As a founding managing partner of Celesta Capital, Sriram Viswanathan knows a thing or two about deep tech.
Founded in 2013, San Francisco-based Celesta is focused exclusively on deep tech, built around the conviction that major technology shifts require innovation at the infrastructure layer. With assets under management of $1.1 billion, the operator-led venture firm has made 110 investments to date with 43 exits — an impressive exit-to-investment ratio.
Notable exits include chip company Credo, AI processor Habana Labs, semiconductor company Innovium, drone systems company IdeaForge, and digital cell biology startup Berkeley Lights.
However, with startups staying private longer than ever, it’s clear that the secondary market is becoming an increasingly critical engine for venture, particularly in capital-intensive sectors such as deep tech.
Viswanathan sat down with Crunchbase News to discuss not only why the secondary market is accelerating now, but also how, in particular, that is impacting deep tech companies and where opportunity lies in deep tech in general.
“Celesta has been doing deep tech from day one, and we were always investors in hardware and semiconductors and systems and the like,” Viswanathan told Crunchbase News. “And so in a way, we’re thrilled that what we do is literally at the core of what the opportunities are and what people want to do.”
The interview has been edited for brevity and clarity.
Crunchbase News: Why do you think the secondary market has been so active, and what’s driving demand?
Viswanathan: It’s a pretty obvious outcome that when you have more capital chasing fewer deals, more investors are looking for exposure. Right now, this is kind of a pendulum. There are times when capital is abundant, and there are times when it’s not.
And so the entrepreneurs struggle in trying to get their companies funded. We’re right now in a period where, thanks to AI and infrastructure on the data center side, there’s just an incredible amount of capital flowing into the innovation space. AI and healthcare and data center buildout — all of that is just exploding. As a consequence, what you’re finding is that there are fewer good companies to go after. And as a result, some of the more discriminating investors are looking for greater exposure and letting earlier investors get out.
You’re seeing early-stage venture capital firms that traditionally looked at M&A or public listing as an avenue. Now, I think there’s a substantial amount of secondary transactions that rival M&A, which is, I think, a good development because now you have a newer mix of investors, you have deeper-pocket investors coming in and providing the liquidity for the early investors.
Let’s talk about the secondary market as it relates to deep tech specifically.
Historically, secondary markets have been driven by larger investors looking for proven business models, proven cash flows, and a time to liquidity in the two- to three-year timeframe. Remember, the secondary players are really larger asset groups that have different return expectations than the earlier-stage investors.
But as it relates to deep tech, what’s happening is that there’s a feeding frenzy, because people see what happens in AI training. They think that a similar opportunity might exist in AI inference. At the same time, that is true.
In some cases, you’re going to see certain vertical applications get built, and in the secondary market, people that are taking a bet on the AI space are not really focused on revenue stream or cash flow. They’re looking to see if, “Is it a winner?” It’s a take-all kind of market. And that we have seen happen in the LLMs. We have three large players now. It was one, but it’s unlikely that it will become 10.
I think the secondary market demand will also decrease in AI, over time. But it is different than the traditional secondary market investors who have focused on cash flow and revenue growth. That’s not what’s happening in AI.
Do you think this hot market is going to continue in the near- and long-term?
There’s always a reversal to the mean, as you would expect. Things always swing a little more than the equilibrium will support. As a result, you will find that there will be some froth. Valuations creep up. There’s more capital chasing it, and people are forgetting the back to basics, which are building a meaningful business, meaningful revenue, meaningful free cash, and all of that. And the hype cycle sets in. We may be in that period, in certain segments of the AI infrastructure buildout.
But we’re starting to see a dramatic improvement in how quickly a company can actually test its products. The time to revenue has shrunk dramatically, right? And the cost of development is also reduced dramatically. So when people talk about a one-person, billion-dollar company, you’re going to see a lot more of that, because the cost of innovation has really plummeted.
The flip side of that story is that the race to build out, clearly, is provisioning the supply side of the infrastructure to be in excess of demand in substantial cases. So the demand has to really catch up with the associated economics. And if the demand doesn’t catch up, overbuild happens. So I think we’re likely to see capacity in excess of demand. As it stands today, that’s going to be a pretty big problem. Deeper-pocketed investors are going to be able to sustain it, like the hyperscalers or the neocloud guys, but quite a bit of the other smaller players are going to really get a whiplash.
When it comes to deep tech, what other areas besides AI do you think are really interesting right now?
When I think about deep tech, of course, AI is the hottest space right now, but there are other deep tech sectors that are getting a lot of attention, such as biotech, for example.
We think there are three real broad sandboxes.
There’s a whole area of hardware systems, intelligence, infrastructure and data center AI — all of that physical hardware infrastructure that’s one sandbox, which is above and beyond just AI.
The second area is really the software layer, above and beyond the large language models or the frontier models. You can think of them as AI tools or AI software capability as an operating system, but on top of it rides a lot of applications and services and key verticals, such as healthcare, retail, diagnostics and supply chain. These are applications and services built on top of some very intelligent AI systems. That area is profoundly exploding. For instance, AI-assisted radiology, AI-assisted marketing and go-to-market strategies, AI-assisted financial services and looking at payment infrastructure.
And the third area is how biology is getting influenced by deep tech, whether it’s robotics, or advanced surgical capabilities, or diagnostics. You’re looking at miniaturization and manufacturing of complex systems, such as CT scan and Xray machines, and radiology equipment like ultrasounds — just all kinds of equipment that are getting enhanced with AI and the cloud as a connected device.
How do you think these prolonged hold periods are reshaping behavior for LPs, GPs and founders?
There have been these prolonged hold periods. Everyone wants DPI (distributed-to-paid-in capital). I saw this mug that said DPI is the new IRR. So everybody wants distributions earlier. And as a consequence, secondaries are a very important part of that market-clearing that has to occur.
Given the performance of the public markets — and the associated liquidity that people enjoy — there is a greater impatience by large investors to see traditional venture timelines get shrunk. The only way to do that is to leverage secondaries. So, there’s greater desire, greater opportunities and a greater impatience, which is in a way healthy, because you’ve got to return money to the investor.
Now the flip side of that is that the public markets have had phenomenal growth over the past seven to 10 years. And you know, it’s arguable whether that’ll continue, so you’re going to see people get back to the early-stage illiquid asset class as an opportunity. But right now the public markets are very attractive. I would expect that to change over the course of time, but for now, that is actually bringing a healthy sort of expectation on distribution.
What do deep tech companies uniquely need in this environment?
Again, back to basics. You can’t continue to build on AI as a sector if you cannot show revenue growth, if you cannot show margin growth, if you cannot show scale. So the core technology by itself is necessary, but not sufficient. You have to have technology adoption supported by revenue growth.
For example, see what happened in SaaS. You know, people are paying 20x ARR and you’d be lucky if you got 5x to 7x ARR in a good market. That has changed. You’re going to see that in AI also. People are going to want to say: “Show me revenue growth. Show me profitability growth.”
Related reading:
Illustration: Dom Guzman
Stay up to date with recent funding rounds, acquisitions, and more with the
Crunchbase Daily.
