Bessemer Venture Partners - State of the Cloud 2024

Live from Schenectady, New York, it's SaaS Talk with the Metrics Brothers, Growth and CAC.

And I'm Growth, better known as Ray Reich, the founder and CEO of Benchmarkit.

And I'm CAC, better known as Dave Kellogg, independent consultant, EIR at Balderson Capital, and the author of Kellogg.

And together we are the Metrics Brothers.

And we go together like bangers and mash.

Back in the UK again, Dave?

No, just back to my food analogies. They're easier to come up with than a wink and a smile, um, and they're less controversial than we go together like the Winklevoss twins.

Wow, can you believe it? They are— oh, I think they're 42 years old now. Isn't that terrifying?

Time flies, Ray. It was just the other day they were at Harvard with Mark Zuckerberg, I guess, getting the idea for Facebook ripped off, if I understand the story correctly.

Yeah, I'll have to go back and watch that social media movie again. But speaking of time flies, I still remember in 2015 when Byron Dieter, on his own from Bessemer, published the very first State of the Cloud report. 9 years ago.

Wow, 2015. So this year is going to be the 10th edition then, if I count correctly, inclusively. So let's take, in this episode, Ray, why don't we take a look into the State of the Cloud 2024 report by Bessemer? But before we do that, let's hear a word from our sponsor.

The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast. Okay, so let's dive into the State of the Cloud 2024. As you said, the 10th annual report from Bessemer. But Dave, per many social media articles and even a recent Janelle Ting post, isn't the cloud dead?

Yeah, I saw that post by Janelle. You know, EM Cloud, the Bessemer Cloud Index, she says it's been down 10% year to date. While the NASDAQ has been up 20%. She shows a chart from Goldman Sachs with growth rates monotonically decreasing from 2Q21, 35%, down to around 20% today. She also shows a different chart with NDR, net dollar retention, down from about 120% median in the ZIRP days to about 110% today.

Yeah, and those numbers aren't, aren't good. But if you then look at Jamin Ball, Cloud of Judgment, he's looking at median growth rates of public cloud companies around 12% today. And in the latest benchmark survey, and we do that only on private SaaS companies, Dave, the median NDR is 101%. So, the metrics definitely aren't saying positive things about the industry right now.

Yeah, and I think, look, the industry is definitely in a downturn. And I think that has kind of provided fuel for the fire for a small viral craze on Twitter about an article written by a guy named Christopher Pike called The End of Software. So if you've been on social media lately, you'll see this kind of end of software meme. And in my opinion, the article's not really worth reading. It's about, I don't know, 500 words. It's a Google Doc, and it makes a relatively naive comparison between software and media. So I was not a fan of the article. Ironically, I am a fan of some of the responses to it. Janelle and her post at Bessemer, and then Christopher Janss, a VC in Europe, have both written posts with roughly the same title called Rumors of the Death of Software Have Been Greatly Exaggerated, kind of a Mark Twain playoff. And Aaron Levy from Box did a nice tweet on it. So I did want to talk about the article. You can go read it if you want to, but I would quickly go look for some of these responses if you wanted to understand more about it.

Yeah, but I think we're piling on because the basic sentiment out there is SaaS is dead because AI is coming. And just our last episode where we focused on the McKinsey report around the disruption of generative AI on traditional software, it really highlighted that SaaS and software is not dead because as the generative AI marketplace is going to grow from like $25 billion to $200 billion over the next 4 years, it says that broader platforms And I'm interpreting broader platforms to include legacy SaaS or enterprise software platforms, can capture anywhere from 30% to 45% of the growth that's created by generative AI. So I don't think it's dead. I just think it's going to be reincarnated, Dave.

Yeah, software is evolving. AI is certainly a disruptive evolution to software. AI is most definitely— GenAI software is software, and it's growing like a weed. But the main message to me from our prior conversations is this is evolution, this is even disruptive evolution, it's certainly not death. And software has been evolving for a long time. Both Ray and I have been in the business probably 35 years plus. And Morgan Stanley recently put out a nice piece of research called Revisiting Software 101, and it's got a slide that shows really the 4 dimensions of software evolution, how software is built, where software is run, how software is sold, and what software can do, kind of functionally. And really, AI is big, right? Cloud was all about where is software run, right? How is software built? I mean, back when we were young, right, it was a monolithic app, then we had client-server, then web, then containers, then serverless, right? So how software is built, the environment has changed. Where it's run, cloud has been a huge change. How it's sold, right, from perpetual to subscription to consumption, Also a big change over the past 30 years. You might argue full circle, because I know you started out in timesharing, although we're not allowed to say that, you're not allowed to say that word. And then finally, what software could do, you know, AI and generative AI are bringing new capabilities.

