Eclipse's Lior Susan on $12.5B AUM and the Bet on Physical Industries
Eclipse’s core bet is that venture’s largest opportunity sits in the physical economy—roughly 85% of global GDP, or $100 trillion. Lior Susan argues that manufacturing, energy, defense, transportation, shipping, mining, and semiconductors were neglected while capital crowded into software. Deglobalization, supply-chain vulnerability, government support, and customer demand now make this “the best time in the history of humanity to build those industries again.”
The firm has scaled to about $12.5 billion in AUM by acting as “operators with capital,” not passive venture investors. Eclipse has roughly 90 portfolio companies and helped build 30 of them, while its latest two-fund raise totaled $1.3 billion. Physical companies need operating help across CapEx, manufacturing, supply chains, policy, subsidies, debt, hiring, and systems engineering because “you cannot fake it till you make it” when poor execution appears directly in yield.
Susan rejects formulaic incubation even though company creation is central to Eclipse’s model. “There’s no process,” he says: some founders bring Eclipse an idea, other companies are recruited around an internally developed thesis, and the Rivian carve-out referred to as Mytra in one transcript passage and Mynd in others followed a third path. Each is “a snowflake,” and repeated attempts to manufacture a scalable creation system have failed.
His valuation framework favors market size, differentiation, team quality, and free cash flow over headline gross margin. Susan says Silicon Valley “bullshitted the world that SaaS and gross margin are the most important metrics,” partly by placing continuing engineering costs in R&D rather than COGS. He contrasts crowded software categories with Tesla’s mid-teens gross margin and Cerebras’s roughly 40–50% margin: both can command strong valuations because they solve difficult problems in enormous markets.
Eclipse deliberately avoids sector-box portfolio construction. “My portfolio construction is not to have a portfolio construction,” Susan says; the firm might back five memory companies rather than force one allocation each to defense, chips, and robotics. Its current concentration spans the semiconductor stack, new energy sources, AI infrastructure “picks and shovels,” and physical AI—not foundation models or LLMs.
Physical companies can compound through second and third acts that expand wallet share within an industry. Susan maps Cerebras from chip to system to data center, Redwood Materials from recycling to large-scale energy storage for data centers, and SpaceX from launch to Starlink. Eclipse tries to link these businesses into an internal economy where one deal dollar can become “four or five dollars” across several portfolio companies.
The hands-on promise extends into operational emergencies and a broader US growth thesis. After a cooling leak turned a portfolio data center into “a pool,” Susan flew to the Midwest and spent 72 hours coordinating repairs, down to drying concrete correctly around electronics. He believes physical investment can drive 5–7% annual US growth through factories, energy, construction, housing, and services: “We better not screw up that moment.”
1. Eclipse was built as an operating platform with investment capital
Susan founded Eclipse in 2015 after leaving Flex, then met Pierre while considering a vehicle for building physical-industry companies. Susan was 31, Pierre 85; ten minutes into their first meeting, Pierre declared, “Young man, we’re going to start a firm together.” Susan protested that he had not finished high school and had never invested a dollar. Four months later, Eclipse was born.
Pierre brought experience stretching from Fairchild and National Semiconductor to 35 years at Sequoia. His stories described an earlier venture culture of operators doing few deals, including Sequoia and Kleiner helping save each other’s companies—far removed from today’s calls advertising portfolios that are “up to the right.”
The value Pierre installed was “discipline, discipline, discipline.” Susan connected that directly to manufacturing: “You cannot fake it till you make it” because bad discipline produces bad yield. Pierre also institutionalized board management and quarterly reporting early enough that first-year Eclipse operated more like a ten-year-old firm.
That foundation now supports roughly $12.5 billion in AUM and a recent $1.3 billion split across two funds. Yet Susan still resists the conventional label: Eclipse is “operators with capital,” staffed by people who left industry specifically to build companies in physical sectors.
2. Building 30 companies keeps Eclipse operationally current
Passive capital is insufficient for businesses combining hardware, software, manufacturing, and regulation, Susan argues. Founders may need help with CapEx, supply chains, government, subsidies, debt versus equity, the right engineering team, and the orchestration of a complete system—not merely enough money to build an app.
Eclipse has about 90 companies, roughly 30 of which it helped create, including Bright Machines, Bedrock, Mitra Chem, the Rivian carve-out referred to as Mytra in one passage and Mynd in others, and Peak Energy. Incubation began partly because Susan’s “hands started shaking” to build again; Pierre initially shut the door and lectured him for an hour about discipline before agreeing investors could also create companies.
Susan’s honest answer on methodology is, “There’s no process.” A founder might approach Eclipse with a new HBM-memory concept; Eclipse might identify an opportunity in heterogeneous AI and recruit a team; or it might carve technology out of an incumbent. “Every time I try to build a system around it, I fail.”
Building is also how the investors keep their operating knowledge fresh. Susan believes a former operator becomes stale after ten years away from execution—losing touch with technology, talent, customers, and networks—unless the firm remains directly involved in creating companies.
3. The Rivian carve-out shows how company creation can unlock talent, capital, and neutrality
The company emerged after roughly three years of discussions with RJ about general-purpose manufacturing robotics. Their design thesis was specific: the robot should be mobile and highly dexterous, need not resemble a human, and would require substantial data to train its model.
When Susan learned Rivian already had a small internal team, he proposed a spinout over wine at a robotics dinner. Independence could attract talent unwilling to work on the problem inside Rivian, bring in external capital rather than burden Rivian’s balance sheet, and create a “Switzerland” able to serve every manufacturer.
