In this interview from theCUBE + NYSE Wired: AI Factories – Data Centers of the Future event, Glean co-founder and CEO Arvind Jain joins theCUBE’s John Furrier to unpack what’s really working in enterprise AI today and what comes next. Jain explains why knowledge access remains the first successful AI use case at scale and how Glean’s enterprise search brings AI into everyday work. He details the past year’s lessons with AI agents – from the need for guardrails, security, evaluation and monitoring to democratizing agent building so business owners (not just data scientists) can create production-grade agents.
The conversation dives into Glean’s vision of the enterprise brain powered by an enterprise graph, highlighting the importance of deep context, human workflows and behavior to reduce “noise” and drive outcomes. Jain outlines core building blocks – hundreds of enterprise integrations and a growing actions library – that let agents securely read company knowledge and take actions across systems (e.g., CRM updates, HR tasks, calendar checks). He discusses how organizations are standing up AI Centers of Excellence, prioritizing “top 10–20” agents across functions like engineering, support and sales, and why a horizontal AI data platform that unifies structured and unstructured data – accessed conversationally and stitched together via standards like MCP – sets the foundation for AI factory-scale operations. Looking ahead, Jain says Glean’s upgraded assistant is evolving from reactive tool to proactive companion that anticipates tasks and accelerates productivity.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for AI Factories - Data Centers of the Future
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for AI Factories - Data Centers of the Future.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Mitesh Agrawal, Positron AI
In this interview from theCUBE + NYSE Wired: AI Factories – Data Centers of the Future event, Glean co-founder and CEO Arvind Jain joins theCUBE’s John Furrier to unpack what’s really working in enterprise AI today and what comes next. Jain explains why knowledge access remains the first successful AI use case at scale and how Glean’s enterprise search brings AI into everyday work. He details the past year’s lessons with AI agents – from the need for guardrails, security, evaluation and monitoring to democratizing agent building so business owners (not just data scientists) can create production-grade agents.
The conversation dives into Glean’s vision of the enterprise brain powered by an enterprise graph, highlighting the importance of deep context, human workflows and behavior to reduce “noise” and drive outcomes. Jain outlines core building blocks – hundreds of enterprise integrations and a growing actions library – that let agents securely read company knowledge and take actions across systems (e.g., CRM updates, HR tasks, calendar checks). He discusses how organizations are standing up AI Centers of Excellence, prioritizing “top 10–20” agents across functions like engineering, support and sales, and why a horizontal AI data platform that unifies structured and unstructured data – accessed conversationally and stitched together via standards like MCP – sets the foundation for AI factory-scale operations. Looking ahead, Jain says Glean’s upgraded assistant is evolving from reactive tool to proactive companion that anticipates tasks and accelerates productivity.
In this episode from the NYSE studio, theCUBE’s John Furrier sits down with Mitesh Agrawal, chief executive officer of Positron AI, to examine how AI inference is emerging as the economic engine of the next infrastructure cycle. Agrawal shares updates on Positron’s rapid growth, including its recent Series B funding, major customer deployments and expanding role in large-scale AI factories. The conversation frames inference as the “killer app” of the AI era, where performance, power efficiency and capital discipline are becoming decisive competitive advantage...Read more
exploreKeep Exploring
What recent developments have there been at Positron and in the AI inference industry?add
What was the primary driver of investor appetite for your Series B round—growth, revenue, product, or something else?add
What product are you designing, what problem does it solve, which applications does it target, and how does its memory capacity compare to existing chips?add
>> Welcome back. I'm John Furrier, host of theCUBE here at the NYSE CUBE Studios. Of course, we've our Palo Alto studio connecting Wall Street and Silicon Valley. It's our AI Factory series, part of our AI Leaders Podcast series, ongoing series around featuring the leaders. Mitesh is back. Mitesh Agrawal, CEO Positron AI. Congratulations on the continued success. Great to see you again. Your CUBE alumni back.
Mitesh Agrawal
>> Thank you, John, and thanks for having me again. I really appreciate being here.
John Furrier
>> I love the Silicon Valley-Wall Street connection that we've done with theCUBE Wired, which is a CUBE original brand that we've created. And the program has been very popular because of the leaders, like yourself, sharing. Sharing the content, sharing the data. So, first I want to ask you, what's new? What's changed since we last talked? What's up?