You know what's funny? I wrote a LinkedIn post on this evolution of software about 2 weeks ago before we even, I saw this chart. And the only thing I would add is between web-based applications and containers, The mobile craze also had apps and the app ecosystem. And not that that was truly business-to-business type, but that was another kind of software built and delivered was on the app stores, right?

Yeah, absolutely. Absolutely. I've not seen that post yet, Ray. I'll have to take a look. I usually keep up with your social medias, but you're so active.

I am.

So let's dig in, Ray, to the BVP, the Best of America Venture Partners State of the Cloud 2024 report. You know, back in the day, it was the one name on the byline, Byron Dieter, to there are 13 names on the byline, and Byron is actually not one of them. So there's no single author of this report. Our friend Janelle Tang is up there, as well as a few others we know, but this has been a group effort. One of the things they do in the report is they actually do kind of internal surveying of Bessemer to get opinions on hot issues, which I think we'll talk about. So that's again reflecting the group contribution. The subtitle of this report is The Legacy Cloud Is Dead, Long Live AI Cloud. Right? So this is back to the meme where, for some reason, people are unable to say that things have evolved, or even disruptively evolved. And always people wanna say, the king is dead, long live the king. So they've done that. Hey, they're VCs, we love them, we have friends there, but they are VCs after all, they're prone to drama. And the intro says, you know, this year's State of the Cloud report serves both as a eulogy for the legacy cloud. My, oh my, legacy cloud needs a eulogy.

Yeah.

and a celebration of the current AI cloud moment.

But you know what, Dave? You have to give them a lot of credit because they also win the VC Honesty Award with a quote of the year, and that is, and AI is all we can think about. You know, we've seen VCs kind of follow the crowd on megatrends before, but they said right up front, we're there. That's all we can think about.

A+ to Bessemer for owning that and coming out and saying the quiet part out loud. That was, that was pretty good. Ray, are there any new metrics that are the focus of this report?

You know, it's funny, I went back to that 2015 report and, you know, Byron talked about things like CAC payback period and NDR, NRR, but no, there's no new metrics at all mentioned in this report. It's really kind of a macro-level trend identification of 5 key trends in AI, and it's all about Growth, and I would say market share. Who's going to capture the initial market share that's being created by this generative AI craze?

Yeah, I noticed the same thing, which, which is, at least to me, this report was famous for its metrics orientation in the past, right? They might even introduce new metrics. I know there were early-to-CAC ratio and CAC payback period and, um, probably some other metrics that they touted early on. Um, and it was interesting, this report was Literally devoid of metrics, which I think perhaps is a sign of the times that, you know, I think change happens, then we make our early attempts at measuring it. And when you come up with the CAC ratio, people look at you like you're a visionary. Oh my God, the CAC ratio. And then over time, it kind of mainstreams and everybody wants to know how to calculate it. And do you amortize commissions and, you know, do you do it on net new or new, right? You get this maturation of the industry. And I think the same thing is going on here with AI.

Yeah.

Which is there are no metrics here at all. People are just talking about the trends because there's a giant trend reinventing the industry and we're not ready to debate how to measure it yet. Let's just talk about first what's happening. And they've done that by identifying these 5 trends.

Yeah, you know, can't believe it, but here it is, Bessemer. And one of the favorite metrics they created was the cash conversion score. You don't see any VC talking about cash conversion scores in generative AI right now.

Yeah, yeah, yeah, yeah. Although you still hear about quite a bit in classical SaaS, and burn multiple is very in vogue right now. We can add that to the eulogy, I suppose. But Ray, do you wanna introduce the 5 trends?

