Masterclass: Inside the Deal—Why We Said Yes to Groq

Frequently Asked Questions

FAQ

• Where is the transcript for this webinar?

  Speaker 1:

  Hello, this is Mark Edwards, Chief Investment Officer at Alumni Ventures. I’d like to welcome you and thank you for joining our webinar Inside the Deal: Why We Said Yes to Groq.

  Before we jump in, just want to do some legal housekeeping. We are speaking today about Alumni Ventures and our views of the associated venture capital investing landscape. This presentation is for information purposes only and is not an offer to buy or sell securities, which are made only pursuant to formal offering documents for the fund.

  So next slide.

  By way of introduction, as I said, my name is Mark Edwards. I serve as the Chief Investment Officer of Alumni Ventures and will be leading today’s webinar discussion. In my role as CIO, I manage and oversee all of the investing activities at Alumni Ventures, which involves a team of 40 full-time investment professionals who are located in the venture hubs of Boston, New York City where I am today, Chicago, and Menlo Park.

  This activity involves the gamut of all things womb to tomb for our investing activities — identifying companies that we find are interesting, sourcing and getting relationships with those companies to create the opportunity to invest, all of our due diligence and deal review and investment approval processes, the monitoring of our existing investment portfolio, and then making decisions on follow-on capital deployment decisions and all exit activities. So a lot on our plates with our existing and growing number of portfolio companies.

  I’ve been serving in this role at Alumni Ventures now in my third year, but my involvement both with the company and private markets dates back a good bit further in history. My prior career involved more than two decades working in Manhattan as a private equity professional for a few multi-billion dollar investment partnerships. And here my focus was similar to the investment strategies here at Alumni Ventures, but tended to focus more on larger, more mature companies — cash-flowing businesses — where generally my investment firm would be looking to take a controlling investment in those companies with a plan to improve operations and increase value, and then generally amplify the returns associated with that by putting leverage on the company’s balance sheet. So closely related to the role at Alumni Ventures, but with some important differences — private markets nonetheless.

  And during that time period, I got fairly active outside of my private equity career investing in younger, more startup-oriented businesses, in part as a personal interest and in part I felt like it gave me a window into sources of innovation and disruption that were very relevant to being a value-added board member of a lot of my portfolio companies.

  Through that activity, I had a serendipitous introduction to our founder, Mike Collins, back in the early days when Alumni Ventures was really just getting started. And that led me to not only invest my own money in our first-of-type investment fund, but get actively involved with helping to establish and develop — and sort of stitch into Alumni Ventures’ processes — a lot of features of our investment process that are still important parts of our business today, which includes the deal review process, the deal scoring process, the establishment of outside investment committees, and things of that nature.

  So I didn’t anticipate at that point in time — and at a lot of steps along the way — that the company would grow into the business that it is today that’s managing close to $1.5 billion of outside capital. Nor did I expect that it would morph into an opportunity for the 2.0 version of my career, but it’s been a great journey for me as well as a lot of the other teammates that have been involved and our investors along the way. And it’s a privilege to be here today talking about Groq and profiling this important portfolio company.

  So next slide.

  Quickly, if you’re not familiar with Alumni Ventures or you have some limited information from our outreach and our presence on things like LinkedIn — the company, as I mentioned, has been investing outside capital now for more than a decade. It really came together in 2014, focused on two foundational concepts.

  Rather than wanting to be sort of a me-too participant in the venture business, we really built a business around two strong points of view that we thought were important enough to build the business around. The first was really focusing our attention on creating access to venture capital as an asset class with the same level of professionalism as what a lot of established institutional investors benefit from, but focusing our time and attention on bringing that same experience and access to individual investors.

  And then the second concept is really trying to harness the power of strong networks of people — in this case, mostly focusing around alumni of a handful of selective schools around the country — with the idea that in a business where access is a real important point of differentiation, investing together we would do better than investing as individual practitioners.

  And those two concepts are fairly simple in many ways, but have really been core pillars that we’ve built the business around and have had the benefit of really translating into important points of differentiation for our business in the marketplace.

