theCUBE + NYSE Wired present the East Coast AI Leaders Executive Series | Carmen Li, Silicon Data

Clips
More from theCUBE + NYSE Wired present the East Coast AI Leaders Executive Series

Carmen Li

Founder & CEO

Silicon Data

play_circle_outline Silicon Data: Revolutionizing Financial Products for Compute Markets with Transparency and Risk Management

play_circle_outline Importance of transparency and predictability in pricing for stability

play_circle_outline Products include Silicon Navigator for market prices and Silicon Carbon for carbon footprint

Info
Transcript

Carmen Li, Silicon Data

Carmen Li

Founder & CEO Silicon Data

Carmen Li, founder and CEO of Silicon Data, started the company in April of this year, focusing on the lack of financial and data products in the rapidly growing compute market. Silicon Data aims to provide transparency, efficiency, and risk management for users in the financial services and technology sectors. The company is releasing two products, Silicon Navigator and Silicon Carbon, to help users make informed decisions about GPU chips and carbon footprints. Li's vision is to create a healthy and efficient market for compute products. She is currently hir... Read more

explore Keep Exploring

What was the inspiration behind the founding of this company and what do you guys do? add

What is Carmen discussing in terms of asset allocation and enterprise AI adoption rates? add

What are the details of the two products being released this Friday, Silicon Navigator and Silicon Carbon? add

bolt Powered by CUBE AI

Carmen Li, Silicon Data

search

>> Hello, everyone. Welcome back to theCUBE's coverage here in New York City. I'm John Furrier, your host of theCUBE. Dave Vellante was here last week, we're getting up and running in our East Coast super studio, super pop access point to theCUBE Network, of course, bringing you wall-to-wall coverage CUBE style. I'm here with Carmen Li, who's the founder and CEO of Silicon Data. Great name, I love the word "Silicon" in it, silicon angle. Of course, the team loves the word "Silicon," we love silicon chips, we love silicon innovation, compute, GPUs, XPUs, chips.

Carmen Li

>> Same here. We love everything silicon-based, too.

>> Congratulations and thanks for coming in.

Carmen Li

>> Yeah, thank you for inviting me. This is very exciting.

>> Okay, before we get into some of the meaty conversation around Silicon and what's going on there, explain what Silicon Data does. What was the inspiration, the founding story, and what you guys do?

Carmen Li

>> Yeah, so we are very new. We started April this year, we're six months old, so be kind to us. Really, the inspiration is from when we see the whole rush to Open AI to gen AI use cases and, from my point of view, which is pretty much financial services, financial infrastructure, I see a big lack of financial product development, data product development point of view. You think about a world where translate compute... Well, we're talking compute here, right? But that's translate compute to another asset class as a resource. Just like electricity, just like crude oil energies and natural gas, then you realize that the big difference in those two markets is everything else I mentioned, electricity, crude oil, natural gas, they're very mature. They have the best suite of financial products, data products to get investors, the users for price fluctuation, for use cases, transparencies with data, and then you look at compute, which is growing so fast. In a few years, I really believe that it will overtake any other energy market and there's really not enough financial products, data products to guide users, give transparency, efficiency to the markets. That's what we really need.

>> A bold vision, you're ahead of the curve. You're seeing a financial market, is that right, around the dynamics?

Carmen Li

>> A financial infrastructure and data markets, yes.

>> What would that look like? I mean, trading, GPUs? Some said GPUs are like gold. I mean, and you can't get enough of them. They're on allocation with Blackwell, with NVIDIA.

Carmen Li

>> Essentially, I believe any financial products and data products are really to provide the industry with a mechanism to transfer risks, to manage risks, and give people the transparency, what volatility and what risk exposure they have to help people manage their portfolio of assets and really predict for the future. If you put that lens down, I really think there should be mature financial products. We can talk about the products, talk about the other OTC products, and data products. Think about the market prices, start with that, and any other benchmarking. Carbon, we just talked about it, energy, anything related to compute as an underlying asset class.

>> Who's the target audience for your product?

Carmen Li

>> Great question. There are three major cohort, people, express interest. The number one would be the actual people who use compute. You think about data center owners, they own compute resources essentially, and people actually use compute. Any technology company, any large corporation, everyone is doubling down the whole AI workflow. That's the one sector, and the second sector are the financial services firms. You think about a bank needs to underwrite a huge data center financing. How are they going to manage their risk portfolio? How are they going to benchmark it? The revenue, forecasting. All the financial institutions need a product to help them manage risks, predict future cash flows, and then market makers.

>> Carmen, it's really interesting, because what you're bringing in is asset allocation with compute, and as enterprise AI just gets going, IBM quoted McKinsey saying only 1% of the enterprises are tapping into generative AI, you're going to probably see the leverage of assets and workflows, and pricing that in, I mean...