Eclipse, Rivian, and RJ subsequently carved out the company together. Susan said it was “flying” and planned to show its first product that year—the result of a relationship-driven creation path that would be difficult to reduce to an incubation template.
4. The physical-economy thesis came from China—and arrived later than expected
A year in China convinced Susan that its roughly $20 trillion economy had been built by aligning five forces: talent, policy, capital, technology, and customer demand. A standalone Western company struggles against that coordinated flywheel; Eclipse was his attempt to reproduce some of that alignment at firm level.
Susan expected deglobalization to force countries to rebuild domestic energy, manufacturing, and defense capacity, but conceded that he was wrong about the timing. The shift took longer than expected; COVID first exposed the vulnerability clearly enough for onshoring to accelerate.
His market-size argument is blunt: the physical world represents about 85% of global GDP, roughly $100 trillion. Metal manufacturing alone is a $3 trillion industry, yet when Eclipse started, Susan guessed that quite a lot of capital was flowing into enterprise software and probably zero into metal manufacturing.
Molly O’Shea’s challenge was that newly enthusiastic “tourist VCs” now claim these sectors. Susan’s answer emphasized the difficulty of the work: building a physical company resembles “500 planes that all need to land perfectly,” from wafers and cold plates through PCBAs, full systems, kernel software, customers, and production yield.
5. Free cash flow matters more than the software gross-margin story
Responding to O’Shea’s concern that physical businesses cannot earn SaaS multiples, Susan argued they might do better. “We kind of bullshitted the world that SaaS and gross margin are the most important metrics,” he said; a real public company ultimately trades on free cash flow and EPS, while gross margin is only an indicator.
His criticism is partly accounting-driven: many SaaS businesses sustain growth by continually hiring engineers, but that spending appears in R&D rather than COGS. A business may therefore report an 85% gross margin even when growth remains tightly correlated with cash burn.
Susan instead asks three questions: How large is the market? How differentiated is the product? Does the team have the ability to solve the difficult problem? He said $20 billion is not large enough, guessed a new CRM market at $30 billion, and noted that roughly 1,000 companies compete in CRM.
Tesla, with mid-teens gross margin, can trade like a leading software company; Cerebras, which Susan placed around 40–50%, can trade better than many SaaS businesses. His explanation is not margin purity but “a really tough problem in a gigantic market” for which customers will pay.
6. Eclipse backs clusters, not a predetermined sector checklist
When institutional LPs ask about portfolio construction, Susan gives the same answer: “My portfolio construction is not to have a portfolio construction.” Eclipse could fund five memory companies if conviction warrants it; Susan rejects acquiring one defense, chip, or robotics company merely so the portfolio can claim exposure.
Semiconductors illustrate the cluster approach. Beyond chips sit capital equipment, testing, lithography, packaging, supply chains, and subcomponents. The fact that HBM memory is concentrated among Micron, SK Hynix, and Samsung—while software supports thousands of competitors—signals both the barriers and the opportunity Eclipse believes its operating expertise can address.
Energy is another cluster: The Nuclear Company, Peak Energy’s sodium-ion work, Redwood Materials across LFP and recycling, and an unannounced modern gas-turbine company intended to compete with incumbents. Eclipse also approaches AI from both ends—its infrastructure “picks and shovels” and physical-AI applications—while avoiding foundation models and LLMs.
Susan offered no macro-timing claim: “I know nothing about bubbles and valuation and things like that. I only know how to build companies.” His conviction is narrower but categorical—viewed from the work of company formation, there remains “so much to build.”
7. Second acts and an internal economy can multiply physical platforms
Eclipse initially questioned whether Redwood Materials’ recycling market could become large enough. The next act changed the thesis: rebuild recycled batteries into large-scale energy-storage systems for data centers as CATL and other Chinese suppliers began being blocked from selling storage into “our grid.” Eclipse then led a large financing, while Deepak, a former Tesla CFO, joined Redwood—“the band is back together a little bit.”
Susan sees this expansion pattern repeatedly: Cerebras moved from chip to system to data center; Redwood from recycling to storage; SpaceX from launch to Starlink; and True Anomaly from space defense awareness toward Golden Dome and more vertically integrated programs. Successful physical platforms can introduce multiple businesses and capture more wallet share.
Across the portfolio, Susan said companies signed roughly $40–50 billion of commercial agreements in the prior year, including some double-digit-billion deals. That scale makes customer selection consequential: “You’re married to a very large whale partner,” even as companies have multiple customers in addition to their largest accounts.
The “Eclipse economy” links chips, automated server manufacturing, Oxide’s CPU-based on-prem racks, data centers, nuclear, sodium-ion, lithium-ion, gas turbines, and physical AI. Selling one large deal through three or four companies can turn “a $1” into “four or five dollars”; seeing five-year gas-turbine lead times also directly prompted Eclipse to create a new supplier.
8. Field execution and nationwide growth complete the thesis
Oxide began from the view that security, latency, and cost would keep some computing on-premises while giving customers cloud-like tooling. Eclipse did not foresee AI becoming a catalyst—“Getting lucky is important in this business”—but Susan said Oxide now has multiple customers doing double-digit millions.
The operating promise becomes literal during failure. When cooling leakage left a portfolio data center looking like “a pool,” Susan flew to the Midwest for 72 hours, mobilized service providers, arranged new installations and HVAC systems, and learned how concrete moisture can damage electronics. “I don’t need medals,” he tells grateful founders. “You just must be there with them in the field.”
