Building at the Foundation: The Celesta Capital Deep Tech Thesis
Building at the Foundation: The Celesta Capital Deep Tech Thesis
Venture capital spent the past decade optimizing for speed, lower up‑front capital, faster product launches, and shorter hold times. In the process, much of the industry neglected the very foundation that makes enduring innovation possible – deep technology.
At Celesta Capital, we did not follow the industry’s drift toward the application layer. Instead, we doubled down on the infrastructure, targeting deep tech since our first fund in 2013. Today, we are proud to be veterans in this space as we witness AI catalyzing a profound hardware renaissance.
On behalf of my fellow General Partners Michael Marks, Matthew Marsh, Sriram Viswanathan, and Lip-Bu Tan, I wanted to share the core thesis through which we’ve built Celesta, why deep tech remains a source of durable advantage, and where we see compelling opportunity across the vibrant landscape of foundational technologies.
Venture capital spent the past decade optimizing for speed, lower up‑front capital, faster product launches, and shorter hold times. In the process, much of the industry neglected the very foundation that makes enduring innovation possible – deep technology.
At Celesta Capital, we did not follow the industry’s drift toward the application layer. Instead, we doubled down on the infrastructure, targeting deep tech since our first fund in 2013. Today, we are proud to be veterans in this space as we witness AI catalyzing a profound hardware renaissance.
On behalf of my fellow General Partners Michael Marks, Matthew Marsh, Sriram Viswanathan, and Lip-Bu Tan, I wanted to share the core thesis through which we’ve built Celesta, why deep tech remains a source of durable advantage, and where we see compelling opportunity across the vibrant landscape of foundational technologies.
Our Thesis: Infrastructure Unlocks the Next Wave
Celesta’s core thesis is clear and straightforward by design: breakthroughs in deep tech are the primary driver for every major, economically impactful wave of technological advancement.
The personal computing and mobile revolutions were driven by breakthroughs in processors, memory, battery technologies, power, networking, and others. The mass manufacturing, miniaturization, and proliferation of those devices in turn enabled the app- and SaaS-led value creation of the past decade.
At each instance, for the world to sustainably re-platform around a new major technology shift, the underlying infrastructure must be solid. To power the next century of innovation that we hope to see within AI, autonomy, or renewable energy, we must advance the fundamental technologies – compute, sensing, networking, and materials – that will allow them to efficiently interact at scale with the physical world.
Defining Deep Tech: Utility Over Spectacle
Celesta defines deep tech as businesses built on strong IP stemming from fundamental scientific or engineering advancements.
We avoid moonshots and speculative science. Instead we seek out proven technical breakthroughs that solve well‑understood problems backed by market urgency. We don’t invest in "what if"; we invest in the essential.
We spend substantial time on technical due diligence and IP in order to deeply understand technology. In more than 100 Celesta investments to date, we haven’t backed a company where the core technology later proved non-functional.
From Contrarian to Frontier: Why Deep Tech Wins Now
Silicon Valley was built on the deep tech "silicon" of companies including Fairchild and Intel. For a time, the industry focused predominately on pure software, lured by the high margins and low friction. But we are witnessing a correction.
The AI revolution is not just a software trend; it is catalyzing the most profound infrastructure rebuild since the microprocessor. Fulfilling the vision of AI will not come through simply adding it to existing systems – AI is compelling a complete re-architecture of the technology landscape itself. Every product now needs to be reimagined through the intelligence layer, which demands new compute paradigms, new requirements on power and data efficiency, and new modalities of advanced sensing and intelligence.
These technologies have never been more essential – and the market opportunities surrounding them are more attractive than any time in recent memory. This moment represents one of the most compelling times to be a deep tech investor:
- Margin Shift: As software margins compress under the weight of rising AI "compute taxes," skyrocketing demand is driving historic expansion of gross margins for hardware.
- Defensibility: In an era where AI can generate code in seconds, the only true "moat" is proprietary IP rooted in physics and hard engineering.
