NASA’s Space Launch System rocket launches carrying the Orion spacecraft launched at 6:35pm EDT on April 1, 2026 - Image Credit: NASA/Bill Ingalls
Legacy Hardware, Brownfield Design, and Why Artemis Costs So Much
I was 12 years old in July of 1969. I was in the family farm’s bunkhouse, and heard on the radio that Apollo 11 was about to land. I decided to join my family at the main farmhouse and watch it on our 13-inch black and white TV with them. It was raining hard, and so I hopped on my bike to cross the few hundred yards quicker. Halfway there, next to the milking parlor, I took a spill right into a mud puddle. Drenched and muddy, I walked my bike the rest of the way. Still wet, I was able to join the family to watch the moon landing. It’s hard to describe how proud we all were, not being just Americans but being part of mankind. That as a species we were able to accomplish such an amazing feat.
But after 4 years, 11 crewed missions and six successful Moon landings, the program was cancelled because it took up over 4% of the federal budget and we had more immediate earthly concerns. [1] Bottom line, it’s good to have great ambitions, but it has to be tempered with the reality of funding and other priorities.
The Illusion of a Complete System
Now here we are, 54 years later, attempting to repeat our previous success. And you might reasonably ask: if we did it before, with slide rules and teletype machines and computers with less processing power than the phone in your pocket, why is it taking so long and costing so much to do it again? The answer has almost nothing to do with technology. It has to do with what happens when you tell your engineers they have to build something new out of something old, and then act surprised when the old stuff doesn’t quite fit the new problem.
The INCOSE Systems Engineering Handbook, Fifth Edition, has a term for this. Actually, it borrows a term from real estate, which I find delightful because there is something genuinely architectural about the problem. The SE Handbook distinguishes between “greenfield” systems and “brownfield” systems. Greenfield is the farmer’s field, the undeveloped lot, the blank sheet of paper. You start fresh. Brownfield is the previously developed land, the lot with existing structures and services already in place, and (here’s the part that matters) potentially containing “undesirable or hazardous materials that must be remediated.” [2] The SE Handbook uses “legacy” and “brownfield” interchangeably, and the metaphor of contaminated land requiring cleanup before you can build anything new is, frankly, more apt than I think the authors intended.
When the SE Handbook defines brownfield SE, it says the new system architecture “must take into account the existing system elements and functions, which impose constraints on the overall system definition.” [3] That word “constraints” is doing an enormous amount of work in that sentence. It means your design space, the set of all possible solutions you could explore, has been pruned before you even pick up a pencil. And the pruning wasn’t done by physics or by your mission requirements. It was done by decisions someone else made, for a different system, to solve a different problem, sometimes decades ago.
The Space Launch System is, I would argue, the most expensive case study in brownfield design constraint that our industry has ever produced. In 2010, after the Constellation program was cancelled, Congress passed the NASA Authorization Act directing NASA to build a new heavy-lift rocket. But Congress didn’t just say “build us a big rocket.” The law directed NASA to use existing Space Shuttle and Constellation program contracts, hardware, and workforce “to the extent practicable.” [4] That phrase, “to the extent practicable,” sounds like it leaves room for engineering judgment. In practice, it meant that the RS-25 engines that had powered the Space Shuttle, the five-segment solid rocket boosters developed for the Ares I, the 8.4-meter-diameter tank tooling from the Shuttle external tank, and the workforce that had built and maintained all of it, were not optional design inputs. They were mandated ones.
I want to pause here because I think this is where my students sometimes miss the subtlety. A design constraint is not inherently a bad thing. Every engineered system operates within constraints. The laws of thermodynamics are constraints. Your budget is a constraint. The speed of light is a constraint (though mostly only if you’re building interstellar probes or fiber optic networks). Constraints are what make engineering interesting rather than merely imaginative. The problem isn’t that SLS had constraints. The problem is that these particular constraints were not derived from the mission need or the operational environment or the physics of getting to the Moon. They were derived from the political need to preserve an existing industrial base and workforce. The constraint source matters because it determines whether the constraint is pointing you toward a better solution or away from one.