  And as a result, like I said, we’ve been able to grow pretty steadily through what’s turned out to be a pretty interesting and volatile decade in venture as an asset class to now manage a portfolio representing approaching $1.5 billion of outside capital that we’ve raised from more than 10,000 individual accredited investors and have pulled together a portfolio that represents close to 1,500 individual companies and growing.

  So we’ve gone from a relative startup just a decade ago to now being the largest firm focusing on the needs of individual investors and, in terms of the number of companies that we’ve backed and continue to back, one of the most active venture firms in the U.S. and globally.

  Along the way, we’ve definitely focused on building the business much like the companies that we invest in. And instead of just being a group of investors sitting around the table talking about interesting companies, we’ve really invested pretty aggressively to build up the company to now 120 employees with a strong focus on areas like legal and compliance and fund management, fund accounting, technology—to really build an enduring business around the investing activities that I manage.

  Next slide.

  Along the way — it’s been 10 years — but we feel like we’ve accomplished a lot. There’s been a lot of recognition that we’ve received, both specific to the venture industry and then kind of in the ecosystem of small, rapidly growing private companies as well.

  I mentioned the attribution for the level of activity that we represent in the marketplace — highlighted here by PitchBook, which is sort of the definitive information source in alternative markets and ventures specifically—that we’ve steadily been one of the most active investors judged by deal counts for each of the last several years, and expect that to continue.

  But in addition, there’s a lot of other attributes of our business that have been recognized, and we feel very fortunate to get the recognition. One specifically here is the recognition by CB Insights — another market information resource — identifying us as a top 20 venture capital firm in 2024. And this is really a function of the companies that exist within our portfolio, and puts us in the same category as a lot of firms that really have been sort of pioneers from the early days of the industry and kind of the granddaddies of Sand Hill Road and elsewhere.

  So we feel really good about not only the activities that we’ve been recognized for, but the quality of those activities, which ultimately is the determinant of our investment performance.

  Next slide.

  Actually, before I move on — I do want to call out — I apologize. We’ve put together a list of 50 of our existing portfolio companies, which makes it a little bit more manageable to review and understand just given the breadth of our overall portfolio, which is several multiples of this size. I would encourage people to access and read the report here. It’s a lot of private companies, as you’d imagine, that unless you’re a real student of the venture industry, many of them may not be familiar to you — but some of them will be. And I think it’ll give you a flavor of how we go about our decision-making process in terms of which companies we invest in. And it’s a document that we’re very proud of.

  Groq, which we’ll talk about in the body of this webinar, is one of the companies on our Apex 50.

  Next slide.

  One final point before we move into the body of the webinar — just to clarify and explain the sort of business rationale behind our approach in the marketplace. We focus 100% on being co-investors in the market, as opposed to aspiring to lead rounds. And if you understand the way that the venture industry works — generally when a company makes a decision to raise a new round of financing that focuses on a specific class of securities, generally preferred stock — the first step is to identify a lead investor that’s going to propose the terms on which they’re prepared to invest and sort of negotiate the deal that works for the company in terms of the economics of it and the governance of it.

  But generally, it involves that firm also agreeing to join the board of directors and kind of serve in a more hands-on role to try and aid and assist the company in its developments, but also be a representative looking after the specific interests of the investors.

  That’s a big responsibility and one that generally is the biggest determinant of how active an individual partner can be. It gets very difficult, as more and more board seats pile up, to dedicate the appropriate amount of time and thought into serving that role. And given the importance in our investment strategy of company diversification and the number of investments that we need to make in service of our investors—because overwhelmingly people that entrust us with their money are using us as their only access point to venture as an asset class — we made the strategic decision that it’s better for us not to try and take on those responsibilities as a lead investor, but rather focus on being a large and important co-investor.

  So we’ll go to a company that’s raising a round and encourage them to come to an agreement with the lead investor on the terms that they’re prepared to move forward with — make a determination that works for them in terms of the alignment and board chemistry.

