Mapping the Frontier Stack: Deep-Tech Sectors Shaping 2025 and Beyond

At a glance, venture capital might seem to be in its own world: hyping trends, chasing valuations, and minting unicorns. But beneath the hype, something interesting is happening: deep tech is quietly growing. While the noise of SaaS rounds and chatbots dominates the headlines, a new layer of venture activity is emerging at the intersection of space, defense, data, AI, and energy.

These sectors are no longer isolated verticals, they’re merging into something more integrated, complex, and critical. AI chips in orbit. Microgrids powering drone factories. Satellites feeding real-time imagery into machine learning models. What used to be science fiction is now the frontier of early-stage investing. And it’s unfolding faster than most people realize.

This blog is a guide to that convergence. It’s an attempt to map the market forces shaping what comes next, and to make the case that founders and investors who understand these intersections will be the ones to build the infrastructure of the future.

Space Tech: The Next Gold Rush Off World

Space has always captivated human dreams. Now it’s capturing venture capital. Today’s tailwinds are clear: launch costs have plummeted (some rockets deliver payloads to orbit for around $1,500/kg), demand for satellite services is exploding, and national priorities (from climate monitoring to security) are pumping money into orbit. SpaceTech VC is booming, deals and dollar amounts have more than doubled over four years. In 2025 alone, VCs poured an estimated $3.3 billion in. Beyond SpaceX and Starlink, capital is broadly flowing into smaller players, meaning space is no longer just 'SpaceX and friends.'

Launchers & Manufacturing: Rocket innovation isn’t over. Startups like Relativity Space (3D-printed rockets) and Stoke Space (fully reusable second stages) are chasing cheaper, higher-cadence delivery. Not to compete with SpaceX directly, but to carve out specialized lanes. Meanwhile, contract manufacturing and smallsat assembly are drawing capital as national security re-emphasizes domestic supply chains. The winners here will either vertically integrate or become the “Foxconn” for satellites.

In-Orbit Infrastructure: This is where it gets interesting. Services like in-orbit propulsion, satellite refueling, debris removal, and even compute are maturing. Starcloud’s orbital GPU cluster is one example. Early, yes, but it signals a broader thesis: move more workflows to orbit, from AI inference to edge processing. As orbital assets proliferate, on-orbit maintenance and services may become the DevOps layer for space.

Sensing & Software: Earth observation isn’t new, but what’s changed is who uses it and how. Investors now look for dual-use applications, climate risk, defense intelligence, or agriculture. Where satellite data powers real-time decision making. Startups are gaining traction by embedding analytics into customer workflows, not selling pixels. Expect to see more vertical AI layers built on top of orbital data streams.

Founders still face steep headwinds, as space hardware has unforgiving failure modes. Even with 95% launch success rates, every satellite risks radiation, thermal shocks, or orbital debris. Building an orbital cloud means conquering new physics (powering GPUs in space, as Starcloud’s first mission proved) and regulatory drag from spectrum and export rules. Scaling one orbiting server to dozens will strain budgets and engineering discipline. Timing is critical. Too soon, and customers aren’t ready; too late, and giants like AWS or SpaceX may eclipse you.

Looking ahead, expect consolidation and vertical integration. Companies that today do one piece (launch, manufacture, or data analytics) may bundle up or be acquired. The market map is still forming, perhaps the big datacenter or energy players will expand skyward (Jeff Bezos has even entertained space-based solar farms). In short, SpaceTech today has diverse VC backing and multiple niches rising, from propulsion to payloads, but it’s still an evolving, high-risk ecosystem with an enormous runway.

Defense Tech: The Next Industrial Base

Defense tech is leading the next wave of national security. Geopolitics and budgets are fueling this push. Global military spending hit a record $2.443 trillion, and VC in defense startups surged, roughly $3 billion flowed into defense tech in 2024, a 33% jump over 2023. Big name investors have taken notice, everything from Founders Fund writing a $1B check for Anduril, valuing it at $30.5B, to Andreessen Horowitz carving out $800M for defense bets, and In-Q-Tel (CIA’s fund) committing serious capital. VCs are shifting from sideline skeptics to active participants.