Susan traces that instinct to leaving high school and working summers on an Israeli banana farm at 16. The workers handled a couple hundred roughly 60-pound bunches a day, and he learned welding to repair the metal cranes used in packaging. Despite the heat, it produced the realization: “I need to build a physical thing. This is my thing.”
His closing macro claim is that a Nebraska data center creates thousands of jobs, then pulls through energy, construction, materials, housing, restaurants, and services. With capital, talent, policy support, demand, and technology aligning, Susan believes the US could grow 5–7% annually—and changed his earlier Henry Ford and Carnegie comparison to call this “the best time in the history of this country to build companies.”
Full transcript
You want to back startups that make actual physical stuff.
The firm is at $12.5 billion in AUM, which might have changed already. I'm not sure.
I hope it's changed—to the upside.
We just named 2 IPOs, both in physical industries: Cerebras, which we did, and SpaceX. Both are doing extremely well. What you cannot fight against is that roughly 85% of the world's GDP—roughly $100 trillion—is in the physical world.
You're an experienced banana farmer.
That's probably my number-one achievement in life.
What were the biggest lessons that you learned from banana farming?
It is hot in summer, but it was also the first time I thought, “I need to build physical things. This is my thing.”
Eclipse has 90 companies total today. We built the majority of them. There are some amazing companies like Bright Machines, Bedrock, and Mitra, and a bunch that we haven't announced yet. We never thought of ourselves as a venture capital firm. We call it “operators with capital” because we are operators.
All right. Lior, welcome to Sorcery.
Thank you. Great to be here.
Eclipse is on a run. Cerebras had a huge IPO.
Yeah.
The firm is at $12.5 billion in AUM, which might have changed already. I'm not sure.
Yeah. I hope it's changed—to the upside. Yes.
For people who don't know Eclipse—
Yeah.
Can you give a little bit of the background?
Yeah. I think the short story is that we started the firm 11 years ago. I know that our limited partners count us as venture capital, but we never thought of ourselves as a venture capital firm. We call it “operators with capital” because we are operators with capital.
We all left industries to go and build a firm that would allow us to build companies in the industries that we're passionate about, and those are physical industries. You're right that maybe in the first 8 years, manufacturing, space, semiconductors, defense, and mining were very much in the corner while software was doing really well—enterprise software, crypto, consumer, and fintech.
It definitely changed in the last 2 years. I couldn't be more excited about the opportunities in the world to build, and it feels like the best founders want to build them. It feels like the world needs more of them. We just named 2 IPOs, both in physical industries: Cerebras, which we did, and SpaceX. They're doing extremely well, so these are exciting times.
Really exciting. We just had Andrew on, and that was a really fun—
Yeah.
Interview.
Yeah. He's funny.
He's great.
He's—
He's funny.
He's really funny. Yeah.
He did name, at the end, when I asked, “Who inspires you?”—who are those kinds of people for you?—and he specifically called out Pierre—
Yeah.
Your partner. Can you tell me more about Pierre? I know we were talking about this off camera—
Yeah.
But this story is insane.
Yeah. The story is pretty crazy. We were this odd couple in the first 7 years of Eclipse.
I was leaving Flex, and I was thinking about starting a vehicle that would allow me to build and invest in companies in the physical world. I got introduced to Pierre. This was early 2015. I was 31 years old, and Pierre was 85 years old.
We had a meeting, and as always, I hadn't done my homework and didn't know who the guy was. I saw this older-ish person step into the room, wearing a suit and full of cologne, and he said, “Okay, young man, what do you want to talk about?”
I said, “I don't know. I got introduced to you, but I'm thinking about starting a firm or a vehicle that would allow me to build companies in the hardware space.”
In a very Pierre fashion, after 10 minutes, he interrupted me and said, “Okay, young man, we're going to start a firm together.”
I said, “Sir, I actually—A, I didn't finish high school; B, I never invested a dollar in my life. I'm a builder. I know how to build. I'm excited to build companies in these sectors.”
He said, “Hey, I will teach you everything you need to know about that industry.” About 4 months later, Eclipse was born.
Pierre is a special, special man. As I told you off camera, I'm taking him to dinner tomorrow for his 96th birthday, and he's still mentoring most of us on the team.
He worked with us full-time in the office for 7 years and brought a lot of what I think this industry originally was, from when he started at Sequoia. Don was his VP of sales at National Semiconductor, and they left to start Sequoia. Pierre was there for 35 years.
We had the huge privilege of hearing his stories and learning from him about what that industry was originally designed to be. It's actually very different today. They were all operators. They were all hands-on. They did very few deals. They had very gentlemanly relationships between the firms.
He told us so many stories about Sequoia and Kleiner saving companies for each other, something that—
What?
Can you imagine calling another VC today and saying, “I have an issue with this company. I need you to come put capital in and help me save that company”?
It will never happen. You'd probably get a phone call saying, “All of my companies are doing amazing, up and to the right. Come preempt around.”
Pierre and I were that odd couple. I said on camera on his 95th birthday that, for me, he's somewhere between a father, a partner, a mentor, and a friend—kind of all of the above. He's probably one of the most amazing people I have ever met in my life.
And what kind of work culture did he bring—
Yeah.
To Eclipse?
Yeah. It's—I grew up in the Special Forces. He grew up during World War II. He got himself moved from France to the United States for school, met Gordon Moore, and got into Fairchild. He left Fairchild for National Semiconductor, and left National to start Sequoia.
I say the best 3 values I can think of about Pierre are discipline, discipline, discipline. Where I grew up, discipline was really important, and that also connects nicely to the companies that we like to build in the physical world.
