Speaker A
So grateful to have you. This has now become kind of our annual one-on-one, our absolutely performance review.
Navin Chaddha discusses the transformative impact of AI infrastructure, the intelligence grid, and the future of AI-native technology stacks.
Key Takeaways
- AI infrastructure is foundational to the next wave of technological transformation.
- Value creation currently lies in the 'picks and shovels' of AI plumbing but will move up the stack over time.
- Claude is emerging as a critical platform for AI-driven work and productivity.
- Innovators should focus on groundbreaking, net-new solutions rather than competing on speed or cost alone.
- Physical and environmental challenges like power and cooling are key bottlenecks for AI scalability.
Summary
- AI is a 100x force that will transform how humans work, live, and play by democratizing knowledge and intelligence.
- The intelligence grid is being created, analogous to historical technology waves like electrification and the internet.
- Foundational layers of AI infrastructure include compute, memory, networking, and cloud packaging, with power and cooling as critical challenges.
- Companies building the plumbing for AI applications, such as Nvidia and Claude, have become essential players in the AI ecosystem.
- Claude is positioned as the new browser for work, acting as a gateway for human productivity and AI collaboration.
- The AI stack will evolve with specialized models for different tasks like diffusion and code generation, moving beyond current transformer architectures.
- Founders should focus on creating net-new innovations rather than incremental improvements to existing solutions.
- The future of AI infrastructure involves overcoming physical constraints with advancements in clean tech, power, and cooling technologies.
- The AI-native stack will enable new roles, personalized software, and human-AI teammates that enhance capabilities.
- Adoption of AI tools like Claude is accelerating rapidly, similar to historic technology adoption curves such as Linux.
Full Transcript — Download SRT & Markdown
Speaker A
performance review of Is there anything you first I'm very open to feedback. If there's anything you want to say that I could do better in in our performance review here that we we seem to find ourselves in. I think
Speaker A
Is there anything? First, I'm very open to feedback. If there's anything you want to say that I could do better in our performance review here that we seem to find ourselves in. I think
Speaker A
Thank you. Appreciate that. We're We're going to talk about AI infrastructure. We're going to talk about Open Claw.
Speaker A
you guys do a phenomenal job, which I'm sure everybody in the audience acknowledges. Just keep doing what you're doing and amplifying the impact you're having.
Speaker A
first and foremost, AI is a 100x force. It's going to transform the way humans work, live, and play. There's a lot of fear on what AI is going to do. But whenever there is a new technology, there is
Speaker A
Thank you. Appreciate that. We're going to talk about AI infrastructure. We're going to talk about OpenClaude,
Speaker A
before. So as with any new technology, when a wave comes, and I won't just focus on the information technology wave, let's compare to the electrification of the world. Electricity is the essential of everything we do. Without these lights, without these mics, nothing
Speaker A
personalized software, and the new roles that are coming in AI and that are really here. But let's just start. Where can the biggest value come from along this new technology stack that's been created by AI? Absolutely. Right. So
Speaker A
So, when the intelligence grid gets created, whether it was the cloud wave, whether it was the client-server wave, whether the mainframe era, the companies which build the plumbing on which the applications can run, or the agents can run, become immediately important.
Speaker A
first and foremost, AI is a 100x force. It's going to transform the way humans work, live, and play. There's a lot of fear about what AI is going to do. But whenever there is a new technology, there is
Speaker A
the compute layer, uh followed by memory, followed by networking companies, the foundational model companies, which three have become extremely big, and then finally, all this stuff get packaged in the form of cloud providers, so that it can be
Speaker A
pain. And the pain, if it isn't there, means there isn't going to be any long-term gain. So I'm a fundamental believer AI will manifest itself in the form of human teammates and allow us to do things which humans weren't able to do
Speaker A
And over the next few years, this is going to democratize the use of knowledge and intelligence, and value will start moving up the stack, if you will. So, the opportunities are abundant. It so happens today, people who provide picks and shovels are the
Speaker A
before. So as with any new technology, when a wave comes, and I won't just focus on the information technology wave, let's compare to the electrification of the world. Electricity is the essential of everything we do. Without these lights, without these mics, nothing
Speaker A
don't try to be a faster, better, cheaper version of what exist. See across the horizon and say, "Hey, what can I do which is net new to change the world?" And that's where massive wealth gets created, massive impact happens. So, let's look at
Speaker A
would be possible, right? So we are entering an era which is of intelligence, and the intelligence grid is being created.
