The Real World Wants Its Money Back

Nuclear reactors built in shipyards. Pilotless cargo planes. $1.3 billion to back machines that move things. The most interesting VC money this week wasn't chasing another LLM wrapper. It was funding stuff that has to actually work in physics.

Blue Energy raised $380 million to build nuclear power plants the way ships get built: in a yard, prefabricated, then barged to site. Led by VXI Capital, the round targets a 1.5 GW Texas facility scheduled to break ground later this year. The explicit customers? AI data centers that need firm, dispatchable baseload power and don't want to own the upside or the risk.

The nuclear pitch has been around forever, but the financing structure is actually new here. Blue Energy got the NRC to approve a gas-first, nuclear-later phased approach, meaning the plant generates revenue for years before the nuclear island even comes online. That's the first time private capital markets can finance a nuclear project without government loan guarantees. If it works, it rewrites how the industry gets built.

Reliable Robotics closed $160 million led by Nimble Partners, pushing its valuation to nearly $1 billion, as it assembles what CEO Robert Rose calls a "mountain of evidence" for the FAA to certify uncrewed commercial cargo flights. Eclipse, Lightspeed, Coatue, and RTX Ventures participated. The company has 200+ system commitments from commercial and military customers and plans to run its first paid autonomous cargo route this summer, Albuquerque to Santa Fe, before the full FAA type certificate expected around 2028-29.

The FAA certification play is worth watching. Reliable retrofits existing Cessna 208 Caravans that already fly FedEx cargo, rather than designing a new airframe from scratch. Split the regulatory problem in half. That's the wedge the rest of the drone industry hasn't cracked, and why Boeing (via AE Ventures) and RTX both wrote checks.

Crunchbase reported 37 companies joined its Unicorn Board in March — the highest monthly count in close to four years. Read that in the context of Q1 2026's record-breaking $300 billion in global venture investment. The headline is obvious: the market is hot. The less obvious part is that the unicorn creation rate is accelerating even as the IPO market stays largely closed. That's a growing backlog of companies with a billion-dollar price tag and no clear exit runway. Private market enthusiasm and public market skepticism are pointing in opposite directions. One of them is wrong.

The Physical World Is Finally Getting Its VC Moment

Earlier this month, Eclipse closed $1.3 billion across two funds with a simple thesis: the next era of AI won't live in browsers. It'll move freight, build buildings, fly planes, and run factories. The firm split the capital between a $720 million early-stage vehicle (Fund VI) and a $591 million growth fund (Early Growth Fund III), pushing its total AUM to roughly $10 billion.

Eclipse has been making this bet for a decade. Its existing portfolio includes Wayve (autonomous vehicle software), Redwood Materials (battery recycling), Arc (electric boats), Bedrock Robotics (self-driving construction vehicles), and Mind Robotics (industrial automation). What's changed is that institutional LPs are now buying the thesis in size.

The timing tracks. Four mega-rounds absorbed 65% of all global Q1 venture capital. But outside OpenAI, Anthropic, xAI, and Waymo, there's been a quiet reallocation happening at the margins. Infrastructure deals are up. Physical AI is getting real check sizes. And the underlying logic is sound: software AI has a commoditization problem. Every model improvement chips away at what the application layer built. Hardware AI has the opposite problem. Once you've automated a physical workflow, switching costs are enormous. The machine doesn't just run the software; it is the software.

Eclipse's hybrid model is worth unpacking. The firm doesn't just write checks; it also incubates companies internally. Partner Jiten Behl describes the approach as building a web of portfolio companies that share infrastructure, customers, and data across sectors. Defense clients share demand signals with logistics clients. Robotics hardware serves both construction and manufacturing. The data that flows between them trains better models. Eclipse is betting that the moat isn't in any single product but in the network.

The risk is timing. Physical AI companies need capital to cross from prototype to production, and that gap is notoriously expensive and slow. Eclipse's growth fund is designed to bridge it, but hardware iteration cycles are weeks or months, not days. And the field is getting crowded fast. Kleiner Perkins closed a $3.5 billion AI fund in March. Andreessen Horowitz has been investing in defense and robotics aggressively. First-mover advantage matters in hard tech, but only if you can survive long enough to reach production.

What the Eclipse close confirms is something Blue Energy and Reliable Robotics also signal this week: the infrastructure thesis has expanded beyond software. VCs are no longer just betting on the picks and shovels that support AI models. They're betting on the machines that AI will eventually run.

Blue Energy raised $380 million in equity and debt led by VXI Capital to build grid-scale nuclear power plants using shipyard manufacturing methods. A three-year-old Maryland startup, Blue Energy does not design novel reactors; it redesigns how they get built. Targeting a 1.5 GW Texas facility for AI data center power purchase agreements, the company's phased gas-to-nuclear approach is the first privately financeable nuclear structure in U.S. history.

Omni raised $120 million Series C led by ICONIQ at a $1.5 billion valuation, more than double its $650 million valuation from March 2025. The San Francisco company, founded by former Looker executives, builds a semantic layer for enterprise analytics that functions as a governed translation layer between raw data and AI agents. Revenue quadrupled year-over-year; the company reached profitability last month. Theory Ventures, First Round Capital, Redpoint, and GV also participated.

AcuityMD raised $80 million Series C led by StepStone Group at a $955 million valuation. The Boston company aggregates FDA filings, claims databases, and market signals into a proprietary knowledge graph for MedTech commercial teams, then layers AI agents on top. Sixteen of the top 20 MedTech companies are customers. Benchmark, Redpoint, and ICONIQ participated.

Tava Health raised $40 million Series C. The Salt Lake City company operates a tech-enabled mental health platform connecting employees to therapists through employer benefit programs. The raise reflects continued enterprise appetite for behavioral health benefits infrastructure amid rising workforce mental health costs.

Courier Health raised $50 million Series B to expand its patient experience management platform, which helps pharmaceutical companies track patients through treatment cycles and address drop-off. New York-based, the company operates at the intersection of pharma commercial strategy and patient services — not clinical care, which is how it avoids the regulatory landmines that slow down most health tech startups.

Word on the street is that three late-stage AI infrastructure companies are shopping for $100M+ rounds, all at flat or down valuations from peak 2025 levels. If true, that's a meaningful shift from last quarter's valuation exuberance, particularly for companies burning $10M+ monthly with limited revenue visibility.

OpenAI is reportedly finalizing its $100B funding round with Amazon potentially contributing $50B through a combination of cash and AWS credits. The circular financing structure (invest money that gets spent back on services) is becoming standard practice, though Nvidia's pullback suggests limits to this approach.

On the exits front, two enterprise security companies in the $500M+ valuation range are exploring merger discussions. Both are struggling to cross $50M ARR despite strong investor pedigrees. A combination might create the scale needed for a strategic exit, but board politics and founder egos could derail talks.

Finally, whispers of SpaceX IPO timing accelerating following the xAI acquisition. Sources suggest a mid-2026 public debut at the $1.25T combined valuation, though regulatory reviews of the merger could delay the timeline. If it proceeds, it would be the largest IPO in market history.