Ali Ghodsi, Chief Executive Officer and co-founder of Databricks, explores groundbreaking developments in artificial intelligence technology within enterprises. This discussion, held at theCUBE's new studio in New York City during the NYSE Wired event, illuminates how AI factories transform data centers into operational hubs.
Ghodsi delves into their expertise in AI application with John Furrier, co-founder and Co-CEO of SiliconANGLE Media. They explore key topics such as the role of Agent Bricks at Databricks, the importance of ease of use for developers, and AI's ability to complete essential tasks within enterprises. TheCUBE Research and video host John Furrier bring insightful questions to the discourse on innovation and AI integration.
Key takeaways from the discussion highlight the importance of accuracy and human involvement in AI applications, as emphasized by Ghodsi. They discuss how organizations such as AstraZeneca and Adidas leverage AI to enhance operations like drug discovery and product design. According to Ghodsi, the future of AI involves commoditizing large language models to bolster application development, emphasizing governance and security.
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
Ali Ghodsi, Databricks
In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value.
Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.
In this conversation at the theCUBE + NYSE Wired: AI Factories – Data Centers of the Future event, Ali Ghodsi, chief executive officer of Databricks, joins theCUBE’s John Furrier in the NYSE studio to unpack how practical enterprise AI is reshaping the data center story. Ghodsi explains why Databricks’ Agent Bricks focuses on automating the most mundane, high-toil tasks inside organizations instead of chasing abstract superintelligence, arguing that we already have “enough AGI” for these workflows if reliability is nailed. He shares concrete examples, from an...Read more
exploreKeep Exploring
What questions were asked regarding the functionality and focus of Agent Bricks in relation to AI?add
What are the current challenges and future possibilities of AI implementation in the enterprise sector?add
>> Hello, I'm John Furrier with theCUBE. We are here at the new CUBE Studios in New York City. The NYSE is theCUBE's home on the East Coast. We'll be opening up the East Coast Tech TV Silicon Valley connection with our Palo Alto Studio. Of course, CUBE alumni is here, Ali Ghodsi, CEO of Databricks. Ali, great to see you in our new home here on the East Coast. We're going to bring that tech content here to Wall Street.>> That's awesome. Yeah, this is beautiful. Amazing.
John Furrier
>> Yeah. And you were just talking to Jim Cramer, so you did a little Mad Money hit, that old film tonight.>> Yep.
John Furrier
>> Did he get into the business side? What was he asking? What was some of his questions?>> He asked about agents. He's saying, "Are they working?" I kind of pointed out that with Agent Bricks at Databricks, what we're trying to focus on ... A lot of others are focused on super intelligence, really smart AI that can solve programming, Olympiad level things. We really focused on the most mundane, boring tasks that you're doing inside of an enterprise. So think of it as the most toil things that no one wants to do, and we just want to automate those, and that's what we're starting with. So Agent Bricks, we just focus on that. And the way we do it is instead of testing it on math Olympiad questions, we test it on, "Can you fill out the Salesforce entry correctly? Can you extract this data? Can you summarize this? Can you call into that system?" So very simple agents that we build with Agent Bricks and we just evaluate them on that. And if you just focus on these simpler tasks, you can do remarkable things. I actually think we already have all the AGI we need for these kinds of tasks. We just have to make sure that they work with high reliability.
John Furrier
>> At your summit, Data + AI Summit, one of the big themes was ease of use for developers, but also getting knowledge around the data itself. How's that going and what are some of the results you're seeing in the field? People are talking agent to agent a little bit out there, but some of these low hanging fruit use cases just get it right and accurate. Accuracy is critical. What's the progress since?>> Yeah, it's amazing. Kind of three cases, but the biggest one is these where you have a human in the loop and we'll have a human that reviews the results. So for really important things, like something is going out to a customer, communication, or in the medical field, clinical trials and so on, you have a human reviewing it. But there's also some of the things that we couldn't have ever a human do that we cannot do automatically. I'll give you an example. AstraZeneca, they do drug discovery, sifted through 400,000 documents with an agent. Those 400,000, no human being could have ever done that, so no group of human beings would've able to communicate that with each other, and that helps them actually do drug discovery faster. So use cases like this, we're seeing more and more. Adidas is another one. Adidas is actually now empowering their designers that make the shoes to actually directly have access to agents that have summarized the sentiment across the whole globe in different languages and what they think about the exact shoe designs that they have. And that lets them then innovate and come up with newer designs that they know will actually work in these different markets.
John Furrier
>> I was also talking with Pete Sonsini who started Laude Ventures and Andy and Pete collaborate on this research. I don't know what to call it, almost like an open source incubator. I don't know what to call, research. Talk about the role of these kinds of research hubs that are becoming not so much applied research, but really directed, and how does that affect some of the innovation? What's your vision on that?>> Yeah, and I think it's actually really important, because we're seeing a lot of open source models coming out of China. And here in the US, we don't have that many. There's a few, there was one model open source by OpenAI. There was few models by Meta, but I think it's fair to say that on open source, China is head of United States. So I think what Pete and Andy are doing is excellent in that we should have domestic open source models that then lets us do innovation, do research on those. So we're big fans, big supporters of that. In general, I think the LLM layer gets more and more commoditized either from the open source coming from China or here. So I think that's good for everyone because once that layer gets commoditized, it helps the next layer, which is the applications on top.
John Furrier
>> And accelerates things.>> Yeah.