You cannot fake it till you make it when you're manufacturing something, because the yield will just be very bad. He brought a lot of institutional knowledge about investing and managing a board, and how to write your quarterly reports. In some way, in our first year of Eclipse, we already operated in a way that maybe a firm that's been around for 10 years would operate. Everything was much more institutional.
But he also has a huge heart, although he tries to hide it.
Really?
Yeah. He really cares. You don't need a 90-year-old person to go to the office every day to teach other people how to do venture. You don't.
Yeah. Well, it seems to be working.
Yeah.
So you were founded in 2015.
Yeah.
You raised your last fund—it was split between 2 funds—about 3 months ago.
Yeah.
$1.3 billion.
Yeah. Mm-hmm.
Okay. I want to go through your investing strategy, because you did say you're operators with capital.
Yeah.
It's completely different from the traditional kind of VC game that's happening here in Silicon Valley, because of that hands-on approach.
Yeah.
So when you talk about venture equity and this incubation strategy, how does that work, and what are some of the companies that come out of it?
Yeah. I would say I don't know if that model is perfect for everyone, just to be clear. I do think that companies building in the physical world shouldn't get only passive capital on their board.
You need to know manufacturing and supply chain. You need to know how to deal with CapEx. You need to know how to deal with the government. You need to know how to think about subsidies. You need to know how to leverage debt versus equity.
You need to know how to hire the right engineering team, regardless if it's a mechy, Ws, and how to think about it holistically, because you're building a system. You don't build an app that, if you're a good enough founder, you just take the capital and build a great app, and there is a great outcome.
I also think that if you're looking at the generational change, we started kind of with mobile to cloud. It's like internet to mobile to cloud to SaaS enterprise software. What we've been seeing over the last 10 years is basically this full stack of tech companies taking on a full industry.
Cerebras is taking the compute industry. SpaceX is taking the space industry, and there are a bunch of others. These are much more complex businesses to build, and much bigger if you're successful.
Our model was that we were on the other side as operators, and we thought about what would have been the most optimal things we could have had from our investors in our previous lives. We felt like we just wanted to bring that model: operators on your board, bringing high conviction, a lot of capital, and a lot of expertise and knowledge about how to build those businesses.
That’s the model. On the incubation side, we’re two years into it, and it’s actually also a funny Pierre story. I’m feeling my hands start shaking, and I’m like, “God, I need to build again.” So I go to Pierre, and I say, “Sir, I kind of feel like building a company.” And he shuts the door and says, “Young man, we are investors now. We are not building companies anymore.”
And I’m like, “Can we do both?” And he’s like, “No.” He lectured me for an hour about my lack of discipline, and I was able to convince him. It’s just amazing. I think Eclipse has 90 companies in total today, and 30 of them we built. There are some amazing companies like Bright Machines, Bedrock, Mitra Chem, and Mytra, as well as a bunch that we haven’t announced yet. Peak Energy, which we talked about in the battery space, is also a company that we incubated.
We did it from a selfish point of view because we love building companies, and I didn’t want to stop building companies now that I also have the investor hat. And, B, it’s about keeping our operating skills fresh. When you make the transition from being an operator to being an investor, 10 years into it, you’re not a fresh operator anymore—not in terms of where the technology is, or where the network, talent, or customers are. Building companies keeps us fresh.
What is your process for starting the companies?
There’s no process.
But how do you pick the categories?
I don’t think you can manufacture this thing. That’s what I learned in the last 11 years of doing it inside Eclipse, inside the firm. Every time I try to build a system around it, I fail. Every time I think, “If I only put this thing here, I would be able to build 2 more,” I fail.
Each of those things is a snowflake. In some cases, someone comes to us and says, “Hey, I want to build this new HBM memory, and I want to do it with you,” and I’m like, “Amazing.” In some cases, I’m like, “The world is going into heterogeneous AI. If I can connect multiple pieces of silicon and optimize them from the networking to the kernel to the orchestration, I’ll be able to build the most efficient, high-performance, lower-cost AI data centers. There’s no company that exists, so I’m just going to build it and recruit the team.”
And in the case of Rivian and RJ, we carve out Mynd outside of Rivian with him. Those are 3 very different models I just described, and there are 10 more.
Damn. Can you go into one of those companies?
Yeah.
Maybe Mynd.
Yeah. The one thing we really like to build, I think about it as clusters. It’s an industry that we know. A lot of us built robotics in our operating lives, and at Eclipse we’ve been doing robotics and physical AI for a long time, before it was called physical AI.
We’re friends with RJ. G10, one of the founders, one of the partners of the Eclipse, was on the founding team with him. I’ve been talking with RJ for a long time about robotics and manufacturing, and we probably talked for 3 years before we actually started Mytra together. We were talking about general-purpose robotics in the manufacturing landscape. It needs to be mobile. It doesn’t need to look like a human or be humanoid. It needs to have high dexterity, and you need a lot of data in order to train the model.
As we were talking over the years, he said, “You know, we have a small team inside Rivian that’s focusing on that.” We had a robotics dinner at my house with him and a bunch of other founders, and at the end of dinner, the wine helped. I told him, “What do you think about us spinning out this thing?”
He said, “Huh.” And I said, “You can win all the wars. We can attract better talent that won’t necessarily go work on that problem inside Rivian today. We can attract external capital, so it’s not on your balance sheet, and it’s like Switzerland. You can work with everyone.”
Naturally, the rest is history, and we were able to carve it out and start the company with him and Rivian. The company is just flying. We’re really excited to see all of the pieces coming together, and we’re going to show the first product this year.