And the SE Handbook, to its credit, warns about exactly this kind of trap. In the section on knowledge management and reuse, it identifies what it calls “serious traps” in reuse, and I encourage you to read them because they read like a prophecy written for the SLS program. Was the prior solution intended for a different use, environment, or performance level? Trap. Is the new application outside the qualified range? Trap. The SE Handbook even provides an example that should make anyone in the space business wince: a NASA Mars probe that failed because the development team reused a radiator design from an Earth-orbiting satellite without recognizing that the thermal environment in deep space is fundamentally different from Earth orbit. [5] Same space, different problem. The radiator worked perfectly in the environment it was designed for. It just wasn’t designed for the environment it was sent to.
The RS-25 engines are a fascinating case of this reuse trap in action. These are magnificent engines. They powered 135 Space Shuttle missions with remarkable reliability. They are also engines that were designed to be reusable, recovered after each flight, inspected, refurbished, and flown again. SLS uses them as expendable engines, thrown away after a single use. That’s like buying a Rolex to use as a paperweight. It technically works, but you’re paying for precision you’re not using and discarding value that was engineered to be preserved. And because only sixteen RS-25D engines remained from the Shuttle program (fourteen flight engines plus two assembled from spares), the design was locked into a finite supply of its own primary propulsion. [6] The heritage hardware was, by definition, a dwindling resource. The handbook’s brownfield waste, made literal.
RS-25 developmental engine at the Fred Haise Test Stand at NASA’s Stennis Space Center
The cost numbers tell a story that needs very little editorial embellishment. SLS took over thirteen years from program start to its first crewed flight and consumed roughly $23 billion in development costs. [7] Total Artemis vehicle development expenditures have reached approximately $61 billion in inflation-adjusted dollars. [8] Each SLS launch costs in the neighborhood of $4.1 billion. [9] And a Booz Allen Hamilton analysis found that using legacy hardware actually cost more than new development would have, partly because SLS employed the expensive legacy workforces who had worked on the Shuttle. [10] Let that sink in. The whole rationale for mandating legacy hardware was to save money by avoiding new development. The actual result was higher costs than a clean-sheet design.
Now, set all of that next to what SpaceX did with no legacy mandate, no Congressional direction to use anyone’s old hardware, and no obligation to preserve any particular workforce or factory.
SpaceX developed the Falcon 9 from a blank sheet to first launch in four and a half years for just over $300 million. [11] NASA’s own cost modeling tools estimated that the same rocket, developed through traditional government contracting, would have cost between $1.7 billion and $4 billion. [12] That’s not a marginal improvement. That’s an order of magnitude. And when Congressional testimony explored why the cost was so much lower, the explanation was almost embarrassingly simple: NASA had set “only a high-level requirement for cargo transport to the space station, leaving the details to industry.” [13] In other words, they defined what the system needed to do and let the engineers figure out how to do it, unconstrained by anyone’s legacy hardware sitting in a warehouse in Utah or Louisiana.
SpaceX’s Starship program tells a similar story. When the carbon fiber approach was taking too long, they switched to stainless steel, a material nobody had used in spacecraft since the Atlas rockets of the 1950s, because it was the right material for their design goals of rapid manufacturing and cryogenic performance. [14] That kind of architectural pivot is simply impossible in a brownfield-constrained program. You can’t look at your Congressional mandate to use RS-25 engines and say, “Actually, we’ve decided to go with methane-fueled Raptors instead, because they’re better for reusability and Mars refueling.” The legacy constraint forecloses not just the specific design choice but the entire class of design thinking that leads to optimization.
SpaceX Falcon 9 Launch Through the Clouds
The Falcon 9, as of this writing, has completed over 620 successful launches. [15] SLS has flown twice. The per-kilogram cost to low Earth orbit on a Falcon 9 is roughly $2,700, compared to approximately $37,000 on SLS. [16] And in February 2026, NASA administrator Jared Isaacman formally cancelled the planned Block 1B and Block 2 upgrades to SLS, acknowledging that the evolutionary path for the legacy-derived design was unsustainable. [17] The system will continue flying in its current Block 1 configuration for a handful of remaining Artemis missions, but the future belongs to commercially developed vehicles. Even Congress, which originally mandated that SLS launch the Europa Clipper probe to Jupiter, ultimately allowed NASA to bid that mission competitively. SpaceX won the contract, saving an estimated $2 billion. [18]
The INCOSE SE Handbook observes that organizations accustomed to brownfield development for long periods may need to “relearn” how to do greenfield. [19] I find that observation both poignant and precise. Legacy doesn’t just constrain your current design. Over time, it atrophies your capacity to think beyond the constraints. You stop asking “what’s the best solution to this problem?” and start asking “how can we make the old stuff work for the new mission?” Those are fundamentally different questions, and they lead to fundamentally different systems.