Mitesh Agrawal
>> Yeah. Well, if you look at Positron itself, what's changed is we launched our first product since we last talked. And Positron is a massively revenue-generating company now. And we raised our series B that we recently announced with pretty big strategics as well involved. And we are running towards getting our second silicon out for AI inference, which is all the buzz right now. And then, if you look at the AI inference as an industry, you're seeing a whole class of AI inference chips that are coming into the space saying... Obviously, NVIDIA is the absolute behemoth in the space, but then all of this chip companies are really coming through and saying that, "Look, depending on the workloads, this is going to get more and more bigger and bigger because every workload, the more dollars you save, the more electricity you save, the better it is for those workloads to grow."
John Furrier
>> Just to mention the raise, I think the notes were $230 million series B.
Mitesh Agrawal
>> Yeah.
John Furrier
>> Pretty hefty, over a billion dollar valuation, which could have done better in this market, maybe. I'm only kidding. It's a great number. Inference is obviously the killer app. We saw NVIDIA acquire a quasi accuhire, Groq. Takes them off the table. To me, that's a tell sign. Inference, we predict it's going to be the killer app. But what's going on now? Because you have AI factories, you got large-scale clusters, you've got the hyperscale, you got the neoclouds. You have now a massive AI infrastructure build out where it's not just neoclouds versus the hyperscalers are on-prem. It's like, where do I get the energy and the horsepower to run inference? And we haven't even talked about the edge coming. The edge is going to come in full force.
Mitesh Agrawal
>> Yeah. No, I think you hit the nail on the head, really. Saying that inference right now... And let's just talk about the future, right? I think we all know what's happened and people have discussed... This year you're going to see a huge amount of multimodality. So, video generation. Just yesterday or two days ago, Seedance 2.0 model launched, people are calling it the DeepSeek moment of video generation, content generation. And not just from a content perspective, but world simulation, world building. Building out of factory floors in simulation before even you do anything real with it. So, you know exactly what are going to be the problems of it. All of this application space is just going to really take off. And what that means is more compute needed, more energy needed. And so, like obviously, I'm an energy maximalist, we need as much energy, but this is where putting Positron into the part of the story is we are a very energy-efficient chips. For every watt of power, we are basically trying to get three, four, five times more output for inference. And this is how we find a story into this overall narrative around, well, we need as much energy, but for the given amount of energy, if you can drive more output, the better it is for the production growth that is out there continuously.
John Furrier
>> Energy drives tokens. Tokens feed intelligence. This is where it's going, adds revenue. Agents will eat tokens and get more intelligent. Robots and robotics massively-
Mitesh Agrawal
>> Physical AI.
John Furrier
>> Physical AI is not just robotics, it's everyday life. Moving around as a human, the robot's roll. How do you see that playing out? And will it be a parallel growth or is it phasing in? Do you see more enterprise with agents and robotics more of a clean sheet of paper?
Mitesh Agrawal
>> Well, in terms of order, I think the first order of business this year will be agent growth. So, when you hear all the AI labs talk about it, they talk about automation of software roles and agentic workflows coming in a lot of the digital life of things. I think there's really a lot of development in the models for robotics, what the data is being fed into it and how the robot models will get better. But in terms of really practically impacting our day-to-day lives, I think that will come a little bit later, but already software development, media development, all of those things, that's where you'll see the first really iteration of the growth.
John Furrier
>> Mitesh, what was the driver for your series B appetite from the investors? Growth, revenue, product?
Mitesh Agrawal
>> The big one is that we launched our first product and that while we'll be announcing the specific customer and things in the coming months, but we are one of the first AI silicon companies to actually get a massive deployment with a very, very large customer. We already announced Jump Trading, Cloudflare, PowerSale as our customers. Jump Trading actually co-led our series B bid, became our customer first. They saw the application of the chip and they said like, "Look, we really love what you're building. We would like to actually kickstart your series B and do that."
John Furrier
>> All right. So, the folks learning, explain what the product is and how people are using the product.