I do, and I really like them, and everyone should go and read this State of the Cloud 2024 report because they have a nice image on this. The first one is that the foundational models, the LLM foundational models, are going to set the stage for big tech's new battle of the century. And we're going to go in each one of these. Second is that AI can turn all of us into 10x developers, which is a really interesting statement about the value of developers in the future, but we'll hold on that. Third is multimodal models and AI agents will transform human relationship with software. I guess the software now is going to be like my partner in life. Fourth, vertical AI shows real potential to dwarf legacy SaaS with new applications and business models. And the fifth trend is AI brings consumer cloud back from the dead, which I know nothing about consumer cloud, so I don't know if I can add much value there, Dave.

Awesome. Well, I'd like you to say multimodal model 5 times in a row, Ray, and then you can tell me what a multimodal model is for extra credit. But that's trend number 3. So we'll come back to that in a minute. Let's start by doing trend number 1. I'll take the lead on this one, Ray. So trend 1 is the battle over foundation models. The first battle is always over infrastructure because that's what apps get built on, right? And there are strong increasing returns of market leadership in infrastructure because the world is trying to figure out what infrastructure to use, right? How wide should the railroad tracks be, right? And then we can go build railroad coaches. By the way, most railroad tracks, You know the trivia on that, Ray? I don't know if it's a myth or not. How far apart are railroad tracks? You know this one?

I don't know this one.

The rumor is it's the width of a Roman chariot axle. We'll have to go figure out if that's true. But the first battle is always over infrastructure. And these models are infrastructure. In the AI world, foundation models are the things that you build upon. So almost 2/3 of funding in AI has gone into the foundation layer. OpenAI, Anthropic, Cohere, Mistral. I think they've raised $23 billion. Uh, well, actually, sorry, in aggregate, $124 billion. So Mistral's most recent round. Sorry, I'm gonna mess up the numbers already. You know what those numbers are again? I can't remember.

Yeah, so almost 2/3 of AI funding in 2023 was based upon the foundational model layer.

Yeah.

And that equaled $23 billion of funding at an aggregate pre-money valuation of $124 billion. So $23 billion was almost 60% of total VC money that went into AI last year.

Thanks for bailing me out there, Ray. Uh, that's what I thought it said, but the notes weren't clear in my script. I apologize. So 2/3 goes to the infrastructure, 1/3 has gone to apps and tech/tools combined. The thing that's been unusual about VC investment in AI is it's got a heavy corporate VC contribution. And that is unusual, right? You've got Microsoft, Google, Amazon, NVIDIA, all putting a ton of money into AI Foundation and building partnerships with foundation model companies.

Oh yeah. I mean, you look at things like, of course, Microsoft, that famous $10 billion investment in OpenAI, and then Amazon invested in Anthropic and Hugging Face. And we won't get into large language models versus small language models, but that's what Hugging Face is. And then Google invested also in Anthropic and Hugging Face simultaneously. So it's going to be interesting to see who wins that language model war, Dave.

It will be. And the prediction from Bessemer, because part of this is open source versus closed source, and the general trend is that open source lags closed source for a bunch of reasons. And Bessemer is predicting that closed source models will drive the bulk of LLM compute cycles And the AI model giants will split the economic pie. And I don't disagree with that prediction.

Yeah, but it is going to be interesting to see what some of these open source models, what type of market share and momentum they can share. And I think about Facebook and LLaMA, right? They're trying to really promote an open source model to LLM. So it'll be interesting.

You want to do trend number 2, Ray?

Yeah, trend number 2, because I've always wanted to think I could be a software developer. And trend number 2 says that AI will make all of us 10x developers. And this is— think about these things like Copilot. And GitHub, you know, they came out with their Copilot in conjunction with Microsoft. And it's really allowing developers to not only do some optimization of their code, but it's actually allowing all of us to do some coding. I mean, copilots alone raised $4 billion in 2023. And by the way, that death of software article you talked about, and it was comparing it to media companies and things like Forbes, it will be interesting, Dave, to see if someone like me can actually develop a basic piece of application software using a copilot. And if there are 1,000 people like me, it's gonna make software even more commoditized. So, there is potential here to make software a little more commoditized than it is today.

