David Blumberg, Founding Partner Blumberg Capital and Jay Eum, Founding Partner GFT Ventures: AI & DeepSeek

David J. Blumberg, Founder/Managing Partner Blumberg Capital

David is the founder and managing partner of Blumberg Capital, an early and growth stage venture capital firm that partners with visionary entrepreneurs from Seed stage through their growth journey. The firm was among the first investors in industry leaders including Addepar, Braze, DoubleVerify, Finaro, Fundbox, HootSuite, Nutanix, Trulioo and Yotpo. David is based in Miami with team members across Miami, New York, San Francisco and Tel Aviv. Prior to founding Blumberg Capital in 1991, David managed international investments with the Bronfman Family Office, Adler & Co, APAX Partners and at T. Rowe Price. He also launched business development for Check Point Software Technologies. David earned an A.B. in Government, cum laude, from Harvard College and a MBA from the Stanford Graduate School of Business and INSEAD. He speaks French and is proficient in Hebrew and Spanish. David serves as a member or observer on the boards of DataHeroes, EasyKnock, Jassby, Kloudfuse, Lendio, SupplyPike, Trulioo and Wunder Mobility.

Jay Eum is the Founding Managing Partner of GFT Ventures

Jay Eum is the Founding Managing Partner of GFT Ventures, Global Frontier Technology Ventures. GFT Ventures invests in category leading early-stage startups in the US and Israel focused on AI, data science, and other frontier technologies. Prior to GFT Ventures, Jay co-founded TransLink Capital, a Palo Alto based early stage venture capital firm supporting entrepreneurs developing customer connections and partnerships in Asia. Previously, Jay was Managing Director of Samsung Ventures where he established and led the US venture capital operations for Samsung, responsible for U.S. based investments and portfolio management. Jay started his venture capital career at Vertex Management, the Singapore based venture capital arm of investment firm Temasek Holdings. Jay received an MBA from Stanford University and M.S/B.S. in Chemistry from Seoul National University. Jay serves as Chairman and Board Director of the Asia America MultiTechnology Association (AAMA) and on the National Advisory Council of the Association of Asian American Investment Managers (AAAIM)

Investments

David: And you’re asking what are we investing in? Sure. Well, Blumberg Capital is an international venture capital firm. We’re practicing what I call Classic Venture Capital, mostly early stage. We do have a growth fund that starts sort of late A and above. But our classic early stage starts pre-seed seed in A, managing just under a billion dollars across five different funds. Most of it is B2B software, business-to-business software. Heavily for the last, say, 10 years, it’s been AI mining some sort of data proprietary data and addressing a vertical need in a vertical domain. So we could go into law, accounting, mining (the hard metal mining type), cancer detection, logistics, infrastructure for the data center. All of these areas can be addressed by intelligent analysis of proprietary data sets that then help the businesses in that domain do a better job of something faster, better, cheaper, or very excitingly new, something they couldn’t do before. We do a lot of FinTech. and we just sold a wonderful company called FinNarrow. The ultimate price turned out to be because of stock and shares $1.1 billion dollars sold to a public company called Shift4. The CEO of Shift4 just got appointed to become the head of NASA. So maybe we’ll start investing in some things toward the space area as well.

Again, we have a preference for software. We do diverge sometimes into hardware. Most of it’s B2B. We do once in a while something in the B2C. And our international consists of North America, so US, and our wonderful neighbors to the north, Canada. We invest a lot in Israel. We have an office in Tel Aviv. They are very active. And we’ve done a number of deals in Europe. Now that I’m in Miami, we’re open to deals from South America, Central America, et cetera. But they should generally be those addressing the US market as the major point of entry. If it’s a Brazilian deal for the rest of South America, we probably would not be the best added value investor. But if you’re coming from somewhere else into the US and then from the rest of the world, we think we’ve done a pretty good job of learning that and helping people expand with our innovation council, our business network platform, and so on.