   

  And then very often we have worked and invested alongside a lot of those investors over the course of our history and can do the work to get ourselves comfortable that we share the perspectives on where the opportunities and risks lie and really rely on that firm to serve the governance role sitting on the board. Whereas we will invest in a significant enough amount in the round so that we are relevant to the company and get the same access in terms of completing our due diligence but really look to complement that role as opposed to compete directly with it. And that’s created a dynamic where not only are we able to be very, very specific in building our portfolios, trying to align around a specific number of firms not only based on their overall reputation and resources, but sort of making a connection between one firm and one partner specifically with the profile of the business that they’re investing in.

  But then be very deliberate on trying to build our portfolios around best-of-breed marriages between entrepreneurs and lead investment partners. And if you look at venture as an asset class, it’s well worth the effort to try and access those deals relative to a lot of other investment strategies that you may have in your portfolio and you participate in involving public companies. The distribution of returns in private markets tends to show a lot more separation between the top firms and the median level of performance in the industry. And that’s generally because, unlike public markets where the gravitational pull of the overall market is incredibly strong and it’s very difficult sustainably for a lot of individual managers to separate from that, the way that the venture industry tends to work is that there’s sort of an alignment between the most promising founders with the best ideas gravitating towards the most viewed, prestigious, and accomplished firms.

  And so the difference in returns historically—and it tends to sort of sustain—between the top quartile performers and everybody else is pretty meaningful. And in some periods of time, it could be several hundred basis points, which adds up to a lot of money over time. So our strategy is very much focused on not trying to compete against or displace those firms, but to get access to the same portfolio companies on the same economic terms as they’re investing in and sort of let the performance of the companies speak for themselves. Next slide.

  Okay, so I’m going to turn my attention now to Groq and try and give a profile of the history of this company, how the investment came about for us, and some lessons learned—even though we’re still kind of in the relative early days of our experience with this company, but are very excited about what we’ve seen so far and think the future is incredibly bright.

  Groq is a little bit of a strange name. I understand it sort of loosely translates to understand profoundly and intuitively. It’s an AI infrastructure company located in Mountain View, California—a business that we first backed in 2021 and then invested again in the summer of 2024. And a lot of our portfolio companies, if you spend time with us, you’ll realize—myself, our founder, a lot of our investment team members—really like to talk about individual companies that we’ve invested in.

  We’re fortunate to be in a very exciting arena and work directly with companies that have very ambitious business plans and a lot of very interesting technology underlying them. And Groq is no different, but the reason that we chose to profile this company specifically in this webinar is for a slightly different reason. One of the opportunities and one of the exciting elements of our business—and venture capital generally—is the ability to serve as a futurist. Given the time horizon over which we invest, there’s both an opportunity and, if done successfully, a real potential reward from—rather than being reactive and just making investment decisions based on the full picture of how markets exist at the time of our investment—to really try and look out into the future.

  And by studying current market structures and the sources of innovation and potential vectors for disruption, try and form a point of view on what markets look like several years down the road. And by looking around corners, develop a perspective on where we think we should be investing money today.

  So if the version of that future turns out to be accurate several years down the road, then we’re already very well positioned and have been able to capture value before it’s obvious and actionable to a lot of other investors that have maybe a shorter time horizon or just spend less time doing the forward-looking analysis that we’re doing. And we think Groq is a good example of the power of that in our business model, and I’ll take you through that in the body of this webinar specifically.

  To kind of give you a sense—we tend to think about and talk a lot at AV in sports analogies. And in the case of this messaging around Groq, I’ve been thinking about it a lot lately. I’m a fan—and hopefully some people listening in are the same—of hockey and the NHL specifically. And if you are, you probably saw that recently, about a week and a half ago, Alex Ovechkin scored his 895th goal as an NHL player to pass Wayne Gretzky as the all-time goal leader in the history of the league.

  And there was a lot of pomp and circumstance around that 895th goal out in Long Island at the New York Islanders facility. And among others, Wayne Gretzky was on site to interrupt the game and graciously pass the baton to Ovechkin as the all-time goals leader.

  In the lead-up to that—and also a little bit in the interviews after the fact—Gretzky got a little bit sentimental about his own career and reiterated a concept that has sort of stuck with him as a player now for several decades, but I think also kind of resonates with us internally, and I wanted to share it. Which is: he said his power as a hockey player was not to skate to the puck, but skate to where the puck was headed.