So, I've been hearing about the citizen developer since literally 1985. My first job, I worked at Ingres, which is an original relational database company, and we had a product called Ingres/4GL, which was a 4th-generation language. And the idea is that 3rd-generation languages, like at the time, Fortran and Pascal, Or BASIC even were designed for technical people, but we'd have this 4th generation language so that people could write their own code. In fact, SQL, believe it or not, was also thought to be an end-user query language. So you get a little jaded after 35 years of hearing about the citizen developer. I do think copilots are real, and I think they're going to help because it's a better model. Basically, you're not trying to just get a higher-level language for you to code in. I think we're saying, eh, forget about it, Ray. Let's turn you into a reviewer of software rather than a developer of software. And we're going to have the copilot write code for you. And I do think that's a better model. Yeah.

And I don't know if you remember, but in that last episode, we, in the McKinsey report, it said that 30% of investment in GenAI applications would be done internal by internal IT and software development shops. And I'm like, Whoa, that scares me. But now that I read this report, it started to make a little bit more sense that we can take that average developer in a Fortune 500 company and they can become more software reviewers based upon the process and the business that they're trying to optimize with GenAI versus just being the world's best technical developers, if that makes sense to you.

It does, right? Okay, so let's go on to trend 3, the tongue twister, multimodal models and AI agents transform human relationship with software. Boy, there's a lot of alliteration in there. By modal here, they mean modalities, i.e., voice, image, video, text. Historically, AI has required the model to be specific to the modality, that you could create a model for image generation and image recognition, ditto for voice, ditto for text. And now I think I'm not sure what model broke us through, but of late, the recent breakthroughs have made multimodal models. Oh, I hate saying that. Multimodal models where the model itself can be used for multiple modes. And that's what they're saying here. So what's a multimodal model? It's a model that can work in multiple modalities. It's also saying that AI agents will transform the human relationship with software. And therefore, you're going to see agents for customer service, sales, help desk, HR support, vertical apps, medical, and more. And for me, Ray, this is a very real working today and super practical AI app. So I'm a big believer in multimodal models and a big believer in AI agents.

Oh, I totally agree. I don't know when ChatGPT, and I think it was 4.0, introduced the voice modality. I actually was playing with it and it's like, oh, this is pretty good because I don't need to type slowly. to become a prompt engineer. I can just talk to it. So I think it is the future and I love voice.

So then finally, their prediction on this one is that voice AI applications will unlock $10 billion of new software TAM over the next 5 years. So that's why we see so many startups working on agents for just about anything. You can find 10 startups doing it, whether it's— who's our friends at Piper, the AISDR? Who's that from? I can't remember their name. Qualified. Whether you've got the AISDR, whether you have AI customer service reps, whether you have AI chatbots, for website inquiries, help desk, internal support, you name it. There's 10 startups building a voice-driven AI agent.

Okay. Now, I want to move on to trend 4 because this one, I had to smile a little bit, Dave, if that's okay, because the trend is vertical AI.

No one can— it's a podcast, Ray. No one can see you smile.

Vertical AI dwarfs legacy vertical SaaS. If you— the State of the Cloud 2022 report, I believe, and one of the trends there was vertical SaaS was going to win and be much bigger than horizontal SaaS. Now, even though there's 20 public vertical SaaS companies that have a combined market cap of $300 billion, it really never quite took off the way a lot of VCs thought it was going to. So now Bessemer says, okay, generative AI is what's really going to make vertical software really take off. And it was interesting in this report. I never thought of this, right? That they mentioned that vertical AI can hit that professional services market. So if you look at the GDP of the United States economy, only 1% is software. And I believe it's 13% is made up of services. Things like law firms, accounting firms, other things. And they think that segment of the economy is ripe for automation using GenAI. So they took that vertical software to a new category, and that's the professional services industry. I think that could be huge. And it's one of the reasons I think white-collar jobs are more at risk than blue-collar jobs with GenAI tools.