Jay: Since GFT Ventures is a relatively new firm, allow me to start there. So GFT Ventures, Global Frontier Technology Ventures, we launched in 2021 during the pandemic. The co-founding partners are myself and my partner, Jeff Herbst, have been on the corporate venture side for two decades apiece. Myself, I launched and ran Samsung Ventures back in 2003. And my partner, Jeff, was leading BizDev, CorpDev, M&A, and investments at NVIDIA since 2001. We’ve been friends for 20 years, workout buddies for over a decade, co-investors in multiple AI and data science companies. And during the pandemic, we decided that we wanted to launch a frontier tech focus fund with a strong concentration on AI and data science opportunities, and thus we launched GFT Ventures. We focus on (very similar to David, who we’ve known for close to two decades as well) what we call Classical Venture Capital. I actually like that word as I typically say Old School, but I like the word Classical much better. We try to build partnerships with the entrepreneurs that we work with. We try to be the lead or co-lead, sit on the company’s boards and typically that first price round opportunity is going to be a seed or series A. We typically will initially invest anywhere from 2 to 5 million dollars, 5 to 10 years of the lifetime of those firms. I would say over 70 to 80% are AI focused. And the remaining balance are applications of AI that include Robotics and Digital Health. um We’ve been actively investing now for the last three years. We have 15 companies in the portfolio so far, and we’re actively looking to invest to build out the portfolio to about 25 companies in our first fund.

DeepSeek/AI

David: Yes, I’ll start and I am thrilled! I hope that shocks people because everybody’s afraid and scared. They think it’s dangerous and oh my God, it’s going to cause US prices of big companies like Nvidia and Sam Altman’s Open AI and so on, to to go down. Obviously, one’s public, one’s private, but I’m a Disrupter. I’m a Venture Capitalist. That’s what we do! So is Jay. so is Mr. Herbst, who I know also (these other two guys are fantastic, and I think they’ll share the same opinion.) The length of companies stay like staying on the Fortune 500 list in 1900 was, I think, about 90 years. Now the last time I looked at it, it’s down to about 15 years. So the pace of disruption is accelerating. And I think that cheaper is better for fundamental technologies! So the cloud used to be very expensive, when it was sort of server client type, then it moved to cloud and it got cheaper. There used to be only mainframes and mini computers and PCs. And now in our hand, we can hold the same power as a mainframe or a supercomputer. So cheaper is better and broader markets allow more people to partake in the glory of these machines we use to improve our lives.

And so I want a lot of competition. Of course, China is a unique character because the government is so intervening a lot intheir economy and can take probably data, if they want it. So I’m concerned about that issue, and probably like everybody? Maybe the data was stolen. We don’t know? Maybe there’ll be security concerns about it, but in general, I like the fact that they used, in theory, (and we don’t know if it’s true?) But they claim, they used cheaper, faster, less powerful chips, and they used a smarter algorithm. Well, good for them, and I hope a lot of other people do it. And by the way, they’re not the only ones. There are other companies. Baidu has something called Erniebot, and Alibaba has Quen, and Moonshot has Kimi, and so on. And there are a bunch of others, there are probably others around the world? I want to let a lot of companies compete because the world will be better if AI was cheaper.

Jay: Yeah, you know I have to agree with pretty much everything that David has said. My partner Jeff Herbst and I had lunch a couple of days ago, right after the announcement of DeepSeek. And while the whole market was still kind of in in trepidation and NVIDIA’s market cap had fallen off a cliff! (Despite the amazing run that they’ve had over the last three years?) But fundamentally, there’s a couple of things: So I think we’re all familiar with Jevon’s paradox and this has been used time and time again in the technology scene where the lower the cost, the more usage occurs! And so what that means is that if the Cost of Compute for AI actually becomes more efficient and it’s lower, then that means that there’s going to be actually more usage! So, for example, because of the cost of training and the cost of inference has been so high, there has been a limit in terms of where AI can be applied! It only makes sense from an ROI perspective where it meets the cost where it justifies the cost in terms of value creation. But now if that cost is lower, what that means is that there’s a far broader set of potential applications where you can utilize AI, which would not have been able to be used in the past. So this actually creates a massive opportunity for additional applications to be launched. In fact, it creates opportunities, especially for startups, which are very relevant for David’s firm and our firm, to be able to do really interesting things with far less capital than previously thought.


