Jerry Chen | The New Wave of Cloud Innovation

Hi, everyone. Welcome to Gray Matter, the podcast from Greylock, where we share stories from company builders and business leaders. I'm Heather Mack, head of editorial at Greylock. I'm joined today by Jerry Chin, who is a general partner here at Greylock. Jerry works with startups building enterprise applications and infrastructure, particularly those purpose built for cloud native ecosystems. Thanks for being here, Jerry.

Hey, thanks for having me, Heather. Always fun to chat about cloud stuff. Always.

Jerry's also the mastermind, if you will, behind Greylock's Castles in the Cloud project, which is an interactive database that maps activity in the venture capital funding startup ecosystem with other big three cloud providers. Jerry and fellow Greylock investor Jason Reese just completed our analysis of the 2022 fundraising landscape. Then they spent the first few months of 2023 looking at a few trends and comparing them to what's playing out in the market right now. You can read Jason's essay on the findings on the content section of our website and you can see the full Castles in the Cloud project. The latest figures at Greylock do slash castles. Both of those are linked in the show notes. So, Jerry, let's talk about what's happening. Top line, what did the funding figures tell you?

Well, first, I don't know if I've ever been called Mastermind, at least not since my last art heist. But I'll take what I can get. Heather. It's interesting, but remember, these numbers are about 2022 trends. And Jason and I in his blog and this podcast would talk about what we're seeing since the end of 22. But in 2022, we saw about 17 and a half billion dollars invested across a little bit over 400 companies. Now, to put that in context, that's a huge drop from 2021. So remember, 2021 was probably the peak of the pandemic years when we saw over $50 billion of VC financing. So 50 to 17 or 50 to 18, which a huge drop. But to put in broader context, 17 and a half billion is kind of consistent with the 2020 cloud funding number, which are 15 billion. So 15 billion invested in 2020, a huge spike or peak in 2021 of 50 billion. And in 2022 you can say 17 and a half billion is returning more towards kind of normalized VC investment in the cloud. Right.

So it's kind of to be expected. And do those lower figures translate to lower activity across the cloud ecosystem?

Not really. I think it's a combination that the bigger late stage checks or bigger late stage investments really have gone away. Right? So it's a, because of the macro economy and the public markets, those big dollar amounts have slowed down. But the smaller early seed series A checks are as busy as ever. So I would say back half of 2022, the first quarter, 2023, Greylock and a bunch of our peers I think are as busy as ever looking at cloud companies and other startups and we'll talk about some of those trends. But I would say there's a lot of interesting things happening. So the checks are smaller, which would be indicative of kind of a sea change, a lot more early, early stage companies, but the number has increased dramatically.

Right, and what's fueling all that?

Well, I think there's this thing you probably heard of called AI or large language models right now. Yeah, just, just you know, a little bit of two vowels and you know, a couple constants, AI and LLMs or generative AI. So I mean it's easy to say it's all caused by AI, but I'll, we'll talk about that, we'll talk about other trends because I think it's simplistic to say it's just all AI, but would say the past couple of years, you know, probably ever since Google published the attention is all you need paper, this whole concept of the transformer models have created a sea change of OR in the industry by this huge advantage of these large language models or foundation models. So we're seeing foundation models appear and that's like OpenAI's ChatGPT. There's a bunch of startups like Inflection, Anthropic, Adept, all building foundation models. Google and the big clouds have their own foundation models. Google releases Bard, they're kind of version of ChatGPT. So you will see there's a huge step function advancement in these foundation models which leads to a downstream effect of thousands and thousands of startups building on top of these foundation models or exploring all the changes caused by foundation models, everything from management, security, explainability. So I think we're seeing kind of this bow wave of innovation right now and it's pretty exciting.

Break that down a little bit more. So it's happening both within the startup ecosystem and within the big three, Amazon, Google and Microsoft.