Warfighting AI & Autonomy: Artificial intelligence is the new frontier in defense. Startups using AI for intelligence analysis, autonomous drones, or battlefield automation are in hot demand. For context, the military AI sector could grow from $13B in 2024 to over $61B by 2034. Autonomous drones and vehicle systems, battle tested in recent conflicts, represent another mega trend. For example, the global military drone market, $11B now, is projected to reach $21B by 2034. Companies like Shield AI (drones) and Anduril (sentinel towers, Lattice AI platforms) exemplify this wave.

Cyber & Networks: Defense grade cybersecurity is booming. Armed conflicts have shown cyberwar’s potency, and budgets reflect it. The defense cyber market was $16.5B in 2023, forecast to hit $63B by 2032 (16% CAGR). Startups protecting critical infrastructure or building hardened networks like SentinelOne and Palantir are extensions of this demand. Encryption, anti-drone defenses, and secure comms now draw venture interest.

Integrated Systems & Manufacturing: Governments are pushing to rebuild the defense industrial base. New U.S. policy ‘Arsenal of Freedom’ scraps slow procurement in favor of 85% done solutions. Dual-use tech (initially civilian products proven in-field) can win contracts faster. We see funds emphasizing supply-chain resilience: additive manufacturing (hypersonic hardware 3D printing), resilient microelectronics, and robotics for ship and plane maintenance. In Europe, a new security push has drawn record funding. The EU announced an €800B defense plan and €576M of VC funding into EU defense tech in 2024. It’s truly global. India and Israel are also ramping innovative startups in this space.

Defense founders must navigate classic constraints due to government procurement being notoriously slow and opaque. Startups need patience, and often a pivot to commercial markets, before DOD contracts pay off. Regulations around classified data, export controls (ITAR), and ethical scrutiny of weapons tech add layers of compliance. There’s also fierce competition. Big primes like Lockheed and Boeing now have venture arms and M&A budgets. Even so, exit opportunities are real. 2024 saw $62B in defense M&A (up 82% YoY), and strategic acquirers have paid huge premiums, such as AeroVironment’s $4.1B buyout of BlueHalo at 60x revenue.

Looking forward, expect mergers and strategic partnerships. Large primes will acquire nimble AI/cyber firms to stay cutting-edge, and allied procurement reforms (easier foreign military sales paths) will let U.S. startups tap global markets. Dual-use companies (like Space startups or AI firms with defense applications) stand to benefit most. Defense Tech is maturing fast into a standalone asset class. The moat is technical difficulty, but once proven, startups can scale on both government and commercial demand.

Data Infrastructure: The Nervous System of the Frontier

Data is the nervous system of modern deep tech. Without it, satellites are blind, autonomous systems are deaf, and AI is effectively paralyzed. The problem isn’t generating data, we’re drowning in it. The problem is making it move.

Every year, we launch more imaging, communication, and sensor satellites into orbit, each spitting out torrents of raw input. Satellite imagery, radar returns, IoT signals, drone video feeds, billions of data points are generated daily. Sending all that information down to Earth remains a bottleneck. Radio downlinks are expensive, congested, and slow. That’s why winning the "new space race" increasingly looks like winning the data race.

The next generation infrastructure stack is rising to meet this challenge. Space-based networks like Starlink and OneWeb are no longer experiments. They’ve become critical. In Ukraine, Starlink’s network proved vital for restoring battlefield communications, earning the nickname 'the first commercial space war.' SpaceX has now launched Starshield, a defense-focused constellation with encrypted links and missile warning sensors, tailored for SDA and Space Force use. These networks form the backbone of modern defense communications.

