Hardware is no longer a 4-letter word in venture capital
What most people forget is that venture capital was born in hardware.
The industry's creation myth is not a story about code, but one of silicon, solder, and circuit boards.
Arthur Rock, widely regarded as one of the great progenitors of modern venture capital, made his name by backing Fairchild Semiconductor in 1957 and then funding Intel in 1968. He later backed Apple. His legendary career was built on physical things: chips you could hold, computers you could touch.
Then, gradually, the industry forgot where it came from.
Over the course of the 1990s and 2000s, venture capital developed a simple truth: software good, hardware bad.
That consensus is cracking. Physical AI has emerged as one of the most successful and exciting industries for founders and VCs alike.
And in a sense, venture capital is not discovering something new, but remembering something old.
The end of SaaS and the rise of vibe coding
Something strange happened on the way to infinite software margins.
The SaaS landscape became a sea of near-identical products differentiated by little more than logo color and pricing page layout.
SaaS multiples have compressed dramatically from their 2021 peaks. The BVP Nasdaq Emerging Cloud Index fell roughly 60% from its highs and has struggled to recover. Public SaaS companies that once commanded 20–30x revenue multiples now trade at mid-single-digit multiples.
Then came the final accelerant: AI-assisted coding.
When tools like Cursor, Replit Agent, GitHub Copilot, and a growing constellation of AI coding assistants made it possible for a single developer (or even a non-developer) to ship functional software in hours rather than months, the implications for software investing became difficult to ignore.
If a motivated founder can recreate the core functionality of a vertical SaaS product over a weekend, the defensibility of that product collapses.
Founders are now competing with the possibility that a frontier lab will ship their entire product as a Tuesday afternoon demo.
Hardware offers a different game entirely.
OpenAI cannot 3D-print a turbine blade. Anthropic cannot prompt its way to a novel battery chemistry. Google DeepMind can accelerate materials discovery, but someone still has to build the factory, solve the thermal management, negotiate the supply chain, and ship the physical product.
For founders seeking true differentiation, the physical world is the last reliable frontier.
The great rotation into hardware and deep tech
The deeper water, increasingly, is in the physical world.
The shift is visible across nearly every major fund, and some firms saw this coming earlier than most.
N47 has been quietly building a portfolio thesis around the conviction that the most enduring technology companies would be those that fused intelligent software with purpose-built physical systems.
The results speak for themselves.
- Verkada, reimagines physical security by combining cloud-managed cameras and access control hardware with AI-powered analytics, growing into one of the most valuable private companies in enterprise security.
- Skydio, builds autonomous drones powered by on-board AI. These aircrafts that can navigate complex environments without a human pilot, serving customers across defense, infrastructure inspection, and public safety.
- And Tractian, deploys proprietary industrial sensors paired with machine learning models to predict equipment failures on factory floors before they happen, transforming maintenance from a reactive cost center into a strategic advantage.
Each of these companies shares a common architecture: differentiated hardware that generates proprietary data, feeding software and AI that competitors cannot replicate without building the same physical stack from scratch.
Hardware that creates the flywheel and the moat.
If you are a technical founder with deep domain expertise in a physical-world problem, you have never had better access to venture capital than you do right now.
Why now? What changed?
Several forces converged to make hardware investable in ways it simply was not a decade ago.
.avif)
AI needs physical infrastructure
The most valuable AI companies are companies that control scarce physical resources.
NVIDIA's market capitalization surpassed $3 trillion because of its chips. The explosion in AI demand has created massive markets for data center hardware, custom silicon, advanced cooling systems, and high-bandwidth networking equipment. Companies like Cerebras, Groq, and d-Matrix are raising hundreds of millions to build specialized AI inference chips.
The AI revolution's bottleneck is not software. It is hardware. And every bottleneck is a business opportunity.
Manufacturing got smarter
The rise of contract manufacturing platforms, 3D printing for rapid prototyping, and software-defined factories has dramatically reduced the capital required to bring a hardware product to market.
Companies like Formlabs, Desktop Metal, and Fictiv have built the infrastructure layer that makes hardware startups more capital-efficient than ever before.
