because we're a Google workspace company and it's just right there and it's integrated with everything.

21:10 And in fact, in fact, it's been making me think, and this would be hard to do, that we should force people at our company that we should stop using Slack and start using.

21:19 using Google chat because Jim and I can see all of our Google chat and you can query your chats and get stuff.

21:25 Brad: I, by the way, now you just like one more piece of news that I heard was and I don't have the exact numbers here,

21:32 but.

21:33 Yeah,

21:33 like number of tokens being processed and generated by Gemini have just been skyrocketing over the last six months.

21:40 So they have come back. I mean, they were supposedly nowhere and they have come roaring back.

21:46 Like they're by far, you know, the fastest growing of the foundation models.

21:50 So don't write Google off.

21:55 Scott: I think that's because I have a hypothesis on that I could be wrong.

22:00 I was talking to a guy at Microsoft who works on Copilot, and Google has added,

22:07 they gave you Gemini for free with your workspace account and then simultaneously raised the prices of workspace by two bucks a month.

22:17 So it's free,

22:18 but we're charging you more.

22:20 Brad: Exactly.

22:20 Scott: Microsoft did not do that. They strategically,

22:23 this guy told me they strategically, you know, it was a competitive thing, decided not to do that.

22:27 And so I wonder if it's just like it's there, it's in front of people and they can use it.

22:32 And so and also Google has released,

22:34 I don't know if you've seen, if you've just gone to Google and searched,

22:37 they have this new thing called AI mode,

22:39 which

22:39 Brad: Oh, yeah.

22:39 Scott: basically you're not searching anymore,

22:41 you're just doing an AI prompt.

22:42 And so they're nudging their customers into that space.

22:46 Brad: Yeah,

22:46 for sure.

22:47 And also,

22:48 you know, you see the, you know, another example of like, is OpenAI focused on the right things?

22:53 They released

22:54 released a version of chromium browser that's the the ai the open ai version of the chromium browser it's

23:01 Scott: Mm hmm.

23:01 Brad: like that's not gonna last like like you're gonna compete with you're gonna compete with google chrome like

23:09 Scott: And

23:09 Brad: no

23:09 Scott: as a matter of fact, I have a button up here in the upper right of Chrome now,

23:14 which.

23:15 uh goes right to gemini so

23:17 Brad: Yeah, exactly.

23:18 Scott: like this this is the same story of the browser you know microsoft windows and the browser is it isn't quite monopoly power but google has that you know frankly google and apple have this footprint that they can push ai on people in a way push their products on a people that way that nobody else can i mean between android and chrome and then

23:41 And iOS,

23:42 it's like, that's your game.

23:44 I think if I'm looking at investing in anthropic or investing in open AI, I'm saying to myself,

23:49 open AI is going to lose,

23:51 anthropic is going to build a nice business and the winners are going to be the major platforms.

23:57 Brad: Yep.

23:58 That sounds like not a not that controversial of an investment

24:02 Scott: Well,

24:03 Brad: thesis.

24:03 Scott: that is my armchair. I will take my $7 and go put it where I can put it. All right,

24:11 Brad,

24:12 this has been fun again.

24:13 Brad: Yes, as always.

24:14 Scott: News roundup.

24:17 Oh, what do we do when

24:18 Brad: What do we do? We smash that subscribe button.

24:20 Scott: we rate and review?

24:22 Brad: We get agents. We program our agents to rate and review.

24:25 Scott: Yeah,

24:25 we should try that.

24:27 Brad: All right, everyone.

24:28 Scott: All right. Talk to you later,

24:29 Brad.

24:29 Brad: Bye.

24:30 This episode of How Many CTOs Does It Take podcast was edited by Saffron Eftagal.

24:35 Our theme song was AI generated by Brad.

24:39 Please follow us on all the social media platforms at at HowManyCPOsPod.

24:44 You can find us on YouTube,

24:46 LinkedIn,

24:47 Instagram,

24:47 Facebook, Threads, X,

24:48 TikTok,

24:49 and Patreon,

24:50 or at our website at HowManyCPOsPod.com.

24:54 And remember to rate, review,

24:57 and subscribe on the podcast platform of your choice.