Carmen Li

>> My goal is to make the market as healthy and can develop efficiently. I mean, at least it's my personal goal. I really believe in transparent market will help industry, help everybody, help data center people, help people in the technology space and help people in... Every day. You and me can have lower cost of finding products we access to.

>> When you have a bold vision like you do and you see something early, and I want to get into what you saw early, there's always people who misunderstand it. My favorite quote was from Andy Jassy, now CEO of Amazon. He said for a long time AWS was misunderstood, and he said, "To be a really good entrepreneur when you're early is to tolerate being misunderstood, which is people don't get it." "We have to change the marketing." No, that's a good thing. It's actually a feature, not a bug. What are people responding to? Do people understand, and people who don't understand or misunderstanding what you're doing, how do you respond to that and what would you say if you have to clarify that?

Carmen Li

>> Right. Obviously, I'm not consumer products, I have a lot less client base than AWS, so my product is so niche where, if you're looking at... You can use my different product, different use cases. If you look at someone, if you are the center of excellence for big corporation, you're looking at ROI for any of AI investment. For you, okay, think about the cost, how much I'm paying for all the GPU for the training for the RAG? That's one component, you probably need some data exposure or you just try to figure out, "Okay, this is one workflow. I can put in 20 different data centers I currently have or I can utilize one of those hyperscalers, which one should I really do? Should I factor carbon as part of the calculation for the cost?" Same chips, different data center, different carbon footprint. Different electricity price, which they probably know, but also different carbon footprint depends on the energy sources.

>> Yeah.

Carmen Li

>> All those things factor in. Obviously, they're probably not going to in the future treatment products, however, they can use the data product to guide their decisions.

>> You're saying that if I have to make a decision on a major system overall, which they're doing with AI systems, like NVIDIA and others, that they have to look at the power envelope and factor that into the calculation and say-

Carmen Li

>> If you want a completely holistic picture, I'll encourage them at least to the calculation and you can take the next steps. I hate to plug, but most of my products are launching this Friday, which is quite exciting, and it's a beta launch. We want to give people the access and the transparency is also what-

>> Let's get into the product. What do you have right now? What are some of the current things you have on the table now?

Carmen Li

>> There's two products we're releasing this Friday, one called Silicon Navigator, I know you love the name. The second product is Silicon Carbon. Silicon Navigator, think about a giant market price for all the major 50 GPU chips available in the markets. Both retail prices, meaning if you buy them from the market, how much are you paying them? The second is rental prices. Let's say if you make a decision between purchase versus rent, and then you can do your own calculation. How many hours am I going to do that for? 100,000 hours? Should I buy it? Should I purchase it, or should I rent it?

>> Yeah.

Carmen Li

>> Right? And the second is the Carbon where you can put in which chip you want to use, where data center is, you don't have to tell me, just put in a zip code, or choose one for Hyperskiller, which you probably know, which one you are with, and tell me how many hours you're thinking about using it. Well, I'll tell you how much carbon footprint we can... Based on the estimation we have, we give the exact calculation, every single input, and you can tell me if data is wrong. We have to take feedback as well.

>> Are the sustainability mandates factoring in now into this? Is that important? Obviously, you've mentioned it three times. Sounds like that carbon footprint calculation is super important.

Carmen Li

>> I would like to think it's important. I'm not going to tell people what to do. They can do carbon trade, they can offsetting whatever they really want, they can just generate a lot of green energy bomb back.

>> Well, they have goals too, right? Don't they have their carbon goals, sustainability goals?

Carmen Li

>> I think that's probably the reason, that's my guess. People started working back, because you hear less about those talks, but the one thing I want to make sure is, number one, we have the data, and number two, we can figure out strategy, how to tackle that.

>> Not to change the subject, but I've been hearing a lot of arbitrage going on in GPUs, chips being sold in different markets, gray markets, black markets, so there's an underbelly arbitrage going on. Does that factor into anything or is that just market that you deal with, or is that not an issue?

Carmen Li

>> Well, I don't think it's not an issue. If you look at compute compared, again, to any other virtual markets, you will see markets where it's gray or it's problematic, and regulation is to step in. But the way I look at it is, okay, let's say you have one H100 cluster. They actually price very differently in France versus in Singapore versus in... Dubai can buy chips, but in Middle East somewhere, and there's a reason for that. Do you want to ship your data over to train the model or do you want to just be, "Hey, I'm fine with US east."

>> Yeah, yeah.

Carmen Li

>> Right? Availability's an issue and then, obviously, people, they buy and sell clusters, but however you think about NVIDIA's, the way they service their clusters and it's also a component of consideration.

>> Yeah, risk management comes up a lot. Governance is going to be a big part of the data and how this all factored in. What was the generation of the idea? What was the genesis? Did you wake up one day, you've just been itching at this problem, you saw an opportunity, you recognized the opportunity? Take me through that process.