It’s just so crazy to hear how close to the metal you actually are.
Yeah.
I’ve had many conversations with many different CEOs and investors and that sort of thing, but I don’t even know what you said previously, when you were talking about building a data center.
Yeah, sorry.
I don’t know what you’re saying.
I went too technical.
But that’s awesome.
Yeah.
Taking a step back on the macro scale of things, I talked about this with Andrew on the pod. I talked about this also with Tony Kim of BlackRock.
Yeah.
We were going through all the different categories, and it’s a common theme that’s very ever-present between everyone. The original industries that built up Silicon Valley really were in the shadows for decades.
Yeah.
And now they’re back again. Now you have—I’m not going to say tourist VCs—but you have a lot of these competitive VCs—
Yeah.
—coming in and claiming territory in them.
Yeah.
But how—if you’re in the founder’s seat, how do you decipher who to work with and where they actually have expertise?
The reason I started Eclipse was actually from a fundamental view. It was also on the back of me spending a full year in China and seeing how they innovate in the physical industries.
People don’t know this, but China essentially built a roughly $20 trillion economy based on the physical world. They built the entire system on these 5 forces: talent, policy, capital, technology, and customer demand. They align those 5 forces in a way that lets their economic flywheel fly, and it’s very hard to compete against them as a standalone company.
My view was that if we don’t do it in the Western world, we’re kind of screwed. Of course, it was also from a selfish point of view because I love building these businesses. I was never a software guy; I was always a hardware guy. There’s just no reason not to do it.
After talking with a bunch of firms as I was leaving Flex, it was clear to me that a lot of my generation came much more from the software world. That world sounds very scary to them: low gross margins, high CapEx, and long sales cycles. But that’s what I love doing.
My view was also that as the world is moving more from globalization to deglobalization, countries will not be able to rely on other countries for their energy, manufacturing, and defense. It was clear to me that it was going to happen. I was wrong about the timing—it took much longer than I thought. COVID was really the first thing that showed the vulnerability, and as a result, we started seeing much more onshoring kick in.
What you cannot fight against is that 85% of the world’s GDP—roughly $100 trillion—is in the physical world. If you’re a founder and you’re thinking about the biggest problem you can solve, you think, “What is the largest TAM?” The largest TAM is manufacturing. The largest TAM is energy, defense, transportation, shipping, and mining. Agile software, nothing about all of the agile software people in the world, is just combined less than 20 billion. Vulcan Forums that working on metal manufacturing, metal manufacturing is a $3 trillion industry.
When we started Eclipse, I don’t have the numbers, but my guess is that quite a lot of money was going into enterprise software and probably zero was going into metal manufacturing. It just didn’t make sense to me. That goes back to the question of what founders who are starting companies want to think about.
It’s fucking hard. It’s really hard to build these companies. It takes time, and you have this 500-plane situation where all of them need to land perfectly. That’s what we do. The wafer needs to show up to the cold plate, to the PCBA, to the full system, to the kernel software, to the customers, to get the customers, to get the yield, and everything needs to work perfectly.
It’s really hard to do, but I cannot do anything else in life. That’s the only thing I’m passionate about.
It is interesting. SK Hynix recently—
Yeah.
—went public.
Went public. Yeah. Yep.
And when you look at the memory world, there are literally 3 players.
Yeah.
We’ve gone so long with only 3 players, and there’s a massive shortage—
Yeah.
—as you would know—
Yeah.
—in the chip industry.
Totally. And that was kind of my point. It was like, “Okay, great: Micron, SK Hynix, Samsung,” if you’re looking at HBMs. When I started Eclipse, I was like, it’s really odd because there are 10,000 SaaS companies. How can it be that there are only 3 memory companies?
When you actually go deep, you’re like, “Okay, you need CapEx, you need a lot of engineering, you need the 5 forces I was talking about.” That’s what Eclipse is expert in, and we’re building new memory companies. It feels like the best time in the history of humanity to build those industries again.
So how do you then overlay the venture model on top of these industries? And then how do you refute the thesis—or the pros, I guess—that people say these industries aren’t going to produce SaaS multiples? I know that’s technically flipping now.
I actually think it’s—we’re going to introduce—
With AI, then—
—I think it’s better than SaaS multiples.
But could you just explain that?
Maybe it’s a little bit connected to the “tourist” comment that you made before. I think we, as Silicon Valley and the venture industry overall, kind of bullshitted the world that SaaS and gross margin are the most important metrics.
Naturally, if you are a public company, if you’re a real company, what matters is free cash flow. It’s actually not gross margin. Gross margin is an indicator of potentially what your free cash flow is going to be. You’re going to trade based on that, and EPS is based on your free cash flow.
Naturally, if you tell me you have an 85% gross-margin business and you are, you know, not in agile software, you are sales software now, we're trying to move so they don't think that I have anything really against agile. You're building a new CRM again because there is a million of them. The TAM itself is not huge. My guess is $30 billion. There are 1,000 companies doing CRM.
What happens with a lot of those SaaS companies is that their growth was correlated with the way they were burning money. Why is that? We mechanically moved the engineering spend from what’s supposed to be in COGS in any normal business to R&D. It was an accounting trick, so your gross margin is high, but you still need to continue hiring engineers as the business is growing. In our world, you cannot do it. There is hard COGS, and accounting will not let you.
Maybe your gross margin will not be as high as a SaaS company, but at the end of the day, I’m a fundamental entrepreneur. On a fundamental basis, I always ask 3 questions: How big is the market? $20 billion is not big enough. How differentiated is what you do in that market? And do you have the right team to actually go and solve that really tough problem?