Speaker A
than the cloud. It creates other issues. So, if the key thing around CPUs and GPUs is going to be everywhere, it's creating massive demand for power, massive demand for cooling, and network and memory are the bottleneck. It's no longer the compute.
Speaker A
So, when the intelligence grid gets created, whether it was the cloud wave, whether it was the client-server wave, whether the mainframe era, the companies which build the plumbing on which the applications can run, or the agents can run, become immediately important.
Speaker A
Cooling technologies are going to be extremely important. Power loss technologies are going to be extremely important. Similarly, the world will move from copper to optics.
Speaker A
So, if you look at the first wave since the launch of ChatGPT 3 to 4 years back, the value has equated to essentially the plumbing layers, to the picks and shovel providers. Some mega companies have been created. They start with Nvidia at
Speaker A
if you look at models, basically, they're doing a phenomenal job on the transformer architecture. But, are we done? No, we aren't done. For AGI to happen, there going to be specialized models which will happen for diffusion, new models will come up for code,
Speaker A
the compute layer, followed by memory, followed by networking companies, the foundational model companies, which have become extremely big, and then finally, all this stuff gets packaged in the form of cloud providers, so that it can be
Speaker A
discuss what do the builders in the crowd and what do builders in the world do? It is incredible how Claude, even just in the last three or four months, seems to have grabbed the hearts and minds of essentially every engineer. It I know
Speaker A
consumed by people. It's really like you had the electricity grid, you had the railroads, you had the roads, the information highway, or the intelligence highway, has just gotten built, and value has equated to the plumbing stack.
Speaker A
they've now come and said with what Claude has done the last few months, "I was wrong" or "I'm sorry" or "I've changed" and I am now going full speed with what Claude is doing with me. How did that happen and and where do you see
Speaker A
And over the next few years, this is going to democratize the use of knowledge and intelligence, and value will start moving up the stack, if you will. So, the opportunities are abundant. It so happens today, people who provide picks and shovels are the
Speaker A
Claude, the way I look at it, I would give two analogs. One for humans, it's really a companion and an AI teammate which not only is making me efficient but is allowing me to do things that I wasn't able to do before. The second
Speaker A
biggest companies. How do I avoid building something that just in 6 or 12 months just immediately gets absorbed by one of these huge AI companies? Yeah, absolutely. So, I think my advice to founders in the audience, and in general, has been
Speaker A
is going to throw darts at me now. I think uh Claude is the new browser for the internet. And what I mean by that is when Netscape happened as the first browser, it organized the internet. It became the gateway
Speaker A
don't try to be a faster, better, cheaper version of what exists. See across the horizon and say, "Hey, what can I do which is net new to change the world?" And that's where massive wealth gets created, massive impact happens. So, let's look at
Speaker A
So Claude I think is becoming the gateway for work. Where whenever it comes, we saw this with the iPhone too with the App Store. Some applications for work will be handled by Claude.
Speaker A
the plumbing infrastructure. It's very hard to go compete now with the AI accelerators just looking at the bottom of the stack with companies like Nvidia. But this wave on the creation of the intelligence grid is orders of magnitude bigger
Speaker A
And Google search appeared where you could search for what you were looking for, your intent. Now, with Claude, I look at it as a gateway to how a human gets their work done.
