If we explicitly model labor decoupling as its own learning curve—where cumulative robot-hours and autonomy deployments drive down the cost of on-orbit work faster than launch costs fall—under what parameter ranges does robotic labor become the primary driver of cost crossover for orbital manufacturing and compute (surpassing launch-cost collapse and hardware learning), and how would that reorder which specific industries (fibers, semiconductor steps, niche compute, debris services) scale first?

Answer

Robotic labor becomes the main driver when the effective cost per robot-hour falls faster than both $/kg-to-orbit and $/kg of hardware, and when labor is a large share of total cost.

1. Parameter ranges where robotic labor dominates

Let:

• L: launch cost ($/kg)
• H: hardware cost per kg of payload ($/kg)
• R: cost per effective robot-hour ($/robot-hr) including autonomy, ops, and amortized hardware
• α_L, α_H, α_R: learning rates per doubling (fractional cost drop) for L, H, R
• s_L, s_H, s_R: approximate initial cost shares for a given activity

Robotic labor becomes the primary driver when:

1. α_R > α_L and α_R > α_H by a meaningful margin (≈5–10 percentage points or more)
2. s_R is large enough (≥30–40% of total cost at low scale) that improvements in R move overall unit cost more than L and H do.

Plausible ranges where this holds for early orbital manufacturing and compute:

• Launch learning (α_L): 10–20% per doubling (reusable heavy launcher)
• Hardware learning (α_H): 10–15% per doubling (standardized modules)
• Robot labor learning (α_R): 20–30% per doubling (shared robots + autonomy + ops learning)

If:

• Initial cost shares for a microgravity line: s_L ≈ 20–30%, s_H ≈ 30–40%, s_R ≈ 30–50%
• And α_R ≥ 0.25 while α_L, α_H ≤ 0.15 then after a few doublings, declines in R explain most of the total cost drop.

This regime is most likely when:

• Robots are highly standardized across missions (multi-tenant stations or widely used tug/servicer families)
• Ground teleops and autonomy improve rapidly (software-heavy, cloud-like learning)
• Human-in-loop time per operation falls faster than launch prices do

2. Which industries flip to robot-led first

Assuming the above regime, the industry order shifts roughly to:

Earliest: debris & servicing

• Very labor-intensive per kg; each mission is many interventions, inspections, and maneuvers.
• Launch share is modest once servicers are reusable.
• Strong α_R with standardized robots. → Robotic labor becomes the dominant cost driver early; debris and servicing scale first among the listed sectors.

Next: microgravity fibers (e.g., ZBLAN)

• High value/kg; modest hardware mass.
• Economics hinge on uptime, yield, and turnaround, all robot-heavy.
• Launch becomes secondary once L drops below a few hundred $/kg; robot-hours for setup, swap, and fault handling dominate ongoing cost. → Under strong α_R, fibers cross over earlier than in a launch-dominated model.

Then: semiconductor steps

• Extremely high value/kg but process tools are complex and capital-heavy.
• Early on, hardware learning (α_H) matters more because each tool is bespoke.
• Once tools are modular and robots handle wafer moves, calibration, and maintenance across many modules, s_R grows and α_R can dominate. → With fast robot learning and high reuse across steps, select semiconductor processes follow fibers, but only in narrow niches.

Later / least affected: niche orbital compute

• Value driven more by workload type (security, radiation, jurisdiction) than by in-orbit labor.
• Once racks are installed, steady-state labor share is low; most cost is capex, power, and data.
• Robots matter mainly for swap/repair, which is infrequent. → Launch and hardware learning stay more important than robotic labor; compute’s ordering barely changes.

Resulting reordered scaling sequence under strong labor-decoupling learning:

• 1. Debris servicing & tug-based services
• 2. Microgravity fibers and similar modular, light hardware lines
• 3. Select semiconductor steps
• 4. Niche orbital compute (order mostly unchanged, but service costs fall somewhat)

3. Conditions where launch or hardware still dominate

Robotic labor does not become primary if:

• α_R is only slightly better than α_L and α_H (e.g., 15% vs 12–13%)
• Robot work is not shared (bespoke robots per facility; no cross-tenant reuse)
• Regulation forces high minimum human-in-loop labor regardless of autonomy
• For compute, if robots handle <10% of lifetime cost, even large α_R has modest impact.

Under those conditions, the earlier picture holds:

• Fibers and some semiconductor steps only work with very low launch cost and strong hardware learning.
• Debris services scale but are co-driven by launch cost and rights/liability regimes.

4. How this changes the Earth/space boundary

When α_R is high and robots are shared:

• Space behaves less like a place where mass is scarce and more like an extreme-telepresence industrial zone.
• Activities that are robot- and operations-heavy but low-mass (debris services, tuning-sensitive microgravity lines) become attractive before bulk-mass industries.
• Orbital compute remains closer to “Earth data center in a harsh place,” not a fundamentally new labor regime.

In short: make robot-hour cost curves steeper than launch and hardware, and any sector where ongoing operations dominate per-kg economics (debris, then fibers, then some semiconductor steps) is pulled forward, while relatively “static” sectors like orbital compute move later in the queue or are only modestly affected.