But the real shift is what’s happening underneath those satellites. Cloud providers have quietly rolled out global ground station-as-a-service offerings. AWS Ground Station, Azure Orbital, and others let satellite operators push imagery or telemetry directly into cloud compute regions in near real time. SAR imagery streams from orbit to the cloud, where it’s processed within seconds, delivering critical flood maps to emergency responders without ever touching a local server. Companies like Capella Space are already productizing this, a space-based SaaS for Earth observation.

That’s the horizontal expansion of data infrastructure. The vertical one is DataOps, the evolving stack of tools that automate, monitor, and manage the pipelines that feed AI systems. Platforms like Snowflake, Databricks, and Airbyte have turned data movement into a software discipline. 

Edge processing is also reshaping the equation. Satellites and UAVs now run inference at the point of capture, sending back only the insights. In defense use cases, this logic matters because decisions need to happen mid-flight, not back on a server farm. Startups are building hardware accelerators and inference stacks that make this possible. Think of a drone identifying tanks or terrain anomalies mid-flight, or a LEO satellite flagging anomalies in missile defense zones before downlink.

Startups are racing to enable this future, yet headwinds remain. Interoperability between systems is a mess: different payloads, data formats, and spectrum bands make seamless integration difficult. Ground infrastructure still lags with many remote stations lacking fiber, and building connectivity in conflict zones or orbital theaters is no easy feat. 

But these are solvable problems. The tailwinds are massive: $100B+ of forecasted investment in data infrastructure globally, plus strategic demand from defense agencies and commercial cloud providers alike. Sovereign clouds, AI-specialized data lakes, and global mesh architectures are no longer sci-fi, they’re the scaffolding for everything else in the frontier stack.

AI Infrastructure: Powering the Big Models

AI infrastructure is one of the most overfunded and overhyped sectors in tech today. But peel back the buzz, and you’ll find a real shift: the building blocks of AI, chips, compute, and models are increasingly colliding with space and defense.

Let’s start with the basics. Building AI models is no longer just about smart algorithms, it’s about having access to compute, fast networks, and scalable tools. And the demand is staggering: AI infrastructure companies raised over $6.8 billion in 2024, nearly doubling from the year before. Most of that is going to a mix of high-end GPU cloud providers (like CoreWeave), observability tools, and model deployment platforms.

So how does that tie back to space and defense?

First, compute power itself is becoming strategic. As training and deploying AI models gets more expensive, space is emerging as a surprising frontier for compute infrastructure. Take Starcloud, they launched the first NVIDIA H100 GPU into orbit this year. Why? Because space offers three critical advantages: always-on solar power, no land or cooling constraints, and a massive opportunity to offload workloads away from Earth’s strained data centers. It’s early, but these orbital clouds could one day power real-time analytics for satellites, drones, or intelligence operations.

Meanwhile, defense is embracing AI at every level from battlefield decision tools to synthetic training environments. But here’s the twist, these systems aren’t just built by traditional defense contractors. Increasingly, venture-backed AI infrastructure startups are powering them. Anduril, for instance, builds AI-native defense hardware, but it relies on robust data pipelines and inference layers to make its systems responsive. Palantir, too, now pitches itself as a full-stack AI platform for government clients.

Even at the hardware level, national security concerns are driving AI strategy. Chips used in military drones or signal processing systems need to be fast, secure, and export-compliant. This is fueling interest in U.S.-based chip startups and making AI infrastructure part of broader tech sovereignty debates.

Another area to watch is model deployment. The shift from building to using AI models means companies, including those in aerospace and defense, need tools to host, fine-tune, and monitor models across complex environments. Startups like Anyscale offer serverless AI platforms that could eventually run edge inference on everything from satellites to submarines. These aren’t just backend tools, they’re enablers of real-time autonomy.

There are, however, headwinds. Space-based compute is unproven and extremely expensive. Government procurement cycles are slow. And the big clouds (AWS, Azure) will continue bundling AI into their own platforms. But the upside remains huge. We’re watching a new playbook emerge: startups that combine deep AI infrastructure chops with a thesis around dual-use, serving both commercial and defense clients, are raising capital and landing contracts at record pace.