  And the notion was that he had this supernatural ability while he was on the ice to study the flow of the game and instead of being reactive to things that were already taking place, anticipate the flow of the play to position himself to be in a position before anybody else realized what was going on—to either score or assist or make a play that was ahead of everybody else.

  And it certainly served him well in his career. I don’t claim that we as an investment organization—either me personally or members of our team—are going to claim Gretzky-like supernatural ability, but this notion of “skating to where the puck is headed” is an important sort of premise of our investing strategy and one that we invest a lot of specific time trying to execute and refine over time. And hopefully, as we get into the specifics, you’ll sort of see the benefits of that if you can execute it successfully.

  So back to Groq. As I mentioned, Groq is an AI infrastructure company located in Mountain View, California. The company was founded back in 2016 by an individual by the name of Jonathan Ross. Jonathan, before founding the company, had spent a lot of time as an engineer at Google. And while he was there, one of his responsibilities among several was working on a project associated with what was called a Tensor Processing Unit. And this was sort of an early iteration of a very specifically designed circuit focused on handling Google’s internal machine learning and AI-related workloads.

  So the technical problem that is at the core of Groq’s solution and current business strategy is something that Jonathan has been working on for quite a long period of time—and in the context of Google, working with the epicenter of some of the strongest engineering talent that exists in the Valley. So very, very backable founder.

  Simply put—and we’ll get into some of the specifics—Groq has developed a very intentionally designed chip called a Language Processing Unit, which is exclusively focused on trying to process inference workloads for large language models. And we’ll get into specifics of what that actually means in plain language.

  But as you can see here, the bottom line is: by taking a fresh look at the anticipated hardware needs of an emerging segment of the market—associated directly with the proliferation of AI—they took a clean look, and rather than innovating with marginal, incremental improvements over existing technologies, really started from scratch with something purpose-built to serve the AI inference workloads of this emerging technology.

  And as a result, they have a solution—and these numbers, it’s a fast-moving industry, a fast-moving company, so you may see some slightly different performance statistics out there publicly—but nevertheless, they have a solution that is directionally nine times faster than some of the competing hardware solutions, and is able to accomplish that processing speed at a third of the power consumption and a fifth of the cost of some of the existing technologies. So pretty transformational from a performance standpoint. 

  Here is your cleaned-up and corrected transcript, with all time codes removed and spelling, grammar, and word-splitting issues fixed. The language and structure remain exactly as originally presented:

  I thought it made sense just to take a step back and kind of clarify some vocabulary so people can sort of understand and dimensionalize where Groq fits into a pretty broad and important ecosystem. So in the operations of large language models, there are sort of two important but distinct workloads that need to be addressed with the computational infrastructure associated with powering the operations of large language models.

  One is training and the other is inference. And as the name implies, training involves the upfront and ongoing process of having a model ingest huge quantities of data in the form of text and images and audio, things like that. And it’s this training process that arms the model with the functionality to understand language, generate content, perform tasks—all the things that it’s set up to accomplish.

  By contrast, inference involves the actual process by which a user orchestrates a specific prompt and asks a question or makes a request of the model, and then it uses that information to go into itself and respond with an answer, an image, taking actions, whatnot.

  So in simple terms, I like to think about it as studying for the test versus actually taking the test. And as is frequently the case with hardware, a one-size-fits-all approach to both of these processes is not optimal. Just like you wouldn’t score your best round on the course if you only had a six iron, you’re not going to optimize the operations of an AI strategy with a single monolithic approach to infrastructure handling the disparate needs of both the training processes and the inference processes.

  So Groq has carved out a lane for themselves where they are the low-cost provider of fast inference, and we happen to think that it’s a very exciting lane and one that we’re very excited to see develop as we monitor this investment.

  Next slide.

  So a little bit of background—and this goes back to the Gretzky comments. The company, as I mentioned, was founded in 2016, started raising outside money a few years later as they really started to develop traction, but still very much a company in product development mode and relative stealth in the scheme of the Silicon Valley ecosystem.

  Our team, really led by a group of investment professionals associated with our MIT-related fund called Castor Ventures, had been doing a lot of desk work and discovery trying to anticipate interesting areas for what they saw as an emerging and irreversible trend towards the proliferation of machine learning and AI dating back to the 2019 timeframe. And Groq was one of the companies surfaced through that.