Because Tesla is what—mid-teens? It kind of trades like the best SaaS companies in the world. Cerebras is, like, what it is—I can’t really remember now—40% or 50% gross margin. It trades way better than most of the SaaS companies because they solve a really tough problem in a gigantic market, and they get paid for it. That’s what I do for a living.
So I want to go through some of your portfolio—
Yeah.
—because you have some really fun names in there. Cerebras is obviously one of them.
Yeah.
You have Ursa Major—
Yeah.
—Vulcan Forms, which you mentioned, Bright Machines—
Yeah.
—Augury, Skyrise, Oxide, which is really interesting. We should talk about that one.
Talk about it, yeah.
WAVE, Bedrock, True Anomaly, also very cool.
Yeah.
And the same with Redwood Materials—
Yeah.
—which is the next one I have, and Arc, The Nuclear Company, Verkada. You also have Anduril on there. But I guess, to start with all of those and the different categories you’re going after, let’s base it off the latest fundraise. Of the physical industries that you’re going after, what are the main categories?
When you raise money in our world, we raise money from what’s called E&F—endowments and foundations—and pensions, institutional investors. They always have this question in the diligence about portfolio construction.
Yeah.
I always have the exact same answer: My portfolio construction is not to have a portfolio construction.
The main point here is, we might do 5 memory companies now. That will not be a very good portfolio construction. But people are calling me and saying, “Hey, can you introduce me to Bedrock?” And I’m like, “Okay, why?” They’re like, “We must have a physical AI company in the portfolio.”
I’m like, “I don’t need to have anything. I need to have the companies I’m obsessed with,” unrelated to trying to say, “Oh, I must have 1 defense, 1 chip, 1 robot.” We don’t think like that.
What I think is really interesting in the market right now is that we went really deep historically on semiconductors, and I think people think about semiconductors as chips. That’s kind of what you have. There’s so much around that: capital equipment, testing, lithography, supply chain, packaging, and subcomponents that go into this thing.
That world is really, really big, and it’s growing rapidly. We just talked about SK Hynix as a good example, so we really like it.
We also really like a lot of new types of energy sources. We think that energy is such a gigantic market. We did a play in nuclear, and we have a play in solid ion, like Peak Energy and TNC. We have Redwood in LFP and recycling, and kind of a second act.
We just did a gas-turbine company that’s building a new, modern gas turbine, going to compete directly with GE, Von Ovo on a very large gas turbine. I think, in the energy sector, there’s just a lot to do.
AI is definitely another area. We’re not doing anything in foundational models or LLMs. We kind of catch AI on both sides: picks and shovels for AI, which we do a ton of, and then we take AI to the physical world with physical AI. There is so much to do.
The only thing is, I know nothing about bubbles and valuation and things like that. I only know how to build companies. But I’m telling you, from building, there’s so much to build there.
What are you most excited about?
I told that to my daughter. She sees all these people coming to our house all day—the people I’m building companies with. She’s 11 years old, but she can take you down on term-sheet negotiation. Dangerous. Dangerous.
She asked me, “What do you think? Which phase are we in building companies?”
Damn.
Yeah. Yeah.
Wow.
Yeah. Yeah, yeah. Because she was born into it.
Yeah.
Our oldest was born—
She’s like, “Dad, how’s the AUM doing?”
Yeah, exactly. “How are we doing at a DPI level?” Yeah, yeah, yeah. She’s very close to asking that question as well. That’s actually her mom. It’s not her, but this will be her as well.
She was actually born into Eclipse, so she’s exactly the same age as the firm.
Oh, my gosh.
It’s kind of super cool. We have an Eclipse Kids Day, so every year we take all of the Eclipse kids to visit 2 or 3 of our companies.
Oh, that’s so fun.
Then we end up at my house with a little pool party, and it’s actually next week. It’s really cool to see the kids looking at how a robot like Mitre moves a pallet, or seeing Reliable, how planes are moving, or seeing Cellaris, how drugs are being manufactured, and then they…
Those kids are asking some really good questions. So she’s knowledgeable. Anyhow, what I told her was, “Listen, we are probably in the most interesting time that humanity has ever had around building. We are at the cutting edge, at the very edge, of building those businesses, and we better not screw up that moment because we have capital, talent, government support, customer demand, and technology.”
I do believe the U.S. can grow 5%, 6%, 7% every year. I’m going to work really hard to see it come true. I think growth, both for companies and countries, solves a lot of problems, because you can spend a lot of time focusing on where to cut spending as a country or as a company, but you can also grow much faster.
I think it’s maybe the first time—I used to say Henry Ford and Carnegie, and I actually changed it to: this is the best time in the history of this country to build companies.
That’s very optimistic.
Yes. Yes, I know.
That’s good.
Yes. I’m super optimistic about this country and opportunity.
We just had David Friedberg—
Yeah.
—here recording, and he went off the rails on that.
Yeah.
You’ll have to listen to that one. I can’t share too much, but—
Yeah.
—it was great, and he was also talking about how, for the most part, in America, the economic system is backwards. You should not be getting 40% tax on labor and 15% on capital gains. It should be the opposite.
Yeah.
It should be totally the opposite.
Yeah.
I do want to go through some of your portfolio companies. Let’s start with Redwood Materials.
Yeah.
That’s a really interesting story—
Yeah.
—in the energy space. It came from Tesla.
Yeah.
How did you guys get involved?
