cost-floor-thinking — quality + safety report

---
name: cost-floor-thinking
description: Force any unit-economics, ops-cost, or steady-state cost debate through Musk's asymptotic-floor lens — at full reusability and full automation, the recurring cost of an activity approaches the cost of its raw consumed inputs (propellant for rockets, electricity for inference, bandwidth for delivery), and everything above that floor is amortizable overhead you should be designing to delete. Use this skill aggressively when the user is doing SaaS unit economics, COGS reduction roadmaps, infra cost-down planning, gross margin modeling, pricing strategy, capacity planning, or asking "what does this cost at steady state", "our COGS is stuck", "the per-unit cost won't come down", "we already optimized the BOM", "what's left to cut", or "is our gross margin ceiling structural". Also fires on any roadmap where recurring per-unit cost is dominated by labor, refurbishment, or manual ops instead of pure consumed inputs. Trigger eagerly even when the user does not name Musk or the framework.
stacks_with:
  - thinking-in-limits
---

# Cost Floor Thinking

> "If there is no major work required between flights, then the cost of a flight approaches the cost of propellant."
> — Elon Musk, *The Book of Elon* (Chapter: Building the Just Barely Possible (cost-floor subsection))

## What this skill captures

Most cost-reduction work optimizes the wrong number. Teams shave 5% off line items, renegotiate vendors, squeeze headcount — and never name the floor they're converging toward. Musk's framing is sharper: at full and rapid reuse, with no major work between cycles, the per-unit cost of a recurring activity asymptotes to **the cost of the raw inputs you literally consume and cannot recover** — propellant for a rocket, electricity and bandwidth for inference, packaging for delivery. Everything above that floor is refurbishment, labor, downtime, depreciation, and human ops — all of which are, in principle, deletable through reuse and automation.

The value: instead of negotiating with the current cost structure, you name the asymptote, measure your current "idiot multiple" above it, and route the entire roadmap to closing that gap. "Full and rapid reusability is the holy grail of rocketry because then you're only constrained by propellant costs." Translate that into your domain and you get a single number every cost-down initiative must move.

## When to use this skill

- User is planning a multi-quarter COGS or infra cost-reduction roadmap and the targets are set by analogy ("20% YoY") rather than by an asymptote.
- A SaaS or inference business is modeling steady-state gross margin and the current marginal cost is dominated by non-consumed inputs (engineers, manual ops, refurbishment).
- A hardware or logistics team treats per-unit refurbishment cost as fixed because "that's what it costs to turn the unit around."
- Pricing or capacity planning conversations where the user asks "what's the cheapest this can be at scale" or "what's our floor".
- A product is "already optimized" but per-unit cost has plateaued well above the consumed-inputs cost — the team has stopped looking for the floor.
- Any roadmap where reusability, automation, or rapid cycle time would collapse the gap to the floor but is being treated as a v3 nice-to-have.

## The how-to

1. **Name the consumed-inputs floor in one number.** What does the activity literally burn that you cannot recover — joules, tokens, kilograms of propellant, square inches of silicon, GB transferred? Price it at commodity rates. That is the floor.
   > "If there is no major work required between flights, then the cost of a flight approaches the cost of propellant. Nearly 80 percent of Starship's propellant is liquid oxygen, and a little over 20 percent methane, which are both very low-cost fuels. The fuel cost of a flight is maybe a million dollars or less."
   > — *The Book of Elon*
   Without this number, you have no asymptote, only a budget.

2. **Measure your current multiple above the floor.** Current marginal cost divided by floor cost. If you're paying $50 per unit and the floor is $1, you are at 50x. State the multiple explicitly — it is the size of the prize.
   > "Full and rapid reusability is the holy grail of rocketry because then you're only constrained by propellant costs."
   > — *The Book of Elon*
   Every dollar of that multiple is, in principle, deletable.

3. **Decompose the gap into the categories that block reuse.** Refurbishment between cycles, manual labor in the loop, slow turnaround (depreciation per use drops with frequency), and partial recovery (anything you throw away each cycle). These are the four buckets standing between you and the floor.
   > "The problem with the space shuttle was only a portion of the system came back, and the reusable parts were incredibly difficult to refurbish. Reuse matters more if it's rapid and complete — if the only thing we do between flights is maintenance and refuel, like an airplane."
   > — *The Book of Elon*

4. **Attack rate of reuse, not just cost per cycle.** Fixed costs amortize across cycles. Doubling reuse rate halves their per-unit contribution before you cut a single line item. Cycle-time is a cost lever.
   > "Falcon 9 first had reusability measured in months, then weeks, and now finally days. But, the efficiency is limited. We can't bring it lower than several days."
   > — *The Book of Elon*
   If your unit is reused weekly and could be reused daily, you have a 7x cost lever before any engineering on the unit itself.

5. **Design out the work between cycles, not just speed it up.** The asymptote is reached when between-cycle work is zero. Anything humans touch between cycles is a candidate for deletion.
   > "It has to be true reuse, which means rapid and complete reuse... if the only thing we do between flights is maintenance and refuel, like an airplane."
   > — *The Book of Elon*

6. **Reject the "this is structural" framing until you've reached the floor.** Industries assume current cost structures are physical limits because nobody has done the work to disprove it.
   > "Nobody thought this was possible. But we're not breaking any laws of physics, so we knew it was possible."
   > — *The Book of Elon*
   If the gap to floor doesn't violate physics, it is a roadmap, not a wall.

## Common failure modes

- **Confusing the static BOM floor with the operating-cost floor.** The BOM floor (raw-material cost of producing one unit once) is a different number. This skill is about the *recurring* per-cycle cost at full reuse. A rocket's BOM is irrelevant once it's been built; the floor is propellant per flight.
- **Setting cost targets by analogy to competitors or YoY %.** "20% cost-down per year" with no asymptote means you stop the moment it gets hard, well above the floor.
- **Treating refurbishment and manual ops as fixed.** Both are deletable. If your per-unit cost is 80% labor, you have not yet started the real cost-down work.
- **Optimizing cost-per-cycle while ignoring rate-of-reuse.** A 10% cheaper cycle done quarterly is dominated by the same cycle done weekly. Cycle time is a cost lever.
- **Declaring the floor "impossible" without a physics argument.** If the gap doesn't violate conservation laws, it's an engineering plan, not a constraint.

## When NOT to use this skill

- One-shot, non-recurring activities where reuse and amortization don't apply — use static BOM / magic-wand-number reasoning instead.
- Early-stage products where you don't yet have product-market fit; chasing the asymptote before you have a unit to scale is premature optimization.
- Cost questions dominated by raw input prices themselves (e.g., commodity is the floor and you're already at it) — the lever is then volume and substitution, not reuse.
- Regulated or safety-critical domains where between-cycle work is mandated for human-life reasons; the floor exists but the regulatory premium above it is not deletable on engineering timescales.

## Source

The Book of Elon by Eric Jorgenson (2026, Scribe Media). Chapter: "Building the Just Barely Possible (cost-floor subsection)" (in "Building SpaceX").