Canada is building a rocket. Here's how it actually works.

The interesting claim in that keynote isn't about rockets. It's that neither SpaceX nor Rocket Lab is really a rocket company, and that if you copy them as rocket companies you will fail.

The evidence Goel puts up is a revenue breakdown. At SpaceX, Starlink is roughly 65% of revenue and about 70% of Falcon 9 flights carry it. Rocket Lab's space systems division is around 70% of revenue, with launch at 30%, and roughly three-quarters of a two-billion-dollar backlog sitting on the satellite side. Both companies became their own best customer.

The number that makes the point sharpest: an external customer pays about $75M for a Falcon 9. Internally it costs SpaceX about $28M. That gap is the whole business. You cannot get it by buying launch — you get it by owning the vehicle and then flying your own payloads on it.

So NordSpace runs two near-separate companies under one roof — Launch Systems at roughly 80% of resources, Space Systems at 20% — and builds the spaceport itself rather than renting one. The stated reason is sovereignty: the argument is that a sovereign launch capability requires a company that is itself sovereign at every layer, including who funds it.

Both SpaceX and Rocket Lab reached orbit for about $100M USD, with Falcon 1 and Electron respectively. That figure is the benchmark the whole plan is scaled against.

Three vehicles at different maturities. Taiga flies now as a pathfinder and doubles as a hypersonic test platform and a future Tundra third stage. Tundra is the one that matters — a light-lift orbital vehicle targeting 1,100 kg to low Earth orbit, first mission 2028. Titan is the medium-lift future, a cluster of the same engine.

Named for Chris Hadfield, patent pending, and described as 100% made in Canada. It is 3D printed in-house, post-processed in-house, and fired at the company's own test site. NordSpace published real numbers for it at the conference, which is unusual and worth reading closely.

The design priorities Goel names are not raw performance. They are cost, speed, testability, and what he calls forgiveness — engine-out capability, so that losing one engine does not lose the mission. That is a manufacturing-led philosophy, and it is consistent with printing engines in-house rather than buying them.

Behind it sits the AMA Lab — Advanced Manufacturing for Aerospace, established 2025 — scaling metal additive and automated composite manufacturing to orbital scale, covering engines, tanks and primary structures. Partners are the National Research Council Canada, Fraunhofer, and NGen. From May 2026 all of this runs out of a new 60,000 sq ft campus in Markham built to seat 235 staff.

The Atlantic Spaceport Complex sits near St. Lawrence on the Burin Peninsula in Newfoundland and Labrador. Ground broke in August 2025. SLC-02 handles suborbital; SLC-01 is the orbital site with at least two pads planned. Environmental approval is done, and the site is expandable 10 to 15 km out for a medium or heavy-lift future.

The claim is that it is unmatched in Canada on four axes, and the inclination range is the one that does the most work.

Three partners carry the parts NordSpace chose not to build. Kongsberg Geospatial runs launch control and range safety, including real-time geospatial decision support across launch, airspace and maritime operations. C-CORE, based in Newfoundland, provides ground stations for telemetry, tracking and control, and secure space-domain communications in the North Atlantic and Arctic. LeoLabs brings space domain awareness radar — Canada's first ground-based sub-metre-scale SDA capability for low Earth orbit, with dual-use launch vehicle tracking.

Those three names tell you where the line got drawn on vertical integration. Goel is candid that they walked back several things they thought they would build in-house.

This is the castle. If the thesis is right, this side eventually matters more than the rockets.

Terra-Nova is the first satellite, launching in 2026 on SpaceX's Transporter-18 — which is itself a nice illustration of the problem they are solving, since Canada's first sovereign-launch company currently has to buy a ride. It carries two in-house payloads. Chronos is an edge-AI imaging system: three cameras with an NVIDIA GPU behind them, doing real-time wildfire detection and Arctic monitoring rather than downlinking raw imagery. Zephyr-EP is a small electric thruster for deorbit and manoeuvring, developed by leaning on the rocket propulsion team — the clearest example of the two divisions actually feeding each other.

The platform scales to 10–100 kg satellites for hosted payloads and small constellations, and the stated goal is explicit: boost Canadian satellite output while generating internal demand for Tundra. That is the flywheel, said out loud.

Canada has never sent a dedicated orbital or surface mission to the Moon. NordSpace has three planned — two surface, one orbital — and the first is privately funded and built by two engineers.

This is the part I find most instructive. NordSpace is over 90% self-financed by its founder, who previously built and exited an events software company. The stated reason for refusing venture capital is that VC pressure pushes a company toward the next round and toward liquidity, rather than toward missions with long-term value. The company is 100% Canadian-owned, and the argument is that this is a precondition for sovereignty rather than a preference.

Team size was about 30 at the keynote, targeting 60 by the end of 2026 and 250+ full-time by 2030. The existing team carries over 200 years of cumulative space experience between them, drawn from MDA Space, Rocket Lab, SFL Missions, Firefly, RFA, Orbex, SpaceX, NASA JPL, Yuzhnoye, Magellan, Kepler, Honeywell and COM DEV — people who have worked on everything from the Canadarm to Starship.

The company has also put over $200,000 into student teams building robots, satellites and rockets. Goel came up through student teams himself, which is probably not a coincidence.

The keynote closes with a national roadmap rather than a company one, scoped against roughly $250M per year of Canadian investment. Take it as a statement of ambition rather than a schedule — but the point being made is that the absence of any answer to "what's next" is itself the problem.

I'm a mechanical engineer who spent the early part of a career on P&IDs, refrigeration systems and high-pressure oil and gas, then a decade building and running companies. Reading this closely, three things stand out.

The hard problems here are industrial, not exotic. Cryogenic ground support, high-pressure feed systems, leak and pressure qualification, remote-site construction, supply chain at cadence. These are recognisable problems from process and heavy industry, and the AIT posting says so outright by listing oil and gas and chemical manufacturing among its preferred prior industries.

The spaceport is a logistics problem wearing an aerospace hat. Building SLC-01 on the Burin Peninsula means permits, environmental conditions, remote logistics, weather, and multi-discipline civil, electrical and mechanical delivery — before anything aerospace-specific happens.

Going from 30 to 250 people is where most of this gets decided. Vertical integration is a claim about owning your own processes. Somebody has to build the processes.

I wrote these notes because I want to work on this. If that's relevant to you, my background is here.