Yeah. Actually, Joe, one of our partners, worked with JB for a long time and served on the board, and Greg, one of the other partners, ran operations and manufacturing for Tesla for the first 8 years with JB. So we always had a very close relationship with JB and the Redwood team.
They originally started with recycling. We had some questions about how big that market could be on its own, but we always knew they were super talented. Then they started talking with us about energy storage, basically taking the batteries they recycle and rebuilding those batteries for very large-scale energy storage for data centers.
That was also when CATL and the Chinese companies started getting blocked from being able to sell energy storage to our grid. By the way, good idea. We felt like the moment was right, and we went out and led a very large round there.
We love working with JB and the team. Recently, we’ve actually had other friends join the company. We brought in Deepak as the CFO; he was the CFO of Tesla for many years. So the band is back together a little bit there, and it’s really exciting.
Second act and third act. The first act for Cerebras was to build a chip. Our second act was to build a system, and our third act is to build the data centers that run those things. For Redwood, the first act was recycling. The second act was taking that and selling it as energy storage.
SpaceX’s first act was launch; the second act, naturally, was Starlink. True Anomaly’s first act was space defense awareness; the second act is Golden Dome and other programs that are much more vertically integrated. When you build something in the physical world, if you’re successful, you have the ability to introduce multiple new businesses that will grow your wallet share in that sector.
Most of these come back to AI and data centers, to your point before.
Yeah.
How do you think about customer differentiation, like the moats and the long-term durability? Or are you just focused on building them now and servicing those customers? How do you think about that dependency?
Naturally, in the sectors we operate in, there are no $10,000 ACVs. This is not that kind of business. Last year, we signed many tens—I think roughly $40 billion or $50 billion—in commercial deals across the portfolio. These are very large deals. Some of them were double-digit billions.
Of course you care who the customers are, because you’re married to a very large whale partner, and you want to make sure that the customers will do well. But we talked a little bit about Oxide. What Oxide is building is an on-prem CPU-based rack that gives you a software-based approach to on-prem. You feel like you operate in the cloud world from a tools perspective, but you’re actually running it on-prem.
Our view when starting the company was that not everything is going to move to the cloud. There are a lot of industries that will still want to do on-prem for security, latency, and cost. We didn’t know AI would be a thing, to be clear. Getting lucky is important in this business. We’re trying to get lucky as much as we can.
That business is growing so fast, and it’s not just 1, 2, or 3 customers. They have multiple customers doing double-digit millions of dollars, and they have one customers that are doing many hundreds of billions. This is actually very typical for us to see in our portfolio.
You have a whale account that will do many hundreds of millions of dollars in a deal, or TCV, and then you have multiple other customers in the tens or hundreds of millions of dollars. We’re trying to move as fast as our customers are moving.
Yeah, it is interesting that on-prem is coming back—
Yes.
—into fashion after so many years of large legacy financial institutions and people like that fighting it—
Yeah.
—fighting cloud salespeople, really.
Yeah.
Now their time is back to prove everyone wrong about on-prem. AI sovereignty was in the AI Act and the AI Action Plan, and then it reemerged when Karp was talking about tokens specifically for models and that kind of dependency.
For the most part, owning the stack and, especially in the U.S., owning the manufacturing, the infrastructure, and everything that’s being built for the models—it’s obvious.
Yeah. We actually helped write a little bit of the AI Act and collaborated with the administration on that, and we call it the Eclipse economy.
If you’re looking at AI infrastructure overall, we have companies building chips, manufacturing racks—the automation to build those AI servers—taking those servers and building racks and CPUs, like Oxides, building the data centers and running them, like this new company we haven’t announced yet that I was referring to, and building the energy sources for those data centers, from nuclear to sodium-ion, lithium-ion, and gas turbines. Soon, we’ll announce another energy storage company at scale.
On top of it, of course, they all apply AI and physical AI that will take whatever those data centers produce and use it in manufacturing and robotics. That Eclipse economy is giving us, I believe, an interesting view because we are, in some way, competing with China in our own way. We’re not a country; we’re a firm. But we are, like—
Yeah.
—we do 1 very large deal, and we sell the deal with 3 or 4 portfolio companies. Essentially, $1 becomes $4 or $5 because we’re leveraging multiple companies.
Because we’re so involved in those companies, we can shape how those very large deals happen or how the policy happens. I really enjoy doing that, first, because it’s helping us and helping companies like Cerebras grow much faster. But the other thing is that it’s giving us new ideas about what else we should build.
That gas turbine is a great example. We saw that gas turbine lead times were insane—5 years. All the companies were fully backlogged, and we thought, “We’re actually deploying a lot of them. We should build a modern company to make gas turbines.” So I’m loving it.
Do you get pushback from maintaining and creating your own economy out of it?
Not yet. If we do, that means we probably did something well.
Yeah, I would imagine.
I think Elon is having his own little economy, right, between what he’s doing with SpaceX, xAI now together, and Tesla—leveraging all the companies, all the talent, and the entire customer base.
So I actually think it works well. I'm sure he gets pushback, and we'll get pushback.
Is there anything that we haven't covered yet that you're excited to talk about?
I think probably the only other thing I'll go back to is that I'm pretty optimistic about America. One thing that people don't understand outside of Silicon Valley is that when you build in the physical world, you actually have an impact on the rest of the country. I'm so sick and tired of the pushback on data centers. I think it's so stupid.
Yeah.
When you build a data center in Nebraska, you create thousands of jobs, and it's a circular economy. It's like building a factory. Then you need energy to serve that thing, construction, and materials. Those people need to live, so you need housing and service providers and restaurants. The whole flywheel of the economy is actually kicking in.
