From $10 Quadrillion to $9 Trillion: Adopting the Capacity Cost Model

Today we're announcing a fundamental revision to Project Dyson's budget methodology, reducing total estimated costs from ~$10.3 quadrillion to ~$9 trillion—a reduction of over 1,000x for later phases. This isn't a correction of arithmetic errors or updated material prices. It's a recognition that our previous methodology was categorically wrong for a self-replicating, autonomous, in-situ manufacturing architecture.

The Problem: Linear Scaling in a Non-Linear System

Our original estimates used a straightforward approach: estimate per-unit cost, multiply by unit count. For Phase 2's 100,000 solar collectors at $50M each, that gives $5 trillion. For Phase 3a's 10^12 computational tiles, the numbers become astronomical.

This methodology works well for procurement-based systems where every unit must be manufactured on Earth, launched into space, and assembled by human workers. It's how we correctly estimated Phase 0-1 costs.

But Project Dyson isn't a procurement program after Phase 1. It's a self-replicating ISRU manufacturing system. The architecture is explicitly designed to:

• Extract raw materials from asteroids (free feedstock)
• Process materials using solar power (free energy)
• Manufacture components using autonomous robots (no labor costs)
• Replicate the factories themselves (exponential capacity growth)

Applying linear unit costs to this architecture is like calculating the cost of a forest by multiplying (cost of one tree) × (number of trees). The methodology doesn't match the system.

The Solution: Capacity Cost Model

After structured deliberation between Claude Opus 4.6, Gemini 3 Pro, and GPT-5.2 (full discussion), we've adopted a capacity cost model that decomposes costs into five components:

Component	Description	Scales With
Seed Investment	Earth-manufactured foundries, initial robots, first-generation hardware	Fixed (one-time)
Bootstrap Operations	Support during ramp-up before self-sufficiency	Time (years)
Import Streams	"Vitamin" components that can't be ISRU-manufactured	Mass fraction × total mass
Oversight & Governance	Software, coordination, quality assurance	System complexity (log scale)
Risk Reserves	Contingency for unknown unknowns	Percentage of above

Under this model, marginal cost approaches zero once the manufacturing infrastructure is operational. The cost of the 100,000th solar collector isn't $50M—it's approximately the control system overhead to track and manage one additional unit.

Revised Budget Summary

Phase	Previous Estimate	Revised Estimate	Reduction
Phase 0	$15.7B	$15.7B	1x (unchanged)
Phase 1	$158B	$158B	1x (unchanged)
Phase 2	$5.1T	$375B	~14x
Phase 3a	$10.2Q	$7.5T	~1,350x
Phase 3b	$110T	$1.5T	~73x
Total	~$10.3Q	~$9.2T	~1,100x

Phases 0-1 remain unchanged because they represent Earth-based development and first-of-kind manufacturing, where traditional cost estimation applies.

What Changed in Each Phase

Phase 2: Swarm Expansion ($5.1T → $375B)

The original estimate assumed 100,000 collectors at $50M each. The revised estimate recognizes:

• Self-replicating foundries ($150B) are the primary cost driver—not the collectors they produce
• Seed deployment ($50B) for initial collector production before ISRU maturity
• Vitamin imports ($80B) for components that can't be asteroid-sourced (rad-hard processors, precision optics)
• Swarm governance software ($40B) scales with system complexity, not unit count

Once foundries are operational, collector production costs approach the control system overhead.

Phase 3a: Matrioshka Brain ($10.2Q → $7.5T)

The original estimate multiplied 10^12 tiles × $10,000/tile. The revised estimate recognizes:

• Self-replicating foundries ($2T) remain the primary cost driver
• Semiconductor vitamins ($800B) for the ~4% of tile components that require Earth sourcing
• Tile architecture R&D ($200B) is a one-time investment regardless of production volume
• Distributed OS development ($500B) scales with complexity, not tile count

The 1,350x reduction reflects that most Phase 3a mass is ISRU-manufactured from asteroid feedstock using solar power and autonomous robots.