Speaker A
than the cloud. It creates other issues. So, if the key thing around CPUs and GPUs is going to be everywhere, it's creating massive demand for power, massive demand for cooling, and network and memory are the bottleneck. It's no longer the compute.
Speaker A
And that's what is going to happen here. And let's see if it starts building capabilities of runtime.
Speaker A
So, the next new waves are getting back to physics, fighting laws of physics, and you need alternative sources of energy. So, clean tech was dead. It's going to be back in vogue.
Speaker A
It's the runtime component. And in 15 years, what took Linux, it took 29 days for developer adoption to happen for Open Claw. So, both these could coexist.
Speaker A
Cooling technologies are going to be extremely important. Power loss technologies are going to be extremely important. Similarly, the world will move from copper to optics.
Speaker A
Android, and where does value get created? But if you look at the App Store or the browser, it went free, and it essentially value got created with search, value got created with e-commerce, the whole SaaS ecosystem get built. So, I'm very
Speaker A
And there will be next-generation memory architectures and networking architecture. So, that's just an example of where value at the hardware layer has happened, and where it's going to go in the future. You move up the stack. Now,
Speaker A
We saw them early on partner with big companies, let's say like a like a company like a Figma or some of these other like big tent pole um uh sort of functions in the a worker stack, but now we see them
Speaker A
if you look at models, basically, they're doing a phenomenal job on the transformer architecture. But are we done? No, we aren't done. For AGI to happen, there are going to be specialized models which will happen for diffusion, new models will come up for code,
Speaker A
that will sort of be open to playing with a lot of people? Do you see them more like an Apple that's more closed where they're doing a lot of these things themselves? And does it matter?
Speaker A
there'll be prevalence of vertical AI models, whether it's for cybersecurity, whether it's for healthcare, whether it's for physical AGI, and this advent of these physical models. So, innovation essentially is not stopping and with Claude everywhere, I'm sure we'll
Speaker A
played in three decades, uh nice sound old, which is okay, having participated in three waves from the internet to mobile cloud, and now the AI era, I think we need to be patient. It's a chess game. So first and foremost, I
Speaker A
discuss what do the builders in the crowd and what do builders in the world do? It is incredible how Claude, even just in the last three or four months, seems to have grabbed the hearts and minds of essentially every engineer. I know
Speaker A
With the browser, it essentially became similar. And the advent of SaaS from desktop or client server apps essentially went. So the idea you can own a UI is going to become extremely extremely hard. You have to endorse this stuff. So what should you
Speaker A
even in my own experience with my own engineering team, we had early AI pioneers that were really driving it. But I have had people who are incredible engineers but maybe have been slow to adopt AI and
Speaker A
Second, you need to have a data mode, which creates context and knowledge graph for what you're trying to do.
Speaker A
they've now come and said with what Claude has done the last few months, "I was wrong," or "I'm sorry," or "I've changed," and I am now going full speed with what Claude is doing with me. How did that happen and where do you see
Speaker A
And own distribution. But you need to essentially endorse the cloudification of the world and not fight it. And the same thing is going to happen with open cloud. So let's endorse this wave rather than fighting it. Let's draw parallels
Speaker A
Claude going? Absolutely with engineering? Absolutely. So, first and foremost, it's not just going to be engineering. It's going to happen in most cases of knowledge work.
Speaker A
Right? So we need to keep an open mind, learn from the past, make old mistakes, but be open to making new mistakes. It's going to be continuous AB testing. And now products, as you said, don't take 6 months, they They take 3 months. They
Speaker A
Claude, the way I look at it, I would give two analogs. One for humans, it's really a companion and an AI teammate which not only is making me efficient but is allowing me to do things that I wasn't able to do before. The second
Speaker A
headless software as a service? Yeah, so till now uh let's look at websites, let's look at SaaS apps, which became massive, which were all productivity. They were essentially we were using software to make myself productive as a human. Mhm.