Energy Tech: Powering New Frontiers

The energy sector is no longer just about wind turbines and solar farms: it’s actively infiltrating space and defense. In orbit, startups and researchers are exploring solar-powered satellites and data centers that tap virtually unlimited sunlight. On Earth, the military is deploying renewables and microgrids to keep bases online during outages and combat missions. And intriguingly, dual-use technologies like advanced microreactors and thermal batteries promise to serve both civilian grids and defense needs. Below we highlight key trends, real-world examples, and funding signals at the intersection of energy, space, and defense:

Space-based Solar & Compute: Orbital data centers and satellite power systems use dedicated solar arrays to drive AI and communications in space. For example, Starcloud launched the first H100 GPU into space on a solar-powered satellite, arguing that in space, you get almost unlimited, low-cost renewable energy for computing. Companies like Star Catcher are even beaming concentrated sunlight to other satellites’ solar panels, effectively creating an orbital power grid that can boost a satellite’s output by up to 10×.

Portable Nuclear & Dual-Use Tech: Perhaps most striking is the push for mobile nuclear power. The Pentagon’s Project Pele is a transportable 1.5 MW microreactor designed to run for 3 years without refuel, delivering resilient and reliable electricity to remote military sites. Similar systems are under development commercially. Radiant Nuclear (founded 2020) is building 1 MW TRISO-fuel microreactors that can drive microgrids and replace diesel generators at bases or disaster zones. Radiant just raised $165M (Series C) to mass-produce these transportable reactors, explicitly targeting military installations and remote power. This reflects a broader nuclear renaissance. On the storage front, innovations like solar-thermal batteries are taking off. Portal Space Systems raised $17.5M to build a solar-thermal spacecraft ‘Supernova’ that heats a thermal battery for high-thrust maneuvers. It delivers nuclear thermal propulsion performance without an actual reactor, a perfect example of nuclear-grade capability adapted for space and defense use.

Nuclear Fission & Fusion: Advanced fission and fusion technologies promise longer-term transformations in power and propulsion. On the fusion side, well-funded startups are moving toward reactor demos. Helion Energy, backed by Sam Altman and SoftBank, is building a magneto-inertial fusion machine. In July 2025 Helion began construction of its first plant (Orion) in Washington State, aiming to deliver the first fusion electricity by 2028. This plant will supply 50 MW to Microsoft under a recent power purchase deal, making it the first planned fusion power plant built for the grid. Another leader is Commonwealth Fusion Systems. In August 2025 CFS raised an additional $863 million (Series B2), on top of $1.8B prior, bringing its total to nearly $3 billion. Investors include Google, Nvidia, and Breakthrough Energy Ventures. These firms reflect a global fusion push, hoping to create carbon-free, compact nuclear plants that could one day power space stations, bases, or ships with virtually limitless clean energy.

Collectively, these examples show a convergence of energy innovation with space and defense. VCs and agencies are pouring capital into these overlaps with record rounds for nuclear microreactors, space-power startups, and military microgrids signaling strong belief in energy as a strategic enabler. The EnergyTech slice of the market map is rapidly evolving into a critical pillar for both space infrastructure and battlefield readiness.

The Long View: Opportunities in the Unknown

As we map these sectors, a few themes emerge. First, the interplay between layers is growing: satellites need AI analytics, defense depends on space and data, AI infrastructure demands clean energy. Think of the Frontier Stack not as isolated silos but as layers of a single new ecosystem. Second, timing is everything. Each of these fields is emerging from dormancy (Space and Defense after 20-year lulls, Data/AI after decades of slow development, Energy under climate urgency), so founders ride a narrow window.

For investors and founders, this is a time of boundless curiosity. We’ve invited you in to decode these frontiers. The story of 2025’s Frontier Stack is still being written, with satellites soaring above, algorithms crunching below, and new forms of energy animating our grids. The limits of what’s possible are still being redefined. Keep your eyes upward and forward, the next breakthrough could be orbiting right above us.