Mitesh Agrawal
>> Yeah, the product is we are designing inference chip and system to really focus on inference that is very memory-intensive. And what that means in terms of what people think about applications is video generation, code generation, very high context requirements for inference. These are memory-intensive tasks. People are talking about memory wall, memory wall. This is what the memory wall is in the growth of AI inference. And so, our silicon will become the first ever terabyte plus memory capacity chip in the entire world. For context, NVIDIA Rubin, that comes out later this year, 384 gigabytes or 0.4 terabytes of memory. Positron chip that tapes out end of this year world's first 2.3 terabyte memory or 2,300 gigabyte chip, right? So, that's the product and that chip goes into a system, that is the inference system and those are the applications-
>> Yep. And look, people talk about edge. I want to talk about that. There's one edge which is like the mobile and at the device, wearables, all those things, that is separate. Those chips are going to be different types of silicon and that's where people are saying some tasks will be done on the edge and some in the data center. The other part of edge is being closer to the population centers. Instead of having this massive data centers in the middle of nowhere, you're having data centers and buildings closer to the cities, just like content distribution network like Cloudflare is and that's the other side of the edge.
John Furrier
>> It's like internet days of access points.
Mitesh Agrawal
>> Correct.
John Furrier
>> You had the MAE-East, MAE-West, had these SuperPOPs.
Mitesh Agrawal
>> And the analogy I love is people always worry too much that one thing will overtake another, like edge will overtake data centers or vice versa, but it's always small brain, big brain, both are needed. Your use cases on your mobile and laptops will grow, but also, your use cases in your data centers will grow massively. So, it's actually both of them will see massive growth in the coming days.
John Furrier
>> All right. So, what's on your to-do items now? What are you optimizing for? You got your B round, you got a lot of cash, business is good. Is it robotics? Where's the application? Where's your use cases?
Mitesh Agrawal
>> So, the biggest thing for us is our tape out of our second generation that is end of this year. And then, really plugging that in to the application of where we think that the biggest application growth is video generation and world simulation model, which will feed into robotics, training datasets and all those things. So-
John Furrier
>> Because it's a prerequisite for-
Mitesh Agrawal
>> Correct, exactly. It's a prerequisite to build out-...
John Furrier
>> understanding the reasoning and inference of work.
Mitesh Agrawal
>> Yeah, you need real world physics for robots to understand from. And world-building models, world-simulation models will be basically the baseline and the underlying-
John Furrier
>> And they need memory for that.
Mitesh Agrawal
>> A lot of memory.
John Furrier
>> So you bring up memory. I just interviewed some developers, Sam Partee with Arcade and Harrison Chase from LangChain, San Francisco guys.
Mitesh Agrawal
>> Those are the people focused on agentic application.
John Furrier
>> Now, they have a different memory issue, which ties into your memory issue. Their memory is what the context memory was-
Mitesh Agrawal
>> Context length, exactly.
John Furrier
>> So, there's two memory definitions, looks like there's two tokens, crypto and here. So, memory of what context is like what I'm doing on my history, my experience, agents.
Mitesh Agrawal
>> Correct.
John Furrier
>> You're talking about memory memory.
Mitesh Agrawal
>> Physical memory.
John Furrier
>> Like in the PC, in the servers, now the systems?
Mitesh Agrawal
>> Correct. But those two are intricately linked. The more physical memory you have, the more digital memory, which is what he's talking about, Harrison, because agents will be better if they have context of not just what you did last five days ago, but 50 days and so on. More importantly in the code generation side of things, it's like agents will be better, not if they just have your latest code repository, but what the company has built over five years, 10 years, so millions of lines of code. The more you feed it, the better it will predict the next code that it'll write. And so, both of those are so intricately linked. Your digital memory can only increase if your physical memory increases.
John Furrier
>> And we are using the example of OpenAI doing ads. And so, the ad model in the AI generation is going to look a lot different than Google Ad Words, just like Google Ad Words looked a lot different than a banner ad because it had context, a keyword in the search makes a result come. In AI, the context is, "What am I trying to do? I'm trying to shop, so I don't need a link. I need the answer." I need memory for that.
Mitesh Agrawal
>> It's not even just context of that particular moment that you're searching. AI will have context of like, "Oh, Mitesh just had a kid who was a year old." And already recommendations engines do that, but AI will take it to the next level of prediction setups here.
John Furrier
>> Yeah. And that's why I like the hyperconverged vision around the edge because you now can add more datasets. So, I think you're onto something with the simulation worldview because that's very translatable to other use cases. Your thoughts on that? Just riff on that concept because if I can do a worldview for robotics, why can't I create a worldview for me?