Actually it's pretty interesting because you see the big three players, big cloud, we call them Amazon, Google, Microsoft, all fighting about AI. Now I mean if you look historically, castles in the cloud machine learning, AI was always the largest funded category, both by VCs and the large number of cloud services offered by the big three. So AI machine learning has been the forefront of the battleground between the big three. But also you see competition between startups that are kind of either playing between these giants or enabling. You know, these giants are riding the wave. So you can argue that Amazon was early on the machine learning AI wave with SageMaker, but then Microsoft clearly with a partners with OpenAI really has disrupted both Google in terms of search as well as Amazon as a cloud platform with this tight coupling of Azure plus OpenAI. So I think super interesting what's going on there. And you saw recently in early 2023, Amazon just released something called Bedrock, right? It's kind of its ability to connect to different foundation models from anthropic, etc. So Amazon's trying to democratize access to these foundation models like it did to different cloud services. So it's say the battle between all three of these players is super interesting. You know, Microsoft kind of did the parts of OpenAI, Google responded with Bard, and then Amazon's trying to play, hey, you come to Amazon for all things developer experience, come to Amazon for all things AI experience, right? But I'd say it's interesting. The big three are probably ones to follow, but I would say recently we've heard a lot more about Oracle clouds, the oci, both as a cloud platform like large startups like Uber, etc. Doing some capacity on Oracle, but in terms of training large language models, I hear from a lot of founders, engineers that OCI is actually a pretty amazing cloud to train these large models on. And then I would say the stock that I think we all wish we bought, you know, five years ago was Nvidia. Right? Because between the crypto boon a few years ago and the AI boom, the need for GPUs kind of drive all the training of these large models has never been greater. So I would say on top of just the hardware, Nvidia's released their own cloud, dgx, they released more software on top of their own hardware. So I would say there are the big three, Oracle making some headways into the cloud space, both in just cloud platform as well as around AI. And then key parts of the ecosystem like Nvidia really kind of creating a beachhead. And they're a part of every conversation.

Right now we're talking about AI exists in multiple layers, kind of like the cloud. And so there's these different places for startups to play in. Like where are you seeing this happening on a startup level?

Well, you know, at the greylock, we love layers and frameworks obviously, but simplify it down because this is a very complicated and evolving stack. We're looking at what we believe a new application stack evolving. So gosh, Heather, 30 years ago almost we had the Lamp stack. It was Linux, Apache, MySQL, PHP. That was kind of the default set of tools. You built any web app and then you had kind of the mean stack with Mongo, Node, JS, AngularJS, kind of these modern Webby mobile apps. We had Jamstack the past few years, kind of the, you know, Netlify, Vercel, kind of these JavaScript applications. I don't know what it's going to look like going forward, but for sure we're seeing evolving stack or application architecture around these large language models apps out there. And so we're super excited about how this develops. But for sure there's one layer that is the foundation models. OpenAI cohere anthropic stability. AI companies out there like Adept and Inflection might play in that space. But for sure you see startups and big companies play in the kind of the low level. You're going to see the top level application innovation. So Jasper Even up and Legal Tome and Presentations Harvey and Legal AI applications. Just so many companies out there take advantage of these advantages in AI to build applications. Close to the Greylock portfolio is something called Instabase that started document understanding and AI around there. They're really expanding on top of these foundation models. But this middle layer in between is kind of an evolving middleware stack. So we're seeing companies and technologies like Llama Index that's really defining the category for data ingestion. Data indexing for data querying is like this memory data layer around these AI apps because these, these large language model apps need some level of data integration and memory. We have technologies like Lang Chain and Fixie that are really pioneering how they think about building agents, doing prompt development and prompt operations. And then vector databases, right? Vector databases have been around for a long time, but all of a sudden the idea of storing and querying these vector embeddings is becoming a key piece of all these applications. So you're seeing comes like Pinecone, we've chroma all kind of take hold in terms of this kind of embedding stores. And then when you have a new stack, right, just like when you build mobile apps or cloud apps, you break a bunch of things. Got to fix these things, you got to monitor these things. We're seeing starts like Helicon and Honey Hive around Elem monitoring right there be there be more monitoring Tools, security tools, and management tools kind of all around this new application stack. So look, if we had kind of LAMP stack years and years ago, I don't know what the stack looks like in next three or four years, but we're seeing an evolving set of best practices on how to build these applications.

Got it. That's a lot to keep track of. And then outside of AI, there's some other things that are fueling more of these trends. What are you seeing for sure, Heather?

We're seeing a lot of more innovation, but let's just take a step back and we're still in the early, early innings of cloud migration and companies going cloud native, right? So for us, it might seem like cloud has been around from the beginning since like pretty much all the founders we deal with are born in the cloud and no one's been building their own data centers for years. But just remember, in terms of the total compute storage application used out there, we're still early, early days cloud. But within cloud development, we're seeing a lot more evolution and specialization. What we're calling industry clouds or vertical clouds. So you're seeing different industries from healthcare, government, transportation, they have either a their own needs, right? Like application needs, or they have their own kind of compliance, security needs or data sharing needs. So you think about vertical SaaS, companies around healthcare, government or solution products, around productivity, around developer tools, around security. So you're gonna see some specialization. And what's happening is, is both the needs are evolving and how to build these applications, but also the cloud providers themselves can now cost effectively build specialized services around these different verticals, different industries. And so Jason, who's one of the vester teams, wrote a lot about industry clouds and how they're evolving as well on our blogs and podcasts. So it's super interesting to see how this is going to be. Probably the next evolution of cloud is these verticals and these specializations, right?

Specialization sounds like a good place for startups.

It's a great place for startups, right? Because if you think about how you build sustainable moats, for example, and defend against your competitors is you have to either a own kind of a domain or you have to own a set of data that's proprietary. So think different verticals are different solutions. And a vertical could be like a market, like healthcare, government or financial services, or it could be a vertical solution like around CRM, customer support or B2C support, or owning one problem. So if you own a data set or one workflow around an industry or one problem, you can Actually kind of build an end to end solution and in the cloud that becomes a very viable, very attractive business model.

That sounds like it's getting more complex, more security needs. Obviously it's like an ongoing trend we see across every industry. So what does that mean for cloud?

Well, security is great because it's an evergreen market. Like I said before, typically what happens when you have a new technology, new platform, like moving to mobile, moving to cloud, first thing first is you enable new applications to be built, but also you break a bunch of stuff, right? And so cloud really enabled a bunch of new stuff, but it broke a bunch of stuff. And the first thing try to do is try to fix what's broken with the tools you have. Security is an evergreen need because when you go cloud you create a bunch of new security vulnerabilities or security needs. First thing you do is try to fix these security needs with existing security tools. But largely when we see these platform shifts like going cloud, the old stuff you have, especially on security isn't good enough. And you see this explosion of these security companies. So last year we saw that over $3 billion were invested into security startups in the cloud. So they put over $3 billion into the security subsector, which is huge, but that's only half what they put in 2021. So that's a huge number. But they put between the last two, three years, you know, north of over $10 billion in security startups. Because that is going to be an evergreen need. As you go cloud, as you build AI to applications, as you go to different verticals, you're going to need to build new security solutions.

So should startups just like always expect that enterprise organizations are going to have limitless budgets for security and it's always going to be something they're going to invest in. And these are, there's just ample opportunity for everybody.

You know, I would say it's not ample opportunity for everybody because unlike a lot of great markets, there's always replacing the old stuff, right? So there's definitely security budget because security and compliance is, you know, it's a top three priority for every CEO. It's a top three priority for every public company board and you're going to see dollar shift from old technology and platforms to new ones. So you see companies like the cloud security ventures like Wiz growing faster than any security company seen a long time. Companies like in our portfolio, like abnormal around email security is growing super fast. As you think about, you know, how email and identity changes when you go to the Cloud. There's a whole bunch of other companies now we're seeing around AI security that are kind of being born in the cloud now, born in this AI generation. Increasingly, as you build more and more application in the cloud, you care about what we call SBoM, your software, Bill of materials, what are the ingredients, where the open source projects you actually built your application out of which are trying to understand the provenance of all this code. So I think it's exciting for me as an investor because we see huge opportunity, but because there's new challenges and also the old technology, the old companies often cannot make that transition. So it's a really attractive area for great founders to have some deep IP to attack. And I think security is one that we're constantly looking at because we think there's always going to be innovation.

And so what are you expecting throughout the next year, you know, generative AI to build its own cloud and then its own security company.

So yeah, if the AI companies could build their own stuff and, you know, we're kind of getting there with projects like AGI and auto GPT, right. Kind of this, you know, AI building AI or you know, they're coding themselves, improving themselves. I mean, I think we're still, luckily for me, I still have a job because we're still a little bit away from that. But we're definitely seeing advances around developer productivity. Right. So some of these AI things can make your developers more productive. So you're going to see actually an interesting phenomenon where startups can actually probably develop farther and faster on the funding they have in different markets. So one, exciting two, we're still in the business investing because we still need to pick the best founders, the best opportunities in front of us. But I think we're excited because I think 2021, it's all a huge spike in venture investing. 2022, kind of a lower level, but return to kind of normal levels historically. And I think 2023, we're probably about the same pace as 2022. I mean, check back with me in a year, Heather, when you do, you know, the cast on the cloud looking back in 2023. But so far I think 2023 will probably be, you know, close to where we were last year. But we're going to see a bunch of new companies be the activity we're seeing just in the first, you know, three, four months of this year around the AI stuff, around security companies, around some of these vertical clouds has been pretty interesting. And so we're super busy. I know my peers and other Firms are super busy and so I think it'll be fun to see kind of how those bubbles on our Castle cloud chart have changed. You know, my early prediction is the AI ML1 is still going to be the biggest circle on our two by two. For those listening, check out Castle in the Cloud on, on Greylock. We have kind of a great visualization of World of Money is going to. So I think security and AI will still be two of the biggest bubbles. I think you're gonna see a lot of activity at the seed in series A at activity level for all the VCs and startups. And then you're gonna see a lot more coming from the big three. Right. I think Google, Amazon, Microsoft are gonna, you know, double, triple down on their efforts in AI. Google, for example, just reorganized their Google brain and their DeepMind research teams to one org. So you're gonna see, I think those big players come out. I think you haven't heard a lot yet from other companies like an Apple or Facebook Meta or Salesforce, which has their Einstein team. So I think a lot of tension around OpenAI around Google, but I think there are a whole bunch of other large companies out there that have great, great resources around AI that will be making big moves this year. That's also going to cause waves, but on those waves will be great momentum for these startups to serve.

You know, as we're talking, obviously this is all reminding me of your very popular essay, the New Moats from a few years back. In that essay you talked about why systems of intelligence are the next defensible business model. You made some pretty prescient observations at the time. This is six years ago, which is like a million years ago in technology. But you said startups using AI would be the ones to build the next business mode. And it all seems to be coming true with the developments in the past year. I'd love to hear if you're going to dive into that further. What can we expect?

Yeah, I think six years is two lifetimes in technology, Heather, but for sure, I mean, I was talking to a founder of the other day about what we got right with the new Moats and what we got wrong with the new Moats and how we had this framework of the system of record, which is your big ERP CRMs. You had the system engagement, which is like your user experience chat or the browser or the mobile app store, and had what we call system intelligence using AI to kind of power these new applications. And I think it's worth revisiting, you know, we got a lot right thinking about this business models. I think you're seeing a bunch of new companies building system intelligence. We obviously got a lot wrong in terms of like, you know, who would have predicted these large foundation models, who have predicted kind of chatgpt as the system engagement. For example, you know, we mentioned chat like Slack as a system engagement for some applications, but now ChatGPT really kind of predicted made chat a legitimate system engagement for your day to day interactions. That plus these plugins that OpenAI is releasing has really created a new kind of stack for applications. So you know, I think it's probably worthwhile rethinking what we got right and wrong. But I still believe system intelligence, these AI powered applications really are core to building what we call the new Moats.

Very cool. Well, I'm excited to hear the next wave of the New Moats, so I look forward to that. Jerry, thanks so much for being here with me today. And again for our listeners, if you want to visit our Castles in the cloud projects@greylock.com castles Jerry, thanks so much.

Thanks a lot, Heather.