Speaker A
thing, the analog which I haven't talked about earlier that I was thinking coming in today, is, you know, I write a lot and I'm going to preempt what I'm going to write in a few weeks. My marketing team
Speaker A
UI is different, the login is different, it's complicated. Very, very complicated. Now, uh I'll borrow from what the Stripe CEO said, we're entering essentially the pizza model of software delivery, where I say my intent as a salesperson is to prepare
Speaker A
is going to throw darts at me now. I think Claude is the new browser for the internet. And what I mean by that is when Netscape happened as the first browser, it organized the internet. It became the gateway
Speaker A
The agent goes, does all the work, and preps me for the meeting. And I didn't go to any SaaS app.
Speaker A
or the access point to all websites and all information. Now, the world is moving from me using computer apps, websites, SaaS apps to getting work done.
Speaker A
So, the DIY era, where the human has to learn all these interfaces, remember all the passwords, download reports, create dashboards, and get intelligent, is getting over. So, that's where essentially this headless and personalized software is going. I have
Speaker A
So Claude, I think, is becoming the gateway for work. Wherever it comes, we saw this with the iPhone too with the App Store. Some applications for work will be handled by Claude.
Speaker A
It's talking to all these softwares through plugins, skills, MCP, some other buzzword tomorrow that you and I will be scratching our head on and just gives me what I need. So, that's what is changing and it creates immense pressure on
Speaker A
So, coding is a good example. It will get into other areas. So, the world on the back end has to connect to this thing. So, we're going to enter a wave of headless apps where similar to the browser, websites were back ends.
Speaker A
Who have a data mode, but this is where new startups are going to come. How many of the client server software companies were able to successfully transition to the cloud era with SaaS. SaaS didn't exist.
Speaker A
And Google search appeared where you could search for what you were looking for, your intent. Now, with Claude, I look at it as a gateway to how a human gets their work done.
Speaker A
your business model changes just from going from perpetual licensing to seat-based licensing created huge difference. The same thing is going to happen now. I charge per seat, it's subscription. The new startup is going to is will be headless,
Speaker A
And the interesting thing will be whenever any platform technology comes, for it to become big, it has to do certain things itself, but then has to build an ecosystem around it for others to flourish.
Speaker A
endorse these platforms which are emerging. Look, with iOS and Android, millions of apps exist. They created a whole ecosystem. Yes, Microsoft owned in the operating system world some productivity apps.
Speaker A
And that's what is going to happen here. And let's see if it starts building capabilities of runtime.
Speaker A
They figure out what incumbents don't have the talent. Don't have the guts. They're public companies to retool themselves. It's a green field opportunity. So, go attack the incumbent rather than fighting what the new browser or the new operating
Speaker A
And the operating system, which is where the difference between Claude and OpenClaude in my mind is. OpenClaude is more the Linux for agents.
Speaker A
So, when big ecosystems get created, you do certain things yourself, but you build a network of developers and builders who support your platform.
Speaker A
It's the runtime component. And in 15 years, what took Linux, it took 29 days for developer adoption to happen for OpenClaude. So, both these could coexist.
Speaker A
months to find the right people to work with, and then 3 months to get a prototype, and and without any iteration.