That's what China did really, really well. I think it's the first time, maybe in history, that we have an opportunity to do it at scale here in the US, and I'm just really excited about it.
That's amazing. Before we close out, I have 2 more questions. One of them is that you're an experienced banana farmer.
That's probably my number 1 achievement in life.
What were the biggest lessons that you learned from banana farming?
Shit, it's hot in the summer. Carrying those things in the summer in that part of Israel is not a fun job.
How many bananas can you carry at once?
There are 2 people: 1 who does the cutting, and a second person who basically grabs the entire thing. I don't know what the name of that is in English. You probably do a good couple hundred of them a day.
Really?
This thing weighs 60 pounds. As a young, 16-year-old rebel and high school dropout, it was good discipline to work there in the summer for 2 years and just work on that banana farm. But that's also the first time I did welding.
Welding at the banana farm?
Yes, because you need to weld those metal cranes that you hang those things on.
Uh-huh.
Then you take them to the packaging. By the way, welding in the summer on a banana farm is a really bad idea, and it's hot. But it's also the first time I was like, "Hey, I need to build a physical thing. This is my thing."
Mm-hmm. The bananas don't melt in the sun?
The bananas don't melt. We cut them green, so they're more polished than the yellow ones that you're buying at Trader Joe's. It actually takes a couple of weeks until they show up for you to buy.
Were you thinking at all about, you know, you're well into your career and you've got a large fund now, that you would start a banana farm?
Would you go back to the roots?
Yeah.
Actually, no, I'm not thinking about it.
I think there are, like, banana problems.
There are banana problems. There are banana problems. My funny thing about that that I always say is, if I were very bad at building companies, worst case, my kids are going back to the banana farm. Actually, it was not that bad, so we—
Have you brought them?
Oh, yeah. They go every summer. My parents still live there, and my sister still lives there in the kibbutz. They go, and it's like, "Hey, Dad works in these banana farms." It's like, go teach them. See how the trees grow, give them a lecture—and they don't care.
To push even further on how you support the portfolio companies, what are the war stories in the trenches there? How are you not just helping out with go-to-market?
We work alongside the teams. We manage a fairly small portfolio, so we can be super hands-on and bring our operating expertise into the day-to-day. I talk with most of our CEOs multiple times on a daily basis.
You get a lot of kicks in the face in the type of companies that we build. A recent example is a company that we're building right now that's involved in data centers. I get a phone call, and they say, "Hey, we have a pool here rather than a data center." It's basically a massive leak in the water-cooling system of that data center.
I'm putting myself on a plane. Immediately, I'm flying to a place in the Midwest to find the data center that absolutely looks like a pool. It was 72 hours of me pulling together our entire network of service providers, new installations, and new HVAC systems, and figuring out things that I'd never actually done before.
You need to dry concrete in a certain way in order to not have a moisture stain in that concrete that can impact the electronics.
Wow.
It's a thing. I was there for 72 hours without my family or anything else, just on the floor until that thing got fixed.
One thing for us is that I always get a lot of emails from founders or management teams that we work with saying, "You were so instrumental. Thank you so much. That was heroic." I'm like, "Guys, I don't need medals. I do it because that's what I'm passionate about." When you know how to build these companies, you just must be there with them in the field.
Are they surprised that you go and do this?
Not anymore.
Not anymore?
Not anymore. No, not anymore. They're actually expecting me to do it now. There's an expectation: "Lior, by the way, this thing—" I'm like, "Well, it's not a big problem. You should be able to solve it by yourself."
"Our power's out. Please come help."
Yeah, exactly. Exactly.
You're like, "What?"
"We ran out of snacks in the kitchen." I'm like, "Okay, that's not a Lior thing to fix."
"How do you reboot the Wi-Fi?"
Yeah, exactly. Unplug it and plug it back in. Exactly.
That's amazing. Wow. As we close out, I have one final question. It's one of our partner questions. So Sorcery is sponsored by Brex. They're the performance corporate card. Can you believe that? It's amazing.
Yes.
Molly O'Shea Spending smarter and moving faster. Anyways, so I take this question on performance on the personal side of things. I asked Andrew this, and he said Pierre and some other individuals.
I believe that you're really a subset of the people you surround yourself with—who you're inspired by, who you're mentored by, your friends, your family, those kinds of things. As you set out to build a very successful and ambitious career, you're obviously building over and over again. Who are those people for you?
Pierre was a huge one for me, but since Andrew stole the thunder, I won't be able to use Pierre. That would suck.
I actually spend quite a lot of time with Sam at OpenAI. I know that there's a bunch of points of view about him and about the company. I actually think he's really special, and I think he's also a really good guy with a good heart and a great builder.
I think a lot of the things that OpenAI did, and other amazing companies like Anthropic, are really important to our ability to build more companies and our ability to shape communities. To see my parents using ChatGPT is unbelievable. So, definitely a source of inspiration for me.
That's amazing.
Yeah.
I wasn't expecting that one.
Yeah, yeah.
Maybe you were going to say a robot or something.
I don't know.
Yeah. One that we are building.
Well, Lior, thank you so much.
Always a pleasure.
It was such a pleasure. I loved all the stories.
Thanks for having me.
Hey, it's Molly. If you enjoy our interviews, check out our newsletter, sorcery.vc, where we deliver a once-a-week top deals and tech headlines email, and also go deeper on our podcast interviews. Subscribe to Sorcery today. And don't forget to subscribe to the podcast on YouTube, Spotify, Apple, or wherever you listen. Link in description to sign up.