Record money for the hard, physical layer — starting with space

Space startups raised a record $20.8B in Q3 2026 across 216 deals, per Seraphim — nearly triple Q2’s $7.4B.

Launch and space-access businesses took $16.4B (120 deals); the biggest rounds were Hadrian ($1.37B), Stoke Space ($1B), and K2 Space ($500M).

France’s Infinite Orbits raised €52M for in-orbit satellite servicing, with demand from the French Space Command, Eutelsat, and SES.

A launch shortage — too many satellites, too few rockets — is drawing capital into the whole supply chain, from manufacturing to in-orbit servicing.

Defense and sovereign demand is now a primary driver, as governments prioritize domestic space capability.

Private capital is surging even as listed space stocks lag — most public names remain loss-making.

The action is in infrastructure and servicing (Infinite Orbits, Impulse, Hadrian), not just satellite operators.

European sovereignty plays — like Infinite Orbits — are a distinct, fast-growing lane.

Risk: a record quarter invites froth; Seraphim itself flags that funding may moderate, and public-market weakness is a caution.

Eka raised a $150M Series A at a $1.45B valuation (co-led by a16z and F-Prime); total funding ~$232M.

Its force-first “vision-force-action” model trains in physics simulation, skipping human-demonstration data entirely.

The system runs on a sub-$1,000 gripper and is pitched to manufacturing, logistics, and warehouses.

Manipulation — reliably grasping unfamiliar objects — is the stubborn barrier between robot demos and real work.

Eka’s simulation-first bet contrasts with the data-hungry imitation-learning camp (Physical Intelligence, Skild).

A Series A unicorn priced on a manipulation model shows capital is backing the robot “brain,” not whole humanoids.

Eka (private) offers exposure to the physical-AI “manipulation layer” — a picks-and-shovels angle on robotics.

Ordinary investors find robotics hard to access — most of the value sits in private rounds like this one.

Risk: it’s early, priced richly on a demo, and must prove the model holds up on live factory lines.

Coherence Neuro raised a $45M seed co-led by the OpenAI Startup Fund and Founders Fund.

Its implant aims to detect brain-tumor signals and, eventually, treat them with electrical stimulation.

It has tested recording briefly in three surgery patients and will fund a long-term glioblastoma trial plus an FDA breakthrough-device bid.

It pushes BCIs into oncology — a far less established use than restoring communication or movement.

It’s part of a wave of AI companies betting on brain data (OpenAI also backs Merge Labs).

A closed-loop “sense-and-treat” implant, if it works, would be a genuinely new class of device.

Coherence (private) is a high-risk, high-upside bet on a new BCI category, with marquee backers.

The raise rides broad BCI momentum (Neuralink $650M, Precision $250M).

Risk: human data so far is three patients for ~30 minutes — the oncology hypothesis is early and unproven.

Hephaestus exited stealth with a $17.2M raise led by CiRi Ventures.

It makes rare-earth-free HTS magnets, heading to manufacturing in 2027.

It’s also developing a negative-triangularity reactor targeting net energy gain (Q>1) by 2030.

HTS magnets are the enabling component for compact fusion — and a supply bottleneck for the whole field.

Removing rare-earth dependence cuts a critical-minerals risk that shadows the clean-energy supply chain.

It’s both a picks-and-shovels supplier to other fusion developers and a reactor builder in its own right.

Hephaestus (private, European) offers component-level exposure to fusion, with nearer-term magnet revenue.

Selling magnets to the whole sector spreads risk beyond a single reactor bet.

Risk: early-stage and capital-intensive, and its 2030 net-energy-gain target is ambitious and unproven.