Speaker A
One could be the operating system, one could be the browser, where things become headless, and the question becomes in this fast-moving thing, what does the platform own? W
Speaker A
that they wanted, and then the customer success person goes and begs the product team to include it, and they take it to the road map meeting, they meet with engineering. All this takes weeks, months, and then finally it gets
Speaker A
into a sprint, but we're 2 months out from, you know, the sprint to get into an open sprint, and then it takes somebody 2 3 months to develop it. You know, if you're lucky, you might be 6 months before you get what you, you
Speaker A
know, this request you had now. Now it's I or where the future is going, I need this. An agent is the product manager. It speaks to me. It mocks it up for me. It does the back and forth. It's
Speaker A
all the discovery. It takes that and puts it into a Jira ticket. The other eight another agent comes in. It takes that ticket. It builds it. It puts it into review. Another agent uh looks at that review and approves it and gets it
Speaker A
into production. And that'll happen maybe in a span of 1 day. Absolutely. And I call this the collaborative intelligence era where humans and agents or machines will work together to work at blitz scaling to create things that humans essentially on their own
Speaker A
would have taken much longer, but more product, more innovation, more wealth. So, everybody wins. And then my call with Vibe Coding to entrepreneurs in the room is in the past, only developers could start companies, right? Now, everybody can be
Speaker A
a builder. Everybody can be an entrepreneur. This is the great equalizer. So, this is a call to action to people to jump in and leverage this technology that is democratizing a lot of things and focus on how you're going to solve real
Speaker A
problems for real customers own distribution and own some competitive moats, but build on these platforms rather than fighting them. The the the jobs that existed uh in these early-stage companies was one thing and it they've completely changed now with AI and there's there's
Speaker A
different roles. This thing that I just described, you've got these you know, we had six six customer six support, we had product, we had uh front end engineering, back end engineering. We had people doing code reviews and QA. I
Speaker A
mean, there's all these different things and now all of it's kind of melted together. What are the most critical roles that you see in an early stage company for somebody to to take on a adapt or to bring those kinds
Speaker A
of people into their company? Yes, so I think this is where new leaders, new managers are going to emerge who will know how to get the best out of humans and leverage these technologies which are companions to be able to do things
Speaker A
much faster at a fraction of the cost and keep innovating. So, this is a call for humans essentially to reskill themselves in an era like we used to talk about globalization offshoring and teams had to manage talent here, they had to manage talent
Speaker A
in India, they had to now the world will be hybrid. Some of it will be agent, some of it will be humans. So, you need to rise above the noise and essentially become architects first and foremost of what the intent of
Speaker A
your company's trying to build and then become orchestrators of essentially which music humans are going to play and which instrument the agent is going to pay.
Speaker A
Endorse this, start with architecture and then get into orchestration, but get closer and closer to in this world what you should be doing. When cloud happened nobody bought servers to build a SaaS app. Nobody bought the network, nobody bought power. It was
Speaker A
available as an API. Same thing is happening now. The manifestation is it does intelligent work.
Speaker A
I don't need to use these tools to make myself more efficient. I just delegate, hey, you go do this.
Speaker A
And that's what is going to happen. These services to me are like APIs that I need to endorse, architect, and be a system thinker, and then sit back and orchestrate the stuff.
Speaker A
But at the end, if I'm a B2B company, my product is essentially getting bought by a human. So, the human-facing skills, the management skills, are going to grow.
Speaker A
And there was a time developers, I don't know how many in the room were there at that time, had to learn machine language, assembly language, then like C, then C++, now you just sit in natural language and say this is my intent.
Speaker A
And it's much faster. And if you look at system programmers, they were a million 20 years back when I was in school or doing entrepreneurship.
Speaker A
They're still a million, and they're 29 million application developers. Now it's going to a billion.
Speaker A
So, skills are getting democratized. Sky is the limit on what can be done, and that's my call to entrepreneurs and builders. Sky is the limit. Jump in. Endorse this stuff because you have the advantage which somebody else is still scratching
Speaker A
their head, and incumbents are going like, what do I do? They don't want this stuff to work. They don't want this stuff to work. And like you have to come and endorse this, write this, and build the company faster than they can at a
Speaker A
fraction of the cost, and let it show up in your pricing. And that's going to reimagine how work gets done. And there are budgets.
Speaker A
There are budgets which are essentially going to shift to you. So, that's the big opportunity and this is a 100x force. Please endorse it. Don't be afraid of it and ride this horse. You are the jockey. Ride this horse.
Speaker A
Naveen Jain, may you have fun. Thank you. Thank you. [applause]
Topics:AI infrastructureintelligence gridClaude AINavin ChaddhaMayfield FundStartup GrindAI-native stackfoundational modelsAI computeAI productivity