  We weren’t able to invest, even with those insights, until 2021, but spent the period of time once we had developed conviction on the outlook for the industry that we were trying to invest behind to make contact with the company, build a rapport and a good relationship, and be ready to move aggressively when they did raise money in 2021. That was a Series C round that valued the company at just over a billion dollars and was led by Tiger Global and D1.

  So between identifying the company and making our first toehold investment—back to the need for patience in our investing business—it probably was more than two years, but very excited to be involved in that round. And then we were able to invest again largely because we were already a participant when the company raised more money in 2024.

  The date that’s not on this slide, which I think is important for everybody to understand and internalize, is November of 2022. Because what changed between the Series C and the Series D round really was the release date of ChatGPT. And this kind of served as the coming-out party for large language models and AI more generally, by extension, in the minds of a lot of consumers—and even enterprises the same—and triggered the perennial Cambrian explosion of interest and capital being attracted to AI as a theme.

  To put things in perspective, at the time of ChatGPT’s version one announcement, Nvidia—which obviously has been, from a public company standpoint, one of the huge beneficiaries of the advancement and proliferation of AI, specifically related to infrastructure—at that point was a public company that was one of the most well-covered and well-recognized businesses in public markets. And yet despite being covered by an army of analysts that literally spent sleepless nights modeling out the company’s share count and tax rate to the right of the decimal point for every quarter, overwhelmingly that community of analysts missed the coming wave of AI-related demand specific to their GPU product line.

  And in the period of time from November of 2022 to where we are right now, Nvidia has gone from a $16 stock—giving effect to the stock split that took place earlier this year—and commanded about a $400 billion market cap, relative to where it stands now in the mid-90s in terms of share price and a $2.4 trillion market cap.

  And that’s obviously down from where it peaked out. So kind of hiding in plain sight was the opportunity to see $2 trillion-ish (a little bit less) of incremental market cap created by a well-understood public company. And it just goes to show the magnitude of this technology trend that we’re living through right now and the value in being a little bit ahead of the crowd.

  In terms of where we are right now with Groq, the 2024 investment gave them a significant amount of available cash to transition from a business that had been working to establish product-market fit and refining their go-to-market strategy, to really putting their foot on the accelerator and getting their LPU out in the marketplace. So 2025 and beyond is really a time period where the company is moving from development activities to widespread scaling and commercial penetration. And based on what we’re seeing, we’re super excited—and in fact, have rarely seen a company get the traction over the period of time and the dollars involved as what we’re seeing with Groq.

  Next slide.

  So why inference, and why infrastructure generally?

  I talked about the thought process that our team had been going through in 2018 and 2019, trying to come up with some thematic approaches to investing in the anticipated growth and opportunities related to AI. And this is one of several strategies that we’ve adopted and have invested behind. But looking back on some previous technology platform shifts that gave rise to winners like Cisco, one important element for us to evaluate was the so-called “picks and shovels” approach to investing in the AI boom.

  And this is language kind of borrowed from the Gold Rush, which is still front and center in the lore of this part of the country in California. But the notion is, rather than trying to invest specifically in the prospectors going out and trying to identify their own claim of gold, take the approach of Levi Strauss and some other famous and enduring companies and sell the supplies that arm those prospectors to go and pursue their dreams.

  And while hardware and picks and shovels can be a challenging business—there are a lot of very significant technical challenges—an industry as competitive as semiconductors requires sustainable points of differentiation. And similarly, it’s a business that tends to involve some “not invented here” types of obstacles and very long and arduous sales cycles. But for the companies that are able to navigate through that and grow into companies with large installed bases—significant moats from competitive disruption because of the rip-and-replace considerations and costs with their customers—and then generally businesses that have significant IP protection and things that we like to see, they can be very powerful and enduring franchises.

  And that was a big part of our draw to having a component of our investment activity associated with AI be focused on the infrastructure arena. And certainly, Groq is one example of that.