The practical barriers to starting a hardware company have fallen just as the strategic reasons to start one have strengthened.
Geopolitics rewired supply chains
The CHIPS Act in the United States, the European Chips Act, and similar programs in Japan and South Korea have poured over $100 billion in public subsidies into semiconductor manufacturing and advanced hardware.
Governments are now co-investors in deep tech, de-risking private capital in ways that would have been unimaginable during the pure-software era.
Climate and energy created trillion-dollar markets
The energy transition is fundamentally a hardware problem.
You cannot decarbonize the grid with a SaaS dashboard. Companies like Form Energy (iron-air batteries for grid storage), Commonwealth Fusion Systems (compact fusion reactors), and Sila Nanotechnologies (next-generation battery materials) have raised billions by tackling challenges that require breakthroughs in materials science and engineering, not better user interfaces.
Defense and space opened up
The Department of Defense has aggressively courted startup suppliers through initiatives such as the Defense Innovation Unit and SBIR programs, while the commercial space market has exploded.
Anduril, now valued at over $14 billion, builds autonomous defense systems that rely on significant hardware components. Relativity Space is 3D-printing rockets. Shield AI is building autonomous fighter aircraft.
These are category-defining companies attracting the largest rounds in venture capital.
The new hardware giants
The shift from theory to practice is best illustrated by the companies raising capital today.
SpaceX remains the archetype. A hardware company that has generated more venture returns than most software portfolios combined. Its valuation, reportedly north of $350 billion in secondary markets, makes it the most valuable private company in the world.
Anduril Industries raised $1.5 billion at a $14 billion valuation in 2024, building everything from autonomous drones to modular command-and-control hardware for the U.S. military.
Figure AI, the humanoid robotics company, raised $675 million in a round valuing it at $2.6 billion, backed by Bezos Expeditions, Microsoft, NVIDIA, and OpenAI.
Hadrian, building automated factories for precision aerospace parts, raised $260 million from Lux Capital, Andreessen Horowitz, and others.
Each of these companies would have been uninvestable by mainstream venture firms ten years ago. Today they are portfolio centerpieces.
Shooting the moon
The capital intensity that terrifies most investors — the hundreds of millions burned on R&D, the agonizing years before first revenue, the brutal cash requirements of building physical production at scale — is not a bug. It is the moat.
Every dollar burned is a dollar your competitor also has to burn, and most of them cannot.
The race does not go to the swift. It goes to the last company standing.
Consider the economics. A SaaS startup that raises $10 million can be replicated by another team with $10 million and a decent engineer. The barriers are low, the clones are endless, and defensibility is a polite fiction maintained by switching costs and customer inertia.
But a company that has spent $500 million developing a novel semiconductor fabrication process, or a new battery chemistry, or a reusable rocket — that company has built something its competitors cannot replicate without spending $500 million of their own.
And by the time they do, the incumbent is another generation ahead.
This is why hardware markets tend to consolidate into oligopolies or outright monopolies in ways that software markets rarely do.
TSMC does not have fifteen competitors. SpaceX does not face a new launch provider every quarter. Intel dominated for decades not because its software was better but because its fabs represented tens of billions in sunk capital that no one else could match.
The exhaustion of extraordinary capital is the strategy.
For founders, the difficulty of building a physical AI startup is precisely what makes it defensible — not just against other startups, but against the frontier AI labs themselves.
OpenAI can ship a new software feature overnight; it cannot ship a new chip architecture, a new manufacturing process, or a new propulsion system overnight.
The harder the problem, the wider the moat, and the less likely it is that a foundation model renders your life's work obsolete with a single update.
Founders who choose hardware are building more than companies. They’re building things that cannot be prompt-engineered out of existence.
Conclusion
For deep-tech founders, the calculus has shifted in ways that would have been unthinkable five years ago.
The venture capital industry wants to fund you.
The frontier AI labs cannot compete with you.
The markets you are entering reward exactly the kind of deep, patient, technically rigorous work that the software era undervalued. Start the hardware company. The ecosystem is ready for you in a way it has not been in a generation.
