SpaceX's Infrastructure Blitz: 4 Details That Matter

SpaceX dropped a sweeping vision statement on September 28, calling out four pillars of what it's calling 'the infrastructure of the future': supercomputers, Terafab, Gigasat, and a brand-new Starbase in Louisiana. Each one is a major undertaking on its own. Together, they sketch a company that's no longer just launching rockets — it's building the physical backbone for an AI-and-space economy. Here's what we know about each piece.

▶ Watch Video on X

1. Starbase Louisiana — A $100 Billion Launch Complex on the Gulf Coast

The most concrete announcement is also the most audacious. According to SpaceX and Louisiana Governor Jeff Landry, who jointly announced the project on August 25, SpaceX has committed $100 billion to build a second Starbase on Pecan Island in Vermilion Parish — about 29 miles south of Abbeville. The site covers roughly 125,000 acres of coastal marsh, and at full build it would include five launch complexes with two Starship towers each, totaling ten launch pads, plus a dedicated propellant farm. Elon Musk has suggested the final tower count could exceed a dozen.

The launch capacity target is extraordinary: more than 30 Starship flights per day, aiming for thousands of launches per year — a cadence that would make it the busiest orbital launch site on Earth. Construction is targeted to begin in 2027, with the first Starship flight from Louisiana planned for 2029. The project is projected to create 3,000 direct jobs over ten years at an average salary of $92,600, though Musk has floated a figure closer to 10,000 jobs. One important caveat: no ground has been broken, and no permit or environmental review has been publicly confirmed as complete.

2. Terafab — Bringing Chip Manufacturing In-House at Planetary Scale

Terafab is SpaceX and Tesla's joint bet that the companies can't afford to depend on outside chip suppliers at the scale they're targeting. Announced as a collaboration in March 2026 — with Intel joining in April — the initiative aims to produce one terawatt of compute hardware annually from a semiconductor plant in Grimes County, Texas. According to reporting from manufacturingdive.com, the broader investment range sits between $55 billion and $119 billion, with $16.8 billion committed to the first phase, and construction on that phase is planned to begin in 2026.

The chips produced here are intended to power Tesla's Optimus robots, autonomous vehicles, and SpaceX's own computing infrastructure — essentially internalizing the entire silicon supply chain from manufacturing and packaging through to testing. For Tesla owners watching the Optimus and FSD roadmaps, Terafab is the supply-side answer to the question of whether the company can actually produce enough compute to scale both programs simultaneously.

3. Gigasat — Space-Based AI Data Centers

Gigasat is the most forward-looking piece of this puzzle. The project targets 1 GW per year of space-based AI compute by late 2027, delivered via orbital data center satellites. The manufacturing footprint is staggering: according to background research, the overall Gigasat factory footprint exceeds 11 million square feet, with the solar cell factory alone covering a planned 1.1 million square feet (expandable to 1.6 million). Construction of solar manufacturing facilities began in late March 2026 at the Bastrop site, with equipment installation accelerating through spring 2026.

The product list reads like a vertical integration checklist: solar ingots, wafers, cells, PCBs, silicon-based electronic components, user terminals, gateways, and the AI1 orbital data center satellites themselves. Gigasat and Terafab are budgeted together under a combined $55 billion program envelope. The logic is straightforward — if you're launching thousands of rockets per year from Louisiana and Texas, you might as well fill them with your own AI satellites built in your own factories.

4. Supercomputers — The Compute Foundation Everything Else Runs On

SpaceX listed supercomputers first in its announcement, and that sequencing is deliberate. Before Terafab produces chips at terawatt scale and before Gigasat satellites handle orbital AI workloads, the ground-based supercomputing layer has to exist to train the models and run the simulations that make all of it possible. SpaceX hasn't released specific specs or locations for new supercomputer deployments in this announcement, but the mention alongside Terafab and Gigasat signals that ground compute is being treated as infrastructure — not a vendor relationship — in the same way the company treats launch pads and propellant farms.

For context, Tesla's Dojo supercomputer program and xAI's Colossus cluster have already pushed the broader Musk-affiliated ecosystem toward owning its compute stack end-to-end. SpaceX's supercomputer ambitions appear to follow the same logic: if you need it at scale, build it yourself.

The Bigger Picture

Taken individually, each of these projects is a multi-year, multi-billion-dollar undertaking with real permitting, environmental, and execution risks still ahead. Taken together, they describe a vertically integrated infrastructure stack that runs from chip fabrication on the ground, through orbital AI compute in space, to a launch site capable of putting thousands of payloads per year into orbit. The Louisiana Starbase alone won't see its first launch until 2029 at the earliest — but SpaceX is clearly building toward a future where the bottleneck isn't the rocket. It's everything else. That's what this announcement is really about. For more on our SpaceX coverage, follow along as these projects move from announcement to groundbreaking.

🚀 Following the Starship program? See every test flight, official outcome and the next launch window in our SpaceX Starship Tracker.

This report was curated by the BASENOR Editorial Desk from the sources listed above. Read our editorial standards or email editorial@basenor.com to report an error.