Yeah, so look, Ray, I'm always a big believer in focus. I like vertical SaaS. I mean, today's top 20 companies have a public market cap of more than $300 billion. So if you're an investor, that's a lot of market cap to go chase. I just think when you get vertical, you get closer to customer solutions. And as a necessary— it's almost a necessary step. You eat into the services that would have adapted a horizontal solution to the vertical, right? You know, Veeva is obviously a super successful vertical SaaS company. The other thing about vertical SaaS is it's sticky as heck. It's very hard. Once these people get seated strongly in a vertical, it's very hard to pull them out. So I like vertical SaaS. It's closer to a solution. It drives more value for the customer. I think AI could help a lot with vertical SaaS. It's gonna make it easier to build, and it's gonna let you chip into the services market through the verticalization process.

I totally agree. Now, I'm going to do something I typically don't do, and that is just hand off a topic because I just feel I'm not qualified to talk about it. And that's around trend number 5, that AI brings the consumer cloud back from the dead. What say you about this trend, Dave?

Well, I was doing all the even-number trends, so I got to avoid this, Ray. So look, consumer SaaS has not been a great market for the past decade. And by consumer, we mean for individuals, not like SMB. I'm not talking about Notion or consumer-driven enterprise solutions. I'm talking about real consumer SaaS. So companies that provide cloud-based storage, compute, and apps directly to individual customers. And that's how Bessemer defines it. Their prediction is that AI will not only reinvent a lot of our favorite pastimes like social entertainment, shopping, travel, but will also help us discover and reimagine new ways for people to connect, buy, play, and explore in the world. And if you look at kind of their 8 years of the Cloud 100, you could see some of the companies they consider to be consumer SaaS. So they would put people like Dropbox, who's worth, you know, multiple billions of dollars, or Dashlane, who's worth at least a billion. You've got some companies that have had success, people like Grammarly, right? Or Calendly was worth $1.4 billion. So you find these people there, but it's still been a little of a tough space. And the number one reason you use the consumer is to kind of seed an enterprise and then go build an SMB, then a mid-market. and then an enterprise business. So neither one of us here are experts in consumer SaaS. I don't know if that's cause or effect on why it's done how it's done, but I do think Bessemer makes a good point that this could change consumer SaaS, both in terms of what consumers want to do and the ease with which you build these solutions and ideally bridge them into enterprise solutions.

Yeah, I just want to make sure that both Facebook and Apple can give me a VR headset when I'm connecting, playing, and exploring the world.

Ray, that's what you want to go through spreadsheets in. You want to fly through spreadsheets with numbers all around you.

Ooh, I can see macros over here and macros over there. I love it already.

You can float through your VLOOKUPs. Just think about that for a minute, Ray. Oh man.

Well, I think we need to wrap up this episode of SAS Talk with the Metrics Brothers with overall, I really thought this was a very, very good to great report because it's talking about where the industry is going in the future, and it really made me think. So that's one of the things I really liked about it, Dave.

Yeah, I recommend reading this report. It's a great report with a lot of history behind it. Second, look, a lot of our audience are practitioners, and the temptation for practitioners is to be so busy doing your job, especially in tough times, that you don't spend enough time with your head up looking forward. So I'd recommend you to flip through this report. But more importantly, if you're a RevOps person, if you're a marketing person, if you're a sales leader, do spend time exposing yourself to new AI solutions in your category, because today they could probably give you a competitive advantage. And if you wait 2 years, you're going to be at competitive disadvantage, right? Now is the time to look at the impact of AI solutions on your job and to go start using and experimenting with them so you kind of don't get left behind. as we go through this AI transformation.

Totally agree. But by the way, no SaaS metrics in this year's report. So you've heard it here first, Dave. I'm going to capture the AI magic number for a metric next year. What do you think of that?

Okay. Well, we won't be able to cover the 11th version of this thing if they don't put some metrics in, but hopefully they will next year.

Okay. Hey, great talking about the future and the evolution of the SaaS industry to integrate a lot of great GenAI capabilities. Talk to you later, Dave. Thanks again.

See you, Ray. Take care. The Metrics Brothers AI, SaaS, and More is a production of the Metrics Brothers Growth and CAC and a member of the Benchmarkit Podcast Network. By accessing this podcast, you acknowledge that the Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided or the humor content of the jokes told. Ray, information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional or for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance upon the information presented here. Ray Grothreich is based in New York City and available on X @rayreich. Dave Kellogg is based in Silicon Valley and available on X @davekellogg. Kellogg. Schedectity, which is French for unspellable, is not our actual production location. Thanks for listening.