John Furrier
>> I was talking to my friend John Markoff about a decade ago. He wrote a book called What the Dormouse Said. It was about the computer revolution and how the hippies and the one, just time-shares. And it was kind of a spawned by research from the Cold War, from the Russian-American Cold War, and a lot of Lincoln Labs, Draper Labs, Cal, UCLA, they're all working on these grants. We don't have that right now. And we're in an AI build-out, and the private sector seems to be funding it. And people, entrepreneurs are stepping in like Pete and Andy and others. Does that change the view of capital structures and funding? Because there's a lot of big debate about this. I see it's a good thing for America. NVIDIA doing a deal with Nokia. That's not Huawei, that's Bell Labs basically.>> Yeah, exactly. I think we're going to see changes happening here going forward. What we're going to see is that open source Chinese models are going to disseminate, and what you're going to have is that you'll see research labs and universities using those Chinese models. And as a result, I think Western nations will say, "Hey, we need our own models to counter this." So I think then we need more open source models. Well, private industry doesn't want to open source as much because they want to monetize. So as a result, what we're going to see is I think we're going to see more and more ... Like we used to have it, as you said, Lawrence Berkeley National Lab, Livermore and so on, all of these. In fact, most of the supercomputers for the last 40 years, the largest computers on the planet were publicly funded in these kind of labs. Not public companies, not private companies. So I think we're going to see a resurgence of that to counterbalance the Chinese open source models.
John Furrier
>> It's kind of in a weird way, tech for good, it depends on what lens you look at it. I want to get your thoughts on the enterprise. I know you don't have a lot of time. Appreciate you taking the time when you visit the exchange. The enterprise has been slow to adopt by some say, "I don't believe the stats about failed projects. All experimentation fails at some level." But there's been definitely a focus on, "Hey, we can't just put AI into the enterprise."
As you said, there's some toil that needs to be and brittle IT systems. You guys have really nailed the data lake. You also have huge client base with OpenAIs and the neoclouds as data lake customers. As you have that knowledge, what's it take for the enterprise to get on the fast track and get that flywheel going? What are you seeing? What are key blockers? What's the accelerant?>> Yeah. I mean, look, the AI models I think are already powerful enough. I think we already have AGI, actually. We just moved the goalposts. But if you had asked us at UC Berkeley 2009 when we were working on these things, what does AGI look like? And he showed us what we have now. We would've said, "Yeah, that is AGI."
Now, we moved a lot, because we don't like some of the results. And we're like, "No, this is not quite." And now, we're also coming up with new names like ASI, super intelligence and so on. But really, we have all we need. What's missing then? Why are these enterprises? What's missing is, one, we need to bring the context of what's happening inside of the enterprise with proprietary information. That's not public on the web. This is, how do you operate inside of a company? What are the standard operating procedures? What's the sensitive data that you have that you haven't put on the web? So the large language models don't understand it. So we have to bring that context to them. And when we do that, we have to do that in a governed, careful, secured way. So I think governance and security is extremely important for any enterprise. So that's been actually our number one priority the last three years, because how can we give context to these LLMs with very sensitive information while at the same time make sure that there's no leakages, that we have the lineage, that we can have an audit trail and that we're documenting it? Because there's also laws coming up everywhere. Everything from EU AI Act, to those going to be regulation here in the US. So I think security and governance in one sentence and getting the context inside of the enterprise to the AI.
John Furrier
>> Yeah. My final question goes to CEO Ghodsi and Professor Ghodsi. If the AI factories continue to get bigger, the CapEx, enterprises adopt and we believe there'll be a hyper-converged edge box where all protocols to all spectrum and ethernet converge with a mini AI factory box. When you have full hyper-conversion at the edge and all the systems redesigns at the superhuman level, what happens next? Because you have the keys to the kingdom looking at the data layer because all you got to do is put a mini factory somewhere in between. What does that look like? What happens when that convergence happens at the edge?>> Yeah. Well, look, I think we are moving towards that future. And are we also moving towards the future where we have quantum and so on? Are we going to have models that are so small that we can actually run them at the edge and so on? We're working towards that. But honestly, the thing that keeps me up at night now is much more mundane. It's like we already have pretty smart AI. Can we just make it useful right now in the enterprise? Because right now, they're not actually being used as much as they should. Given the capital investments and given what people are actually doing with it, there's a big difference. I just want to make them applicable. I think all this other stuff will happen too, and it's exciting.
John Furrier
>> Yeah, that's headroom. On the enterprise, quick follow up on that. What would be that key thing that you would wave the magic wand right now in the enterprise? What's the psychology? Is it technical debt?>> No, I think that they get it. It's really security and governance. Making sure that you can actually ... Because there's a tension here. This is very sensitive data. Can we give it to the large language model or can we not? Well, what happens if we give it to it? So there is a tension. And if we have enough guardrails around it and if we can do audit tracking and we can have the lineage and all of that, so that governance is extremely important for enterprises. That's really what we need to nail. And then how we evaluate whether we succeeded or not. I built something, is it successful or not? Would you say I think it's successful? Do you think the way to evaluate it is to have benchmarks and evaluations on the actual task that we're doing? Not on math Olympiad, physics Olympiad. Humanity's last exam is the name of the hardest exam that anyone ... No one has got more than 30, 40-
John Furrier
>> Practical evaluation.>> Yeah. Did you do this mundane task correctly or not? That's what we need to nail. So benchmarks and evaluations on the actual tasks.
John Furrier
>> I guess Jensen was right, agents will be the HR department. Check the evaluation.>> Yeah.
John Furrier
>> Thanks for coming on our CUBE here in New York City.>> Yeah.
John Furrier
>> Breaking it in. Thanks for coming on. Ali Ghodsi making an appearance inside our new studio here at the NYSE, NYSE Wired and theCUBE partnership. Thanks for watching.
John Furrier