  The other thing that we appreciate—now more than ever—is that given the amount of company formation and innovation and things that are taking place in private markets relative to what’s available in the public markets, we feel fortunate to be able to review, analyze, and make investment decisions with a portfolio approach at a lot of younger companies that still, if things go well, have their best years ahead of them.

  And certainly what we’ve seen over the last several years in public markets with the coining of the “Magnificent Seven” is that companies that are disproportionately benefiting from the early stages of the AI revolution have obviously had huge upticks in value, and have become quite crowded trades.

  And so we feel fortunate, even though we certainly have tremendous respect for all of the companies making up these areas of the public markets, to be in a position where we can be looking for companies that aspire to make up the Magnificent Eight over time.

  On the right, there’s a graph here that puts into perspective what is probably most compelling about the lane that Groq is focused on. If you think about the description I gave of training versus inference on the previous slide, one of the things that’s so exciting about the prospect for future demand for Groq is that, as you can imagine, in the early days a lot of the activity that’s been fueling the incredible amount of capital expenditure being dedicated to data centers and other resources has focused on training workloads.

  Because we’re in the early innings of the development of these foundational models and all the computational resources that come with that. But as AI continues to penetrate the enterprise environment and gets taken up by consumers who are seeing its value, increasingly the demands of inference are going to start to equal—and eventually overtake—training as an area for investment.

  And so Groq has anticipated this back to its early days. It was a key insight dating back to the time at Google, and back to the idea of looking into the future and trying to skate to where the puck is headed. We’re seeing this demand start to play out, and if you compound an already large number at close to 20%, you realize that by 2030 we’re talking about a significant business opportunity.

  And Groq, as it stands right now as a company that anticipated this, has a significant lead relative to the competition. And we think it’s going to translate into terrific investment performance for all of us.

  Next slide.

  So exclusive access—how we got in.

  As I mentioned, on the one hand, there’s a big part of our business—and I think it’s true generally of a lot of other venture and private market investors—that there’s a lot of companies that are constantly raising money. There’s a lot of deal flow that finds us through our network and relationships. And a lot of those opportunities are very attractive. They’re attractive in part because they are shovel-ready and ready to go. And that does make up an important part of our opportunity set, and we certainly don’t discriminate against the most promising businesses that fit that description.

  But by the same token, we always feel that the best opportunity is the one that we have to go out and find ourselves, and demonstrate patience and persistence and often a little bit of creativity to figure out a way to get in.

  Venture is still a little bit of an investment business that operates like a speakeasy—where top companies try and keep themselves behind a velvet rope and operate with stealth. And in those situations where they can turn away investors and have more money than they really need to raise, it’s incumbent upon us to demonstrate that we can not only deliver on our promises from the standpoint of contributing capital, but bring a lot of value as well.

  And it’s an element of our business where we actually overperform because of the distinct and differentiated aspect of our network.

  This is how it played out in Groq, and the combination of patience and setting the stage with several years of explaining our business model, our interest in the company, our alignment, and what we could bring to the equation as a participant in their cap table gave us access to their Series C, which was not a very well-publicized round and one that had more than adequate demand just based on familiar parties. So a good example is that sometimes it requires patience, but the best things tend to come for those who are willing to invest the time upfront and wait for your opportunity.

  And I will say that one of the advantages of this approach when we do get into rounds a little bit ahead of the more competitive rounds that are more well-publicized and have a lot of people showing up, is that nothing is more beneficial both for a company and for an investor than to have the opportunity to work directly with teams—not when they’re raising money, but when they’re in execution and operation mode.

  You learn a lot as an investor to validate your underwriting case and your initial hypothesis with the passage of a few years before having to make a follow-on investment decision, so you can arm yourself with better insights to make an informed decision. And very often the value of incumbency when a company’s raising more money and wants to reward investors that have been with them for a longer period of time and have been additive—it creates a little bit of a privileged position to be able to get access to deals that become heavily oversubscribed, which was the situation with the Series D that we talked about. So a good example of just the X’s and O’s on the field of the approach that we need to take in order to access companies with this quality.

  Next slide.

  So where are we now and where do we think we’re headed? As I mentioned, the summer of 2024 investment round—which turned out to have the company raising north of $600 million, which is a significant sum of money in venture markets in any environment, but certainly looking at the landscape over the last several years—was a pretty seminal event for a business like this.

