Economics & Viability
Orbital compute economics depend on three converging trends: GPU performance/watt doubling every generation, launch costs falling 10× with Starship, and growing demand for sovereign/secure compute.
Breakeven Analysis by Scenario
H100 + Falcon 9 (today)
$16.36/GPU-hr
Gov/IC only, viable at security premiums
B200 + Starship (2029)
$6.20/GPU-hr
Gov + premium cloud competitive
X100 + Starship (2031)
$3.04/GPU-hr
Premium commercial viable
12-Satellite Fleet (2035)
$2.58/GPU-hr
Mainstream cloud competitive
GPU Performance Roadmap
| Year | GPU | FP8 TFLOPS | TDP (W) | Perf/W vs H100 | Cost/GPU |
|---|---|---|---|---|---|
| 2023 | H100 | 990 | 700 | 1.0× | $30K |
| 2025 | B200 | 2,250 | 700 | 2.3× | $35K |
| 2027 | B300/X100 | 4,500 | 700 | 4.5× | $35K |
| 2029 | Next-gen | 9,000 | 700 | 9.1× | $30K |
| 2031 | Next-next-gen | 15,000 | 600 | 17.7× | $25K |
Key insight: Centradiant's fixed 55 kW thermal budget accommodates every future GPU generation. As compute/watt doubles, the same satellite delivers 2× more GPU-hours without hardware changes.
Launch Cost Evolution (to 700 km SSO)
| Year | Vehicle | $/kg | vs Today |
|---|---|---|---|
| 2026 | Falcon 9 rideshare | $2,720 | 1.0× |
| 2028 | Starship (early) | $1,500 | 0.55× |
| 2029 | Starship (maturing) | $1,000 | 0.37× |
| 2030 | Starship (mature) | $600 | 0.22× |
| 2032 | Starship (volume) | $400 | 0.15× |
| 2035 | Starship (amortized) | $300 | 0.11× |
Ground Station Strategy
Dedicated Ground Station
$9M
CAPEX for Ka-band station + ops center. 5-year TCO: $12M+
KSAT Per-Minute (Recommended)
$12K/yr
Commercial GS-as-a-Service. Global network, no CAPEX. Saves $9.4M in capex.
Satellite Unit Cost (Pathfinder, 2029)
| Category | Cost |
|---|---|
| Structure & mechanisms | $2.0M |
| Thermal / LDR system | $3.0M |
| Compute (64× H100) | $1.9M |
| Power (solar + battery) | $3.0M |
| ADCS / avionics | $2.0M |
| Integration & test | $2.0M |
| Launch (F9 rideshare) | $2.5M |
| Ground test NRE | $1.6M |
| TOTAL (Unit 1) | $18.0M |
85% learning curve reduces unit cost to ~$9.4M by unit 12 (2035).
Fleet Economics (7-Year Projection)
| Year | Active Sats | Revenue | OPEX | CAPEX | Cumulative |
|---|---|---|---|---|---|
| 2027 | 0 | $1.5M | $0.5M | $0.2M | -$0.2M |
| 2028 | 0 | $10M | $2.0M | $17M | -$9.2M |
| 2029 | 1 | $5M | $1.3M | $0 | -$5.5M |
| 2030 | 1 | $12.6M | $1.3M | $0 | +$5.8M |
| 2031 | 3 | $25M | $2.5M | $26M | +$2.3M |
| 2032 | 5 | $50M | $3.5M | $22M | +$26.8M |
| 2033 | 7 | $75M | $4.5M | $20M | +$77.3M |
$179M
Total 7-year revenue
$94M
Total 7-year profit
Year 4
Payback (2030)
Sensitivity Ranking
Impact on breakeven price (from $16.36/GPU-hr baseline):
#1
GPU evolution (-$12.04/hr)
#2
B200 GPU alone (-$6.02/hr)
#3
Starship ($1K/kg) (-$3.83/hr)
#4
Longer life 7yr (-$1.59/hr)
#5
Higher utilization (-$1.07/hr)
GPU evolution is the #1 economic lever, and it's the one variable that improves on its own. Centradiant's fixed thermal budget means every GPU generation doubles the value of the same satellite.
Per-Satellite Economics
Annual Revenue (Hybrid Model)
$12.6M
DoD lease + subscriptions + spot
Annual Profit
$8.1M
63% gross margin
5-Year ROI
+165%
Per satellite over design life
IRR (Fleet)
45–60%
7-year fleet projection