Mitesh Agrawal
>> Yeah. I mean, you actually were making me learn about it, the hyperconverge where not only is just like... Well, other part of that is actually, I do want to quickly just... The security part of it, the cybersecurity and all those things that will have to evolve to make sure all of this data, especially as we put our personal lives into the AI models and things like those. But I think having that quick low-latency through that kind of setup will be very critical as well because people don't like to wait. Speed is-
John Furrier
>> I think the point about security is interesting because functionally you could say, "Oh, I can see that world happening," whatever use case you see. But if you look at cloud, when I was talking to Teresa Carlson, who was the first executive to build out public sector for AWS in DC and she cracked the CIA and a lot of public sector. And then, Amazon had the startups first and then the enterprise. Startups were happy to have no data center, so they win that one, check. Amazon goes off. But they really took off when they nailed security and compliance. I have to ask you, AI is tracking similarly where it's highly accelerated. Security and compliance mean different things than the classical categorical definition. What is your vision on how security and compliance need to be thought about, the bar to hurdle over, the good-enough line? What's your thoughts on that?
Mitesh Agrawal
>> I think on the technical side, there are, first of all, a lot better experts than I am that are building in the security space for AI, especially... The thing here is like you cannot be reactive anymore, it has to be proactive. So, when you hear Anthropic doing their model testing around how the model will behave in certain scenarios? What if it's fed this prompt? They have to do that because they can't just wait until it does it in the wild and that's proactive. A lot of that is very much-
John Furrier
>> By the way, that strategy has helped Anthropic.
Mitesh Agrawal
>> Yeah, very much so.
John Furrier
>> And a lot of enterprise .
Mitesh Agrawal
>> Exactly, right? And that's on the extreme end of what models will do. But coming back to enterprises, they are so critical... Today, OpenAI, Anthropic, they have scraped every piece of web data, but where there's still tons of data available is in the public and private company market data, right? And companies won't want to hand that over until the security protocols are done very, very well. So, any company that thinks about that, that cracks that will actually have a big role to play. But from a technology perspective, there's a lot other better people than me to think about it.
John Furrier
>> Well, Mitesh, great to have you on. Got a very tight schedule today. I want to have you back in.
Mitesh Agrawal
>> I know.
John Furrier
>> I want to do a deep dive on this because I think the inference silicon angle is a really good one. I mean, we are silicongle.com, of course.
Mitesh Agrawal
>> Yes, exactly.
John Furrier
>> Kind of comes to the head. But I think the edge, I think robotics all converge in as edge cases, but huge demand for inference and simulation.
Mitesh Agrawal
>> Yeah. Just the amount of use cases will... It's hard to fathom the exponential multiplication of use cases. So, really have to-
John Furrier
>> Put a plug in for your company. Great. I'll shake your hand.
Mitesh Agrawal
>> Thank you.
John Furrier
>> But put a plug in for, your hiring? What are you trying to do? What's going on with the company?
Mitesh Agrawal
>> I think the biggest thing I'll say is Positron is working on... We are focused on the silicon for the next level of intelligence. And so, anyone that wants to work on making sure intelligence grows and is good intelligence. And one of the internal jokes that we do at Positron is if Positron makes a really great chip that intelligence works on, eventually when they're our overlords, they'll either really like us if they're nice overlords, and if they're terminators, they'll be like, "Okay, these guys did good. We'll not kill them."
Obviously, we just raised, so we are growing. People should look at positron.ai. Very, very interesting space to build in. I think the approach we're taking is different than a lot of the existing companies, NVIDIA, TPU, Amazon. And I think that's one of the interesting things to do. I think when you have some new way of approaching the problem that everyone is trying to solve, I think it makes for a challenging and fun environment.
John Furrier
>> Well, it sounds like a great culture. And getting to know you, it's been great. I think anyone who wants to join a company where it's thrilling, exciting, technically, and solving hard problems-
Mitesh Agrawal
>> Yep, exactly....
John Furrier
>> you're in the right spot. Super exciting. Thanks for coming on.
Mitesh Agrawal
>> Thanks, John. Really appreciate it.
John Furrier
>> I'm John Furrier, host of theCUBE our AI Factory series, doing our part to explain complex problems to you from experts. Thanks for watching.
John Furrier