  The reason the company was raising that sum of money—A, it was upsized in part due to the overwhelming level of demand to get in—but the company had found itself at a position where it had made significant technological progress and manufacturing progress in the period of time between our 2021 investment and raising this new round, where they were really in a position to transition from being a development/product-oriented stage business to putting their foot on the accelerator to really try and penetrate the market and achieve revenue growth and widespread commercial adoption.

  And that requires capital to do that. Anticipating that, the company made a real interesting and opportunistic acquisition early in 2024, which added a go-to-market component to their strategy called GroqCloud, which is the company building out company-owned data centers and then turning around and selling increments of capacity at those data centers to the growing community of application developers focusing on AI.

  And so this is a way to turn capital expenditure requirements for a lot of businesses into an operating expense and democratize access for inference to what they think is the state-of-the-art solution. And obviously, they want to take care of the community of developers that are going to be building applications—to have them heavily reliant on the hardware as a foundational aspect of their business models.

  So this chart on the right demonstrates the response that the company has gotten from the developer community. It has gone from basically a cold start in the early part of 2024 to now having a community of more than a million developers—which is, I won’t say unprecedented, but is an incredible, incredible curve to appreciate in terms of the market reception from the high-frequency users that are looking for low-cost and fast inference.

  In addition to that, the company has more of a large enterprise-focused model where they’re selling their LPU clusters directly to large-volume users that are constructing their own data centers. And similar to the GroqCloud approach, the market reception to that side of their business has been overwhelming.

  One of the early partners for that side of the business has been Saudi Aramco, the Saudi sovereign that is investing aggressively to build out a domestic AI infrastructure in the Kingdom of Saudi Arabia. And they were one of the early customers committing to clusters of LPU in the second half of 2024. Based on the positive experience with those initial purchases, they announced in the early part of 2025 re-upping their orders to represent a billion and a half dollars’ worth of product in 2025 alone.

  So you get the perspective of the sport of kings that the company is involved in right now—the significant amount of capital expenditure dollars that are flowing into building out this infrastructure—and Groq seems early to be getting more than their fair share.

  I will point out that given all of this activity and public announcements from the company, and Jonathan Ross’s articulated vision for the business and their objectives, he thinks that over time, given the capital that’s being deployed and the number of LPUs that are being introduced into the marketplace, Groq is positioned to capture 50% of the global inference market over time.

  And if you look back on the previous slide that I referenced, that’s a $250 billion opportunity by 2030. So this is a company that we think is going to be growing for a long period of time and attracting a dollar volume of revenue that is really not available to most companies, let’s put it that way.

  Okay, next slide please.

  So given that trajectory, where do we think we are right now? I think as you think about AI broadly as an ecosystem—and a complicated, multi-nodal ecosystem that’s growing as rapidly as it is—there are going to be different bottlenecks and choke points.

  And right now, even with all the capital that Groq has raised and is deploying into building out its capacity and its services, the demand for inference is growing so rapidly as these large language models continue to experience unprecedented adoption across all of their different user communities, that we’re in a situation where we’re short on supply.

  Similar to how Nvidia has communicated pretty consistently with their public shareholders that they are essentially sold out on all of their GPUs for the foreseeable future, Groq is very much in the same situation. They’re going after, in many ways, a different market opportunity, but the demand for inference capacity is nearly insatiable.

  To the point where there’s discussion and rumors out in the marketplace that some of the leading innovators in the arena—like OpenAI—are actually making very deliberate decisions about the release dates for product improvements and new functionality, and thinking about their pricing structure to try and temper the demand because there’s just not the capacity to satisfy it.

  So we feel like we’re in a once-in-a-lifetime cycle of investment. And again, Groq is in a very specific lane where it’s in a position to capitalize on it for not only the foreseeable future—in a situation where they’re going to be sold out of all their products as fast as they can make them—but will have a very, very enviable position as things get a little bit more in balance with respect to supply and demand.

  If people are interested in the company, they do a very good job both on LinkedIn and some other social media sources with interviews and whatnot. There are constantly pictures of huge cabinets and racks of LPU being shipped to all corners of the earth, and it really makes you appreciate the sophistication and the size and the scale of this mobilization of dollars and resources to satisfy, like I said, what right now is almost insatiable demand.