Carmen Li

>> Yes, it's actually very intuitive to me, and I'm sure to a lot of people out there as well. I'm from a very traditional background, financial services. I started my career off as a high-frequency trader at DLW Trading, so we create trading platform systems for energies and fixed incomes and index teams, and I realized that this is a market that gives a lot of transparency, efficiency, and people can get in and out position very quickly, very cheaply.

>> Yeah.

Carmen Li

>> Everybody knows about the price, which is, I think, the small market really needs. Then I take a look at the compute markets when I hear complaints from a lot of AI company I was working with, and the complaints, "Hey, this great SaaS model I have, however, my spot prices compute all over the place. $4, $6 tomorrow. There's no way we can control that." In my mind, I was like, "That doesn't make sense. This doesn't sound like sustainable business model for anyone." If American Airline can hatch their oil prices, how can you run business with that? That's very intuitive to me. Hey, the market is so underdeveloped and someone needs to step in to create financial products and as well as data products to serve that particular cohort.

>> I mean, the changing in prices, one, it's a reality in every business, and two, the hedging is interesting. Now you can build the hedge to maybe potentially offset costs.

Carmen Li

>> Again, risk transfer, right? You don't want to risk, someone else may want to speculate on that. That's great that I'm taking the risks. For business people, you just want a stable cost. You know exactly it's going to cost $3 per hour, fantastic. You don't want to go up and down, but you don't want to enter a, say, five-year contract.

>> Yeah.

Carmen Li

>> If you get out, just like every single other features market.

>> It could be a nice revenue stream, too, to offset those costs.

Carmen Li

>> Predict the future, predict your revenue expectation as well.

>> You've got data.

Carmen Li

>> Yes.

>> Again, if you're a big bank and you have huge... Say you're Goldman and you have all these cloud, and Capital One is huge in the cloud, they want to start thinking about training their data. I'm sure they have projects, there's no TCO calculator, right? I mean, remember the old TCO calculator, you type in all those...

Carmen Li

>> .

>> Spits out a number.

Carmen Li

>> This is how much. Yeah, yeah.

>> You're going down that road to do some asset planning with commodity compute, basically.

Carmen Li

>> I like to give them transparency, what's happening in the past, what's happening now, and then we can all talk about what's going to happen in the future. Again, defense supply, right? All I want to do is give you transparency that show them defense supply card.

>> Okay, so you're newly-formed. What's it been like so far? Give us a feeling for who you're talking to, customers you have.

Carmen Li

>> Yes.

>> What's been the feedback?

Carmen Li

>> Yes.

>> Any changes in product strategy?

Carmen Li

>> Literally, my friends, my people I know, they chase after me, because the client is like, "I thought you were going to probably be out in next month," so it was like, "Doesn't take anytime to accumulate. You have to verify and test the model is built and test it." But yes, and chased by my clients, basically, which is an awesome feeling the same time. You feel the urgency, the responsibility to create product actually worked, and track the market and produce a clean great data. No data can be without error, but I want to create as good as it can be.

>> Are you thinking syndication of the platform or are you thinking custom products for specific companies, or both?

Carmen Li

>> Really creating a standardized product. For me, it's about, what kind of product people need to really help them manage risks? Right now, we're focusing on the top thing, top GPU chips people are using most frequently. Obviously, we are not with third party, so we don't really take sides with which chips they want to use.

>> Yeah, demand's a big point of it.

Carmen Li

>> Exactly, what kind of chip company they want to go with. First is, what are people using? If everybody's using one type of chip, that's what we're going to do. If people are using 20, that's what we're going to do.

>> Okay, so the question is, risk management, dealing with the ever-changing GPU landscape or XPU landscape. Broadcom, NVIDIA, they're all building these new chip designs. Memory, HBA's hot, I mean, it's hard to get, but other memories, highly available.

Carmen Li

>> Right.

>> When people have to figure all this out, they have to factor in things like allocation.

Carmen Li

>> That's right.

>> Are the chips available? What's that going to impact on the risk? Take me through that mindset, because those are big determinations into deployment, which has an operational risk.

Carmen Li

>> Yes, I agree. If you talk about new chips coming out, the Blackwell, right? Obviously, there's no data currently, so unfortunately, I don't have any insights.

>> Yeah.

Carmen Li

>> What I will have is a lot of data on the A100, H100, the H200, our L4s, the 4 Niles, all the good stuff. Then that's like the AMD out, but a lot of different chips.

>> Yeah.