In fact, quite frankly, this whole LLM space, as we all know, had been a domain of hyperscalers and massive multi-billion dollar funds that can afford to continue to fund these opportunities. But if you can actually do it far more efficiently with a fraction of the cost, then there’s more opportunities for startups. So that’s the first point. I think the second point is there is a fair amount of undisclosed information, which David referred to as well, because the cost of their ah training is most likely only the final cost of their training. It does not disclose or include any of the prior training costs that led up to this. And as we’ve all been reading, there is a lot of conversation around how much did the DeepSeek team leverage other LLMs, especially OpenAI’s LLM. So if you actually think about the accumulated cost, it may not be as cheap as everybody thinks. Now, having said that, there are certain techniques and algorithms that the company has used very smartly, which, as we all know in the world of tech, will be adopted and copied and utilized across the board. So all in all, I think this is a tremendous opportunity! And for the folks that are worrying about open AI or Nvidia, I think it’s a wake up call for everybody that has been just charging ahead, not worrying about efficiency and just pouring capital at the problem to solve the AI issue. I think it’s a good thing that people are starting to pay attention for efficiency, especially because of the amount of energy that goes in and the environmental effect that will impact massive usage of compute. So all in all, I think this is a positive development.

AI and Energy

David: And I like to make a point about energy and how fundamental it is to our modern civilization. And I’ll ask your audience a question: How many of us would change places with our great, great, great grandparents? Not many. Maybe if you were the descendant of a Duke, an Earl, or a King, but most of us would say our lives would have been much more difficult. So the big change, frankly, in our ancestors, three, four generations ago and us, is energy. We had a revolution because of coal first in England and the Netherlands, and then it became oil, natural gas, nuclear and so on. And I want to just raise the flag to say that all of us in the tech world owe a huge debt of gratitude to those energy producers. And if the AI revolution recently has taught us anything, it is we’re going to need a lot more energy to power our rich world. But there’s another thing I want to bring up, and that’s the poor world. There are approximately 4 billion people on Earth today that have less than 4 hours of electricity available to them, and 700 million people, (that’s a lot of people – twice the US population) plus have zero electricity. So I consider myself, what’s called an “energy humanitarian” and an “energy national security hawk“, and we need energy to be both moral and safe. And we need it for the tech world to be productive. So we need to be building a lot more power plants. Fossil fuels are going to be a big part of the mix. They’re 80 plus percent of the world’s energy now that will probably continue on in the future. I’m not in favor of the green agenda. I’m in favor of the energy pro-human agenda. And I just want us to understand that because we have expensive energy, we don’t have much AI. If we have cheap energy, we will have lots of AI. And lots of AI will solve Cancer. It will solve differential equations that lead us to Quantum, and so on and so forth. Many, many, many great things will happen with cheap energy.

Jay: Well, again, in full agreement, as I said, all of us in terms of just being mindful of the resources required, just to be able to do the same things efficiently with less energy consumption, regardless of the source. There’s a lot of conversation around nuclear energy and other alternative sources. But at the end of the day, just to be able to be smart about things and be more efficient about things has clearly been the way that the technology is developed over time. And so this new development, again, that DeepSeek is launching is no surprise. It just goes to show that innovation comes from when you have resource constraints and barriers and because of the very variety of different constraints that were set to DeepSeek in their Chinese environment, they’ve been able to do amazing things which can be again adopted by everybody else. Its just a net positive.