  And the one element that is still very much on people’s minds—and it will be interesting to see the way that it plays out—is that as the company continues to develop on its roadmap and has plans already in place to introduce a second-generation chip, moving down the path of Moore’s Law with respect to more modern chip architectures and node scale, its performance and cost attractiveness is only going to improve.

  And as the cost of inference comes down, the market forecast in terms of the overall level of demand may be rendered completely obsolete—because generally, as costs come down, more use cases and more adoption tend to be the result of it.

  So we like what we see right now but are realistic that the numbers could actually be a fraction of what they ultimately turn out to be, as Groq continues to lead the charge and ride down the cost curve.

  Next slide.

  So the point of this discussion certainly was not to make anybody feel like they’ve missed out on Groq or any of the other companies in our portfolio. I think the reality is—time will tell and we’re certainly focused on it—there may be a chance to continue to support Groq with their future capital-raising activities, whatever shape they take.

  But equally as importantly, I think it’s emblematic of the reality that we have so many companies in our portfolio where we’ve taken the same approach and made an initial investment with the view that we wanted to be early. We wanted to see how things develop, and if our vision of the future starts to become a little bit more clear, we can leverage not only the insights from that investment but the access that we have to promising companies to put more money to work.

  And then of course, we’re constantly thinking through other scenarios in different areas—including AI infrastructure—but the arena of quantum computing comes to mind, robotics comes to mind, and a number of other areas that have tremendous prospects but are still in their relatively early days of presenting investment opportunities. And we’re constantly on the lookout for companies that could evolve into the next Groq.

  It’s a big piece of what we get up and focus on every morning, and I think is a huge piece of the value proposition that we provide to our investors. So we are going to continue to think through ways—every shift, every time on the ice—on how we can skate to where the puck is headed. And hopefully, there’s many more Groqs to take place in our future.

  Next slide.

  So we’re going to bring things to a close with just a little bit of additional information and a recap of our business. If anything that we talked about today strikes you as appealing and you want to get more information, speak to somebody directly—whatever it is that we can do to be most helpful.

  One of the attributes of our business—now serving north of 10,000 and growing individual investors—is we realized that different people have a bunch of different motivations that bring them to venture capital as an asset class and to our firm specifically as an investor.

  For some, it’s the primacy of the role of technology in their lives and in their investing activities. Others are curious and want to learn more and have a relatively low-friction onboarding process to ask questions and develop confidence and maybe start slowly from that perspective.

  Plenty of investors have been tremendously successful over the courses of their careers, and now their focus is to try and set up future generations. And then a handful of other types of personas that we think about.

  The good news is that, given the fact that we have a pretty broad set of product offerings—we tend to organize our capital-raising activities on annual vintages—so we’re constantly in the market, if you will, with different fund offerings, and really have built up a series of approaches that, no matter what your interest is and your investment inclinations, we have the ability to work with you and steer you to a program that is fairly well-customized to your interest and objectives.

  But in all instances, we really focus on the important aspects of company-level diversification, industry-level diversification, and time-series diversification to set you up for the maximum likelihood of success.

  So it’s probably best done to identify what your objectives are and work with one of our teammates to try and explain things in more detail, and we would welcome the chance to engage with you in that type of conversation.

  Next slide.

  Long way of saying—we’d like to be your venture partner. So here’s some information on this slide in terms of what you can do to take that important next step, should you be interested. There are some fund materials that can be accessed via this link and QR code.

  And then we do have a team of very experienced senior partners available that would be happy to get on the phone with you and explain in more detail our strategy and our offerings—or answer any specific questions. If you’re an existing investor and anything that we talked about today strikes you as interesting, you want to learn more, or you’re just interested in checking in on updates for your existing portfolio and have a discussion on how you may look to allocate more money or rebalance it over time—Stephanie and Hillary are also members of our team that are available to get on the phone and walk you through that as well.

  So—ball’s in your court. We hope that you found today’s discussion interesting, and please follow up with us in any way we can be helpful. Thank you.