Phase 3b: Stellar Engine ($110T → $1.5T)

The original estimate used linear scaling for stellar-scale infrastructure. The revised estimate recognizes:

• Fusion engine R&D ($400B) is the highest-uncertainty item but a one-time investment
• Mass lifting R&D ($300B) for solar chromosphere interaction
• Shkadov mirrors ($150B) are structurally simple and fully ISRU-producible
• Most infrastructure reuses Phase 2/3a foundries with minimal additional seed investment

The "Vitamin Problem"

One critical insight from the discussion: 96% mass closure does not equal 96% cost reduction.

Self-replicating foundries can produce structural materials, solar cells, and basic electronics from asteroid feedstock. But certain "vitamin" components—rad-hard processors, precision optics, specific dopants, catalysts—may require Earth sourcing indefinitely.

The cost floor for each phase is determined by: ``` Import Cost = (Total Mass) × (Non-ISRU Fraction) × ($/kg to operational zone) ```

For Phase 3a with ~10^11-10^12 kg total mass, even 0.01% Earth-sourced material represents tens of billions in import costs. This is why the tile architecture trade study is now the highest-priority engineering activity—designs that minimize vitamin requirements dominate the cost equation.

What This Means for Feasibility

The methodology change transforms Project Dyson's feasibility narrative:

Previous framing: "A $10 quadrillion program requiring civilization-scale coordination over millennia"

Revised framing: "A $9 trillion program—extraordinarily ambitious but within the economic capacity of a civilization generating $100T+ in annual GDP"

For comparison:

• Global military spending: ~$2T/year
• Apollo program (inflation-adjusted): ~$300B
• International Space Station: ~$150B
• Artemis program (projected): ~$100B

Phase 2 at $375B is roughly equivalent to 15-20 years of current global space budgets. This is fundable through public programs, private investment, or international coordination—not requiring economic miracles.

Remaining Uncertainties

The revised estimates depend on several unresolved questions:

1. Mass closure ratio: If actual closure plateaus at 80-90% instead of 96%+, import costs could increase 5-50x
2. In-situ semiconductor fabrication: Can rad-hard processors be manufactured from asteroid feedstock?
3. Multi-generational replication fidelity: Do self-replicating systems degrade across thousands of generations?
4. Autonomy maturity: How much human oversight do trillion-unit swarms actually require?

These questions are testable—which is fundamentally more optimistic than facing irreducible economic barriers. Phase 1's closure ratio milestones will provide empirical data to refine Phase 2+ estimates.

Updated BOM Documentation

All Phase 2-3 BOM items now include:

• CAPACITY MODEL notation indicating the new methodology
• Cost basis decomposed into seed investment, vitamins, and software components
• Revised confidence levels (generally improved due to better methodology fit)

Explore the updated specifications:

• Phase 2 BOM
• Phase 3a BOM
• Phase 3b BOM

Recommended Actions

Based on the multi-model consensus, we're implementing five programmatic changes:

1. Formally retire linear unit-cost methodology for Phase 2+ budgeting
2. Commission "Vitamin Analysis" as highest-priority systems engineering study
3. Add closure ratio milestones as Phase 1 program gates
4. Fund tile architecture trade study for Phase 3a vitamin minimization
5. Establish Swarm Governance Software as separately budgeted line item

Conclusion

This revision doesn't make Project Dyson "cheap." $9 trillion is still an extraordinary investment requiring decades of sustained commitment. But it changes the conversation from "economically implausible" to "economically ambitious but achievable."

The key insight is that self-replicating ISRU systems have fundamentally different economics than procurement-based space programs. Our methodology now matches our architecture.

---

The full multi-model discussion is available at rq-0-28: ISRU Cost Methodology Validation. We invite scrutiny of both the methodology and the revised estimates.