Where We Invest: Strategic Points of Transformation
We invest in infrastructure technologies in four key areas, targeting the entry points where we can exert the most influence on the global technology stack.
1. Semiconductors and Intelligent Systems: We invest in the hardware building blocks – compute, memory, and power delivery – that define the limits of modern system performance. We also invest in hardware-driven intelligent systems – edge AI, sensing, robotics, and autonomy – where advanced hardware is deployed as intelligence in the real world.Example portfolio companies include AuraSemi, Credo, DustPhotonics, Eliyan, PowerLattice, SambaNova Systems, Stathera, Tensordyne, UpScale AI, and VeriSilicon.
2. Next-Generation Infrastructure Software: We back software that exposes the underlying capabilities of computing infrastructure and makes it operable at scale. This includes hardware-aware infrastructure software (orchestration and control planes, resource optimization across CPU/GPU clusters, data movement and management, observability/reliability tooling, and security) as well as application-layer infrastructure that operationalizes AI inside real enterprise workflows. Example portfolio companies include Exostellar, H2O.ai, Normal Computing, ParkourSC, Platform9, and Slate.ai.
3. Bio-Convergence: We invest where engineered technologies converge with biology to deliver more accessible, lower-cost care. That includes leveraging advances in compute, sensing, automation, and data science in diagnostics, imaging, care delivery, and drug discovery – areas where deep tech can speed time to market, reduce cost per test, and improve outcomes by moving care closer to the point of need. We see biology as the next great engineering discipline. By applying deep tech – sensors, automation, and data science – to diagnostics and care, we move from speculative "biotech" to predictable, scalable medical infrastructure. Example Celesta portfolio companies include 1Cell.ai, 5C Network, Berkeley Lights, Alveo Technologies, and Prellis Biologics.
4. Industry Transformers: Our first three pillars focus on breakthroughs in the enabling layers. Industry Transformers are where we back companies applying those breakthroughs – in compute, sensing, robotics, materials, automation – to re-architect large legacy markets. These are industries that have historically lagged in tech adoption because of regulation, physical-world complexity, or fragmented economics, but where deep tech can introduce net-new capabilities to meaningfully lift productivity. Example Celesta portfolio companies include Agnikul Cosmos, Atonarp, Boom Supersonic, Brick N’ Bolt, Crescendo.ai, and Robin Hood.
Intellectual Capital: The Edge in Deep Tech Investing
In deep tech, success often hinges on decisions made early. You must know your non-negotiables and choose the right trade-offs early because the cost of a "wrong turn" is an order of magnitude higher than in software. For example, if your edge is time‑to‑market, ship a product and optimize costs later; if you must win on cost, you can’t hope to reduce costs after adoption – you may never get the chance. The wrong early compromises create more ups and downs that burn cash and time.
This is why a seasoned deep tech investor’s value-add is not just financial, but “intellectual capital”. With a team that has collectively started over 40 businesses and holds 600+ patents, we act as a "stabilizer" for our founders to help reduce growth harmonics – the volatile oscillations between technical breakthroughs and market realities – that can stall growth and sink even the best ideas.
We don’t just coach; we provide the architectural oversight necessary to ensure that early trade-offs (like cost vs. performance) don't become fatal flaws later in the lifecycle.
Looking Ahead: What’s Old Is New Again
When Celesta launched in 2013, deep tech sat outside the venture mainstream. We built the firm around a contrarian conviction that the most durable value would be created by companies solving hard technical problems in the enabling layers of the technology stack.
Today that conviction is stronger than ever. AI, autonomy, modern healthcare, and industrial transformation all run into the same reality: they must continue to advance their underlying technologies in order to drive performance improvements at scale. The opportunity is no longer theoretical. It is showing up in market demand, capital flows, and the growing importance of defensible IP.
We have been building with this view for more than a decade. Now, we’re glad to see the rest of the market join us in recognizing this potential as well.
