00:00We seem to have spent the last few days thinking about what AI means for software.
00:04The market has spent the last few days thinking about that.
00:06The market has decided very quickly to sell software stocks.
00:09Has the selling been too indiscriminate?
00:12How would you characterize it?
00:12Absolutely, it's been too indiscriminate.
00:14The market has oversold software.
00:16So if you look at it, number one, nobody is vibe coding their payroll.
00:19That is a future we're not going to live in.
00:218 to 12% of enterprise software spent today is on software.
00:26That is not going to be subsumed by vibe coding.
00:29The stakes are so much higher for the rest of the spend, number one.
00:32Number two, if you look at prices for SaaS companies, 75% of the public names have raised prices since ChatGPT was released.
00:39So all the signs, and then Google's quarter, which was exceptional, all the signs are proving and sort of indicating that we're oversold software.
00:45Okay, so they're not vibe coding.
00:47So you're saying there's a room for specialist software that's dedicated to the job at hand.
00:54Is that what we take away from the year?
00:56I think that the incumbents are highly capable in many of these markets.
00:59CrowdStrike is a very capable incumbent.
01:01ServiceNow is a capable incumbent.
01:03They have the right to win and deploy AI within the workflow and the customer base that they have today.
01:07And I think the market's missing that.
01:09Does it help or does it hinder your ability to pick winners and losers in this environment?
01:13I mean, we're in the job of the micro, and a lot of the things we think about are new emerging markets versus what's going to happen with established markets.
01:19If you look at the entrance of Google, that did nothing to Microsoft in terms of their dominance with their office products, right?
01:25Search was a completely new category.
01:27So often when we have these product cycles, what we see is that the incumbents do better in their existing markets, and they're irrelevant in emerging markets.
01:34We're mostly focused on the emerging markets.
01:36There's been a long debate amongst VCs and founders about what happens when the models start to push higher up into the verticals.
01:42And we've got a reminder of that with Anthropic and the plug-in in terms of legal services this week.
01:47What does that mean for the companies, the startups that do the wraps around the models?
01:52How defensible are those kind of businesses?
01:54I think the wrapper story is overtold.
01:56So if you look at the labs today, they've done a formidable job of building out their models.
01:59But there are many categories in which you want to be multi-model.
02:02That is number one, right?
02:03Number two, if you take a look at the feature surface that's implied by many of these software products.
02:08So you can do meeting transcription with AI, and many AI companies do that.
02:11But the correct product is to build spreadsheets, to build a word processor, to build a slides product.
02:16So there's just an enormous amount of software to build around the capability.
02:19And I don't believe the labs are going to prioritize both the capability and the software stack.
02:23Let's go back to the beginning.
02:24Your sense is that the markets have overdone it, that we've pushed too far on the downside.
02:29I believe so, yeah.
02:29Okay.
02:30Isaac Kim over at Lightspeed.
02:32Technology, private equity in its current form is dead.
02:34Yes.
02:36Totally disagree.
02:37We've said this so many times.
02:38We said that new media was going to kill old media, yet we're sitting here today, right?
02:41How many 30 years after AOL Time Warner?
02:43We said that PC was going to kill the mainframe and data centers.
02:46More data centers than ever.
02:48And guess what?
02:48Mainframes are still relevant.
02:49Okay.
02:50Yes.
02:50So private markets haven't over-allocated to software?
02:54No.
02:54I don't believe so.
02:55100% sure on that.
02:56There's going to be an enormous flood of software.
02:58In fact, the world is short software today.
03:00Look at all of the industries in which software has not been a relevant player.
03:03Look at legal, for example, right?
03:05We've got maybe a $50 billion software market.
03:07We've got a $550 billion overall market.
03:10Legal is infrastructure for capitalism.
03:12There's no way that is one-end market.
03:14So why is everybody so worried?
03:15I mean, I'm not going to talk about day-to-day stock prices, but I think the market has overreacted
03:19to the perception that vibe coding is going to cause enterprises to rip and replace
03:23their SaaS spend.
03:24They're just not.
03:24Just on the legal piece, so the Lagoras and the Harveys of the world, these companies that
03:28build these legal options with AI embedded, they are defensible from the kind of product
03:33that Anthropik's putting out?
03:34That moat is deep enough?
03:35I absolutely believe they are.
03:36I think what Anthropik is doing is product marketing more than delivering product capabilities
03:40that are vertically oriented to that market.
03:42And if you look at legal, look, it's an industry, not a market.
03:45It is too small to say that this is just one-end market.
03:47This is an industry that's going to have many winners and many sub-markets within it.
03:50Does this also mean that the infrastructure spend that we're seeing and those commitments
03:53around CapEx, that's all rational?
03:54That all makes sense as well?
03:56Look, without speaking to specific companies' infraspend, what I will say is look at Open
04:00AI's announcement this year, right?
04:01They said that they tripled capacity, they tripled revenue.
04:04So what we're seeing is every time they bring supply online, you're seeing 100% of it allocated
04:08by demand.
04:09Do you worry about Open AI, given the progress that Anthropik's made on Enterprise and the progress
04:13that Alphabet's made?
04:14They're invested in, of course, by Andrews.
04:16Look, so far what we're seeing is prices are going up, not down, and we're finding all
04:19the foundation models are finding their own directions to specialize in, right?
04:23So Google's done a really nice job on their multimedia models.
04:25Dano Banana is excellent.
04:26Anthropik's done a nice job on coding models.
04:29And Open AI, ChatGPT's, $950 million weekly.
04:32It wins on consumer.
04:33So far.
04:34So far.
04:35And will they pay?
04:36How much will they pay and when will they pay?
04:37I mean, if you look at the top SKUs of these companies, right, Gemini Ultra is $250 a
04:41month.
04:41The top ChatGPT SKU is $200 a month, right?
04:44Grok, super heavy, $300 a month.
04:46We've never seen consumers paying these prices for software before.
04:50So whilst we're busy talking about software and worrying about software going out of business
04:54and white-colored jobs, what should we be thinking about?
04:56What is the tech story that we are missing because we're too busy talking about this?
05:00Look, I think the world is short software.
05:01Software is a huge both productivity driver and it's a deflationary force.
05:05This is how we get to 10% GDP.
05:08That sounds wild, but I think that it's possible.
05:10Sounds exciting.
05:11It is exciting.
05:11When you say the word short software, you mean something else other than that.
05:14That's right.
05:15That's right.
05:15That's right.
05:16I mean, it's oversold.
05:17There's not enough software in the world and we're going to see more of it.
05:19Yes.
05:20Data center build out.
05:21Yes.
05:22Too much of it.
05:22We're going to end up with a glut.
05:23The prices need to come down.
05:25What's your view?
05:25All I can tell you is that the biggest sort of companies that are investing in data centers
05:29so far are seeing 100% of the supply they bring online be allocated.
05:33So, so far, it seems like a wise investment.
05:35So far.
05:36That's now.
05:37What about going forward?
05:39I mean, what is the sort of market and demand that is latent today for intelligence?
05:44That's what we're talking about.
05:45So I don't see a slowdown in the near term future.
05:47Okay.
Comments