I still remember sitting in that farmhouse, muddy and wet, watching grainy footage of men walking on the Moon. The Saturn V that got them there was, for all practical purposes, a greenfield design. Yes, it built on the knowledge gained from Mercury and Gemini, but the vehicle itself was engineered from the ground up for the specific mission of getting humans to the lunar surface and back. The engineers at Marshall Space Flight Center and their contractors were given a destination, not a parts list. And they got there in eight years, from Kennedy’s speech to Armstrong’s boot print, for about $200 billion in today’s dollars. [20] Expensive, certainly. Four percent of the federal budget, unsustainably so. But they were solving the problem they were given, not solving the problem of what to do with hardware left over from a program that had already ended.
If there’s a systems engineering lesson in all of this, and I believe there is one worth teaching, it’s that legacy is first and foremost a design problem. It constrains your architecture, limits your trade space, introduces technical debt you didn’t incur, and creates the seductive illusion that because the hard work of development has already been done, the path forward should be easier and cheaper. The INCOSE SE Handbook’s brownfield framework understands this. The SE Handbook positions brownfield as a distinct system type with different life cycle approaches, different design review cadences, different team compositions, and different risk profiles. The question for practitioners is whether we’re honest with ourselves and our stakeholders about the true cost of those constraints, or whether we let the sunk cost of legacy hardware masquerade as a head start.
My twelve-year-old self, the one who fell off his bike in the rain to watch history happen on a tiny television, would probably be disappointed that it took us this long to go back. But my Systems Engineering self understands exactly why. We weren’t just trying to get to the Moon again. We were trying to get to the Moon while dragging the Space Shuttle behind us.
Optional Reader Resource
References
Wikipedia, “Apollo Program.” https://en.wikipedia.org/wiki/Apollo_program
INCOSE Systems Engineering Handbook, Fifth Edition (INCOSE SEH5), Sec. 4.3.1: Greenfield/Clean Sheet Systems.
INCOSE SEH5, Appendix C: Terms and Definitions, “Brownfield SE.”
U.S. Congress, NASA Authorization Act of 2010, S.3729, Sec. 302.
INCOSE SEH5, Sec. 2.3.3: Technical Management Processes, Knowledge Management subsection.
Wikipedia, “Space Launch System,” https://en.wikipedia.org/wiki/Space_Launch_System. Accessed April 2026.
SpaceX Stock, “SpaceX vs. NASA: Structural Design Approaches,” July 2025. https://spacexstock.com/spacex-vs-nasa-structural-design-approaches/
Payload Space, “Detailing Artemis Vehicle R&D Costs,” March 2024. https://payloadspace.com/payload-research-detailing-artemis-vehicle-rd-costs/
Aerospace America (AIAA), “Bending the Cost Curve,” April 2025.
NASA Fandom Wiki, “Space Launch System,” citing PopularMechanics analysis. https://nasa.fandom.com/wiki/Space_Launch_System
National Space Society, “Statement on Launch Costs from SpaceX CEO Elon Musk,” May 2011. https://nss.org/statement-from-spacex-ceo-elon-musk/
New Space Economy, “How much would Falcon 9 have cost if it was developed by NASA?” May 2023. https://newspaceeconomy.ca/2022/10/23/how-much-would-falcon-9-have-cost-if-it-was-developed-by-nasa/
Wikipedia, “Falcon 9,” https://en.wikipedia.org/wiki/Falcon_9. Accessed April 2026.
Productside, “Behind The Product: NASA SLS vs. SpaceX Starship,” April 2022. https://productside.com/behind-the-product-nasa-sls-vs-spacex-starship/
Wikipedia, “Falcon 9.” Accessed April 2026.
Hacker News discussion, November 2024. https://news.ycombinator.com/item?id=42070547
Wikipedia, “Space Launch System.” Accessed April 2026.
INCOSE SEH5, Sec. 4.3.1: Greenfield/Clean Sheet Systems.
Inflation-adjusted estimate based on commonly cited Apollo program total costs.
