Orbital Computation · 2026-04-28

🛰️ Orbital Computation Watcher — 2026-04-28

🌞 Overview Energy Signs 1 GW Deal With Meta to Beam Space Solar Power to AI Data Centers
🏛️ SpaceNews DC Forum Frames the Architecture Debate Two Days Before Launch: FCC, Senate, Starcloud, Varda on April 30
💼 Seraphim's $474M Trust Raise Signals Public Market Appetite for AI+Space Convergence
🛡️ True Anomaly Raises $650 Million at $2.2B Valuation for AI-Enabled Golden Dome Space Interceptors
🇨🇳 China's April Sprint: LEO Internet Tests, EO Export, and Atmospheric Monitoring Signal Constellation Acceleration
⚡ arXiv Preprint Models 100+ kW/ton Orbital AI Panel Architecture, Quantifying the Thermal Ceiling

---

🌞 Overview Energy Signs 1 GW Deal With Meta to Beam Space Solar Power to AI Data Centers

Overview Energy, the space-based solar startup that emerged from stealth last December after demonstrating infrared laser power transmission, announced April 27 an agreement with Meta to deliver up to one gigawatt of continuous orbital solar power to Meta's AI data center network. It is the first commitment from a top-tier hyperscaler to space solar at meaningful scale, and it inverts the dominant orbital data center thesis in a structurally significant way.

The architecture places the power source—not the compute—in orbit. Geosynchronous satellites collect near-continuous solar power and transmit it via near-infrared wide-beam lasers to existing ground-based solar farms, enabling electricity generation after dark. "Put the energy part into space," CEO Marc Berte said. "Put the thing that doesn't need maintenance and can last a long time in space." The approach skips the engineering complexity of launching compute hardware, thermal management systems, and inter-satellite optical links that orbital data center constellations require.

Meta's rationale is explicitly AI-driven. VP of energy and sustainability Nat Sahlstrom framed the deal as part of an effort to diversify power sources for data centers supporting artificial intelligence applications. Meta simultaneously secured geothermal, nuclear, and a 100 GWh battery storage agreement with Noon Energy—a hedged, multi-vector energy strategy that treats space solar as one diversification channel rather than a primary source.

The 1 GW commitment is forward-looking. Overview plans an in-space demonstration in 2028, with commercial service beginning no earlier than 2030. The deal is a demand signal to capital markets and a validation of Overview's fundraising thesis before a single commercial watt has been transmitted.

The strategic divergence from orbital compute players is sharpening. SpaceX's AI Sat Mini constellation plans 100 kW per satellite with onboard processors; Starcloud's 88,000-satellite FCC application bets on lease-your-orbit infrastructure; Blue Origin's Project Sunrise frames compute and power together at LEO. Overview argues hyperscalers will keep their compute on the ground and pay for clean, continuous orbital power instead.

Both theses can coexist in a world of sustained AI energy demand growth, but they will compete for hyperscaler contracts and capital as each matures. The bellwether question is whether launch economics fall fast enough—before 2030—for orbital compute to undercut terrestrial total cost of ownership. If they do, Overview's position weakens; if they don't, space solar captures the margin by delivering the scarce resource (clean power) without requiring the orbital data center economics to close.

Sources:

---

🏛️ SpaceNews DC Forum Frames the Architecture Debate Two Days Before Launch: FCC, Senate, Starcloud, Varda on April 30

SpaceNews convenes its Opportunities for On-Orbit Computing event in Washington, D.C. on April 30—the first in-person forum where FCC Space Bureau chief Jay Schwarz and Senate Commerce Committee staff director Brad Grantz will jointly address the orbital data center market alongside industry operators. The gathering reflects how quickly the sector has shifted from hypothetical to structurally contested: the debate is no longer whether orbital compute is viable but who governs it, who powers it, and how fast the regulatory apparatus can respond.

The FCC's agenda is suddenly urgent. Within months the Commission must respond to three massive constellation applications seeking spectrum waivers from standard deployment milestone rules: SpaceX's one-million-satellite filing, Blue Origin's Project Sunrise for 51,600 satellites, and Starcloud's 88,000-satellite application. All three argue they are not warehousing spectrum (they primarily seek Ka-band for TT&C, not user links), which is the technical rationale for each waiver request. How Schwarz characterizes the FCC's milestone philosophy in public on April 30 will signal the regulatory posture for all three applications.

Overview Energy's Marc Berte shares the agenda with Starcloud CEO Philip Johnston—whose company has raised $170 million and staked its thesis on space-based compute being the winning architecture. The coexistence of both arguments in the same room reflects genuine architectural ambiguity in the market. A panel explicitly titled "Powering compute in space: solar ambitions and nuclear realities" brings in Planet's Brian Lewis (Pelican and Suncatcher missions) to evaluate whether solar-only architectures are sufficient. The inclusion of nuclear power signals awareness that certain orbits and mission profiles face the same intermittency constraints as terrestrial solar.

Varda Space's Delian Asparouhov anchors the closing session with a framing of what orbital data centers "will and won't be by 2035." Varda executed the first commercial return of a pharmaceutical synthesis payload from orbit in 2023—the closest analog to operational orbital manufacturing at commercial scale. Asparouhov's operational track record distinguishes his 2035 projection from the filed-but-unfunded forecasts that have dominated this market.

The April 30 event won't produce operational systems or FCC decisions, but it will generate a governance record. Regulators and legislators speaking publicly about their authority over multi-billion-satellite applications—before those applications are ruled on—will create interpretive precedents that shape all three constellation timelines. The meeting is functionally a pre-decisional policy hearing dressed as an industry conference.

Sources:

---

💼 Seraphim's $474M Trust Raise Signals Public Market Appetite for AI+Space Convergence

Seraphim Space Investment Trust, the London-listed vehicle whose shares have climbed more than 85% year-to-date, launched an offer on April 27 for up to £350 million ($474 million) in new C-shares—the largest fresh capital raise in the trust's history since its 2021 LSE listing. The move is a bellwether for whether public markets, not just venture capital, are prepared to price the AI+space convergence thesis at meaningful scale.

The timing is deliberate. SSIT shares have climbed 120% since listing, driven disproportionately by Iceye, the Finnish synthetic aperture radar operator that accounts for 39% of net asset value. Iceye's ascent reflects a structural repricing: SAR imagery satellites generating real-time AI-digestible data streams are being valued as recurring-revenue infrastructure rather than episodic aerospace hardware. Portfolio revenues grew 79% in 2025, with management projecting more than 85% of holdings profitable on an EBITDA basis in 2026.

The portfolio composition reveals where Seraphim sees durable stack value. All.Space (multi-orbit user terminals, 15.9% of NAV), D-Orbit (in-orbit transportation and logistics, 12.4%), and HawkEye 360 (RF geolocation, 10.1%, filed for IPO) together represent the connectivity, servicing, and sensing layers that orbital compute depends on—not the orbital compute spacecraft themselves. Seraphim has also recently formed a global advisory council to guide long-term strategy as geopolitical and technological shifts reshape the space economy.

CEO Mark Boggett identified three capital deployment themes: global security, climate and sustainability, and next-generation space infrastructure. The security framing is load-bearing: Iceye (SAR), HawkEye 360 (RF geolocation), and LeