AWS Marketplace Newsroom | Jerry Chen, Greylock

Clips
More from AWS Marketplace Newsroom

Speakers

play_circle_outline Transition from traditional cloud computing to AI-enhanced cloud services.

play_circle_outline Exploring the Impact of Generative AI and Foundation Models on Business Solutions and Startup Opportunities

play_circle_outline Navigating the Rapid Tech Innovation Landscape: The Crucial Role of Observability in AI-Driven Companies Like Chronosphere

play_circle_outline Shift in enterprise strategy emphasizing AI alongside existing cloud strategies.

play_circle_outline Investing trends towards AI infrastructure and business process digitization.

Info
Transcript

Jerry Chen, Greylock

In this broadcast from AWS re:Invent 2025, Jerry Chen, general partner at Greylock, joins theCUBE’s John Furrier to analyze the industry’s transition into the "agentic era." Chen argues that cloud computing is evolving into a "cloud plus AI" paradigm, where AI-native applications act as solvers to digitize complex business processes. The discussion unpacks the role of agents in augmenting human workflows, handling long-running reasoning tasks and driving the next wave of enterprise value. Chen also reflects on market winners, citing the $3.4 billion acquisiti... Read more

explore Keep Exploring

What is your perspective on the growth and opportunities presented by AI native companies? add

What is the relationship between AI native companies and traditional business processes? add

What is the significance of Chronosphere in the context of AI and observability for cloud native and AI native companies? add

What has changed in business strategy focus over the past 15 years? add

What aspects of the AI market are currently being considered for investment and development? add

bolt Powered by CUBE AI

Jerry Chen, Greylock

search

John Furrier

>> Welcome back everyone to theCUBE's live coverage here at the AWS re:Invent 2025. I'm John Furrier, host of theCUBE. This is our 13th year covering AWS re:Invent. It's been a wild ride. We've been seeing all the innovations, cloud abstracting away the infrastructure. Hello, cloud computing and SaaS. Now, we're in agentic era, abstracting away work with agents, opening up a whole new level of cloud-like patterns. And here to break it down is Jerry Chen, general manager, general partner at Greylock Ventures. Jerry, you've been on theCUBE every year's tradition.

Jerry Chen

>> When is it?

John Furrier

>> When no one knew what Amazon Web Services was or theCUBE. Who the hell is this show?

Jerry Chen

>> It was a bookstore selling compute. What was going on?

John Furrier

>> We called it.

Jerry Chen

>> Oh, come on, John. Thanks. First, happy birthday, belatedly. So everybody, it's John's birthday. And, yeah, it's been amazing to kind of watch this show grow from a hotel ballroom into basically a small city. I feel like it's interesting, with all the AI stuff, it's clearly what's changing technology right now. You can say cloud is dead, long live cloud. It's like you're seeing a transformation from what Amazon and all the cloud vendors were before to what they will become, which is basically cloud plus AI.

John Furrier

>> You and I both have talked about the different opportunities. Obviously, you're making investments in early stage startups. By the way, congratulations on Chronosphere, three point something billion.

Jerry Chen

>> Thank you.

John Furrier

>> Six years coming from the Chronosphere team. We interviewed them at KubeCon many moons ago.

Jerry Chen

>> Yeah. You and Martin at KubeCon years ago.

John Furrier

>> We're like fellow travelers. But I love your perspective on because you have to look at connecting the dots to identify the diamonds in the rough, so to speak, to see which team, which market is going to be there, and we covered the cloud. We saw that SaaS market and good commentary. We're kind of in a similar ... Every few couple years, six years, so Amazon has to kind of shift. This seems like the ground is changing because the game is still the same. It's not a pivot off cloud native or SaaS. It's just an extension trajectory, accelerated trajectory with AI. So it's got the same kind of pattern. Infrastructure as a service, platform as a service, but it's different. It's got AI native. What is your take? Because a lot of startups that are called AI native have a different orientation. Their building blocks are different, but they're still building blocks.

Jerry Chen

>> Sure.

John Furrier

>> What's your take on the AI native growth, new company opportunities?

Jerry Chen

>> I think AI native is a broad term. I think everything is AI native, just like everything that was cloud native. And to your point you made before, they're kind of adjacent on the spectrum. You have cloud native, and I say AI native is cloud native. So these AI companies are using generative AI models, foundation models, or other smaller or distilled models to do things that we didn't think was possible before, machine learning. And quite frankly, what it is, is solvers always solving a business problem. It's taking a business process like order to cash, hire to fire, and you digitize it. What it is now with AI native, you can address a whole rational problems that are business processes that were handled by software plus humans. And because what AI is doing is augmenting the human part of the equation. And so now, you're seeing more and more applications being built that are "AI native." still have some traditional software workflow, but adding the foundation model element to it to add reasoning, to add actions, to solve this long fat tail of business problems.

John Furrier

>> When you look at the adjacency, because I like that leverage because when you're adjacent to these new markets, you get advantages. Are there winners and losers on this wave that maybe didn't make it? There's a lot of speculation that there's going to be kind of a harvest or culling of the herd, so to speak. For the folks that didn't cross over, some will get acquired like Chronosphere into Palo Alto networks, which is coming you and put the first check in again, three point something billion is not an acquihire. That's legit.

Jerry Chen

>> Yeah.

John Furrier

>> It's a legit acquisition. Obviously, Palo Alto Security. You can connect the dots on that, obviously. It's the obvious. But like when there's some companies that might not make it.

Jerry Chen

>> Sure.

John Furrier

>> And then also, there's another concept that we've been kicking around theCUBE called, there's no room for fast followers even because it's so fast. If you're not in that first wave, you can't really wait because waiting is like, what's a strategy in the kind of slower tech scene where-

Jerry Chen

>> I don't think it's ever been slow, John.

John Furrier

>> Well, I mean, six months, a year, let's see how the market develops. I mean, AI is obviously happening. What's your take on those two things?

Jerry Chen

>> On the first one, Chronosphere is a great story. And you should talk to Martin. It is a company that basically sold observability to cloud native companies at the largest scale, highest amount of data volumes, but they did it cheaper, faster, better than their competitors. And what they were able to do is sell not just the cloud native enterprises, boring the cloud companies, startups, large enterprises, but they're also starting to sell to you the AI native companies. So they actually have two of the largest foundation model companies as customers, a bunch of other AI native startups start growing superfast. So just because you're using foundation model, it doesn't mean you don't need management observability. Chronosphere crafted code saying, "Hey, the largest, biggest, fastest growing companies out there need observability." And Chronosphere was providing that. And so Palo Alto Networks smartly saw, they were able to crack the code and sell to AI native companies and acquired them for $3.4 billion is a great story.

John Furrier

>> So when you look at investments now, has this world changed just the investment dynamic?

Jerry Chen

>> It's always changing. A comment on the speed question you threw out there, what's changed is that speed. What it is, is we have a window opportunity. Not necessarily bulbar wave, but call it the overtone window of enterprises adopting software has kind of shifted and broadened that every vertical, every enterprise now has an AI strategy mandate. 15 years ago it was like, "Hey, what's your cloud strategy?" That's what we're talking about. You got to move from on-prem to cloud. Now, boards around the world are saying, "What's your AI strategy?" And so the reason why speed matters and also as an investor matters is you have to take advantage of this curiosity, the forward leaning interests of your customers to actually adopt AI. Because if you're not there to serve that need and be their AI thought partner, somebody else will.

John Furrier

>> So Matt Garman told me the shift is not about bigger models, it's not about cheaper inference, it's not even about scalability, it's about agency. And he says, "The next 80% to 90% of enterprise AI value will come from agents." So he's kind of shifting the narrative to, of course, it's scale, they have scale, but like the agency, the creativity of agents. What's your reaction? Do you think the enterprise is unlocking and this will unlock more enterprise? I mean, not everyone could be a neocloud or a hyperscaler, you don't have to CapEx, but you bring a Nova Forge, you bring me an open foundation frontier model for not a frontier price with my data set. I'm interested. You have my attention. Does that play out? You got frontier agents, the word frontier implying scale, multi-step reasoning, all these kind of bigger items. What's the enterprise vibe there for you?

Jerry Chen

>> For sure. I mean, look, I think to what we were saying earlier that the AI native portion is AI doing the work of humans, which is basically augmenting what humans do. It's two things. One is agent work, like filling out forms, browser and web actually doing work. Number two is all reasoning, right? So a lot of the business processes, the long fat tail was that email would come in or someone would call you up and like John or Jerry had to look a bunch of data, make a decision and open another application. So the agency there are these long-running processes that do or make that work faster, easier, cheaper that was normally done by humans. And so the agency is solving those problems, but I think Matt's right in that agent to agent actions along these long-running actions are kind of the future of these applications, but cost and scale still matter. I think tokens per watt is clearly going to be a metric that we all think about in the next 2, 3, 4, 5, 50 years, who can actually provide foundation models and these tokens cheaper, better, faster.

John Furrier

>> And they got some work on the silicon. The word sovereign cloud is in their AI factories announcement, which by the way, I really love because we've been talking about AI factories for over a year. Amazon capitulates. They're now talking AI factories, which implies on-prem, but hybrid cloud, but the word sovereign's in there. It's not like sovereign country, but like sovereign seems to be the word for governance and policy and like private cloud-like activities. So I'm not saying it's private cloud, but like private cloud has all these things in there. So in a way, sovereign by default is somewhat private or custom. Am I reading into that, or no?

Jerry Chen

>> I think you're connecting some dots here. So first and foremost, data gravity still exists. One of the advantages of Amazon is all your data is in S3, all your data is in their databases. So data gravity still is true. On-prem, data gravity rules, right? That's why there's a bunch of applications still on-prem. Well, I have mainframes. I have databases like Oracle. DBT is still sitting on-prem. So that's all true. So then the question is, what do you define as sovereign? It's probably some privacy, some control, some governance, some visibility. And you can basically solve all those problems in the cloud, like a VPC or BYOC environment can do those on-prem for sure. I think you will see a rise of on-prem stacks to solve that problem. And you can talk about nation states if you want to, but there'll be different ways to solve that problem because it's not like 12 years ago where all the data was in point A, on-prem, moving to cloud. Now, you have so many companies that were born in the cloud. So their on prem is S3. The on prem is Azure. Their on prem is GCP. There's no data centers to go back to. So they want sovereignty as well. They will want a BYOC solution or something like that.

John Furrier

>> So multi-cloud fits there. So talk about your view on the competitive strategy between like, okay, Nova was slow to the game, their language model. OpenAI, Google are the ones that have the ones that I think are relevant. Obviously, Claud is great.

Jerry Chen

>> Anthropic is doing a great job. Yeah.

John Furrier

>> They're doing great jo, but they're on all those clouds. But Amazon has hyperscale competitors. OpenAI is not yet a hyperscaler, but they're on their way to be. They're buying data centers. They plan to build chips. They got an ambitious roadmap. What's your take of Amazon Web Services' strategy to open up Nova a subsidization? Is it a play to keep the customers? They pull in a Microsoft move where you keep your customers happy, just get some ROI, keywords in the keynotes. ROI, transformation, stay on AWS, don't leave.

Jerry Chen

>> I think Amazon, you've talked to Jassy, you've talked to Olsavsky, you've talked to Matt Garman. It's all from Jeff Bezos' philosophy, customer first, right? It's like customer, customer, customer. So I think Nova and their whole suite of products is all trying to serve and meet the customer where they're at. So let's think about what Amazon has. They have scale, they have capacity. So they can serve you models and their goal is all models, right? I think is they want to solve all models all the time. Anthropic, Nova, OpenAI, they want to use the open source models as well, Llama, et cetera. So they want to basically serve whatever their customers want. But they also understand that two things. One, they got to keep their partners honest by offering Nova, which is kind of their version of a model for customers to use and customize. We talk about the pros and cons there. But they want to offer a large menu of choices for customer, because it's not one size fits all. I think that you want to believe a long fat menu can actually help their customers.

John Furrier

>> One of the quotes he had also, "Generic tokens are useless unless they know your business." This is the big, I think, thing I like about this re:Invent is for the first time, they're saying your data, the gravity for the customer, whether it's in S3 or on cloud, Matt even said an earlier, "99% of customers won't buy AI factories." Implying the other guys. They'll use the service, AI factory service, aka, the Amazon Cloud, agent cloud or whatever you want to call it. So the personalization around smaller models. I talked to the Reddit people about their use case, they were using Nova. They got amazing results by tapping into the Nova half-baked model and making it their own on little things like content moderation, stuff, hard fricking problems and without a lot of heavy lifting. So that's like value. So if you can train your business. I mean, that's-

Jerry Chen

>> I think as you raised, it's basically ... I totally believe in trading the models, be it like fine-tuning or the new trends, reinforcement learning, right? Create these RL environments or do reinforcement learning on my data, my environment and a per enterprise, per business use case. I think the other question is, is that for 10% of the customers, 20% of the customers, 100% of the customers? And I think that the question we don't know is this using your data and doing the tuning with Nova or doing something like RL environment, RL tuning as a service for enterprises. What percent of customers will benefit from that from a base generic model? And then there's a race. If you're one of the guys with the big models, of course, you can say, "Hey, it's fine. You can use Nova today, but in a generation, our models can be good enough to content moderation out of the box." And I think that's been true for a lot of use cases, that the next generation models just basically suck up things like coding, completion, speech, stuff like that, where fine-tuning, some of these other techniques don't make sense. Then to your point, John, okay, so what are the things that are left that will always need my business process, always need my data? What are those use cases and how much will customers pay for those things? And I think there's a valid market there, there's a large market there. And that's why we're in this business, which founders and which companies will grab that bell, if you will?

John Furrier

>> Yeah, 25 years in Palo Alto, Silicon Valley format is really simple. The startups talk to the other growing startups, they become customers. There's a little bit of a keiretsu and vibe there. But spending time in New York, opening up our NYSE studio, I've noticed there's a lot of enterprise startups in New York. Because you just walk down the street or take the subway, you got 20 customers. So the AI enterprise market is hot for startups right now, which is great for us because we love it. It's our market we cover. So the question for you is, how hot is the enterprise market, and does it stay hot for startups?

Jerry Chen

>> I mean, I'm bias, I do enterprise software investing, so I think it is hot.

John Furrier

>> It is hot.

Jerry Chen

>> It will stay started.

John Furrier

>> It's legit. 24-year-old kids are working on it.

Jerry Chen

>> Well, I think that's the most interesting thing. I was talking to some young founders the other day and in a generation ago, the younger founders would do more consumer experiences, right? But now, a lot of young founders, be it college grad or college dropouts or people in grad school, they're doing B2B enterprise problems. And it's almost like either A, maybe there are no new consumer experiences to be created and that's not true. Or there's no new social graphs. Who knows? Talk about TikTok and Instagram and the rest. But for sure, this generation of founders, especially young ones, really see the value of taking AI and solving a business problem. And it's neat because not only are they connecting the dots, but the feedback cycles of using AI to solve hiring, business analytics, security, it's so fast. They're getting the dopamine hit that people got from likes from solving real business problems. So I think it's hot. I think it's great. I think New York is a great market. We have investments in New York. We have a small office in New York, and so I think it's great to see founders everywhere.

John Furrier

>> Yeah. It's fun. I think that's a good observation about the young kids because they watch their dad's work and we can fit. I don't want to do that.

Jerry Chen

>> Don't make us .

John Furrier

>> But there's also some pragmatic in this generation between ice baths and health kicks. They're on. They see practical value of the business side of doing it right, like making money and not trying to get the holy grail. It's not a lot of go big or go home in the enterprise. You got to grind a little bit, but the payout's good and they see those business problems. So I think that's interesting. Now on the consumer side, I had a debate with Dave on theCUBE Pod about this. I think there will be massive consumer apps. We just don't know yet what that is. I think it'll have to come out of the woodwork.

Jerry Chen

>> We're certain to see that with Sora and Veo and some of these video generation things and some of the music things like Suno. I think we're going to see new media created. I haven't seen a new social graph created yet, but with maybe new devices are coming out. So you have like the Ray-Ban glass, et cetera. Who knows what Johnny Ive and OpenAI is working up in terms of these wearables, whatever. But I think you're going to see new communities come up, new graphs be created. I mean, it's going to be fun. So I'm looking for that as well.

John Furrier

>> Rise on micro-communities is a new interesting abstraction, I think, opportunity. All right, final question. Great to always chat with you. What are you working on? What are you excited about these days? Where's your attention in terms of like just interest, areas you think might explode, gut feeling, and/or legitimate market targets?

Jerry Chen

>> I mean, look, I think first, we're always founder driven. So if you ask any of you to see what they're investing in, they're going to tell you what these just solve that day. The truth of the matter is, look, it's AI or die, right? It's an AI generation. We're looking at all things AI, both from the infrastructure, coding, like everything is code, right? That's why Cursor and these other things are doing so well. It's regardless what business problem, healthcare, enterprise, software, industrials, it's all code. So code and code adjacent things would definitely be a market for us. And then finally, like I said, these business processes be it verticals like healthcare or legal or industrials or horizontal problems like HR. We are seeing a new generation of software companies attack both these verticals and horizontals. And it's a little intimidating, overwhelming sometimes. I mean, you walk out there on the expo floor, but it's actually a great time to be an investor, a great time to be a founder.

John Furrier

>> Do you like vertical models? I mean, we're seeing like Harvey AI pick a vertical and crush it like legal for instance.

Jerry Chen

>> I love verticals. We love horizontals. They're just business processes, right? So what do we say before? Software is taking a business problem, business process, digitizing it. And with AI, it's going from step A, B, C, D, and E. And with verticals, those four or five steps are just in legal or just in healthcare or just in like home services. They can also be an HR, ERP, IT service management, help desk. But what AI does is they say, "Step A to step E, let's do it faster or God bless it, let's skip two steps in the middle, go step A to Z immediately." So we love all.

John Furrier

>> Jerry, great to see you. Again, we're part of the 20-year anniversary party when we have seven years from now on Amazon Web Services, our 20th year. Our 13th now, but yeah, great to see you.

Jerry Chen

>> I'll see you in seven years.

John Furrier

>> Always a tradition. Jerry Chen on theCUBE, always a tradition. The evolution, it's AI or die. This is what it's all about, and it's a huge opportunity for startups, creatives. AI scales intellect. If you got the data gravity, even better. It's in your head too, that's data. And of course, we're doing our part to share the data with you in theCUBE. Thanks for watching.