Carmen Li

>> What usually people come to me is data center people. They were like, "Hey, I'm going to invest X amount in this cluster, this data center, this location. How am I going to price this? Am I pricing near the Hyperskiller or some secondary cloud provider? What's the visibility?" And then no H100 class is the same, essentially. Two points, different CPU, different memory bandwidth. What we do is we bring all their specifications and then we calibrate to the market and, depends on where they are, we'll give them, "This is the range that it's market price regulation." Not cost-based, but market-based regulation. Again, it's their decision to see how they want to price this. Yeah.

>> The thing that's interesting I find is that you've seen a lot of... What DGX did with Nvidia, you see CoreWeave, we mentioned that before we came on camera, you got these GPU clouds, you got these new services I could buy and sit on them, but there's no power. Okay, oops. Flaw in the design. Cooling are the constraints. You have constraint issues and then, do I buy the product or do I pay by the drink?

Carmen Li

>> Right. Rent it, define where should you rent it from, right?

>> That sounds like you're hitting that, too. If I'm a customer, I can look at that and say, "Okay, I should be paying CoreWeave-"

Carmen Li

>> X amount....

>> "X amount."

Carmen Li

>> Right.

>> They're formulating their pricing strategies or bid?

Carmen Li

>> Again, I'm not advising on price. I think everything depends apply-

>> On the market.

Carmen Li

>> If someone is so great, have the best-notch software and support and everything else, I mean, you can argue they probably charge high at a premium. If someone, "Hey, there's a cluster in my basement from Carmen's home," you probably won't pay much, because it's not very, very reliable.

>> You all make the market price.

Carmen Li

>> Exactly, right? You have to time the risks, how much you can really tolerate, and depends on workflow. Some workflow cannot be interrupted, some need the huge clusters. You might not want to risk that. If you bring some quick inferencing, you put down capital ads, fine, she's charging me 40 cents, who cares?

>> I can see my mind an infographic on siliconangle.com from Silicon Data, "Powered by Silicon Data, current prices."

Carmen Li

>> We'd love to work with you on that. Yeah, we'd love to give exposure to the participants.

>> Well, we hope that can happen soon. If you look at the future, okay, your vision, what's your north star vision? In your mind, you're at a steady state, you get the product out there, you got some new product additions. What does the future look like at a steady state for you?

Carmen Li

>> I don't know if I've ever been in a steady state. Because things move so fast, everyone is completely FOMO-anxious. I don't know if maybe it's by all my friends. Things are moving so quickly with chips and all the large language model update every six months and chips come in every year now, right? It's getting crazy. The way I want to see it is the market grow at a pace where the financing can come in effectively. There's a lot of risk management too, and people have the transparency, have the trust. If you buy anything, let's say from a third party, you have to trust to say, "Hey, it's not like I hope it works," it's more like, "Hey, it's verified by Silicon Data. Hopefully, it works." A third party oversight.

>> You're at the nexus of financing and high-performance compute.

Carmen Li

>> I would like to be the data-

>> From a marketplace perspective.

Carmen Li

>> Exactly.

>> A data provider for that intersection.

Carmen Li

>> Correct.

>> All right. What's next? You got some funding, you got products.

Carmen Li

>> I got really lucky with-

>> Give a plug for the company, what you're looking for. Obviously-

Carmen Li

>> I'm hiring. I'm looking for amazing engineers and product-business people. Obviously, we are based in New York. However, we are pretty much open, be flexible, working from home policies. We love to hire amazing people from every disciplines, if you're an engineer or you're a data scientist person, we'd love to hire you. Obviously, the vision has really become the information, data product center, so we can really help the compute market to grow.

>> Carmen, I have to ask the entrepreneurial journey question, because it's hard to do startups. What's the journey been like for you personally, and share some experiences?

Carmen Li

>> It's been amazing. I would literally tell my friends, "I love every single day, waking up so excited to do work." Really, because it's a problem for my team sometimes, like, "Carmen never sleeps and never stops." The only time I don't do work is when I'm with my kids. I have two daughters, seven, three, so it's amazing to be a mother, at the same time, be an entrepreneur. I'm so happy every day. I do things absolutely I love, which is doing the work. I'm really passionate about this industry, the things I think I can help the industry to move quickly, effectively, efficiently. That's the most important thing.

>> Yeah, the industry needs you. Well, we're certainly looking forward to collaborating. Thanks for coming on here at theCUBE.

Carmen Li

>> Thanks so much for inviting me, this is good.

>> we're at the center of all the finances. Our studio in Palo Alto, connecting Wall Street with Silicon Valley and creating a great community and, thanks for being part of theCUBE and the NYSE Wire being that Brian's putting together.

Carmen Li

>> Yeah.

>> Thanks for coming on.

Carmen Li

>> Thanks so much.

>> All right, cool. Thank you. Okay, we are here, theCUBE's East Coast location. Again, linking Silicon Valley and Wall Street together, merging the communities and building the network, it's an open source network, and looking forward to more content. I'm John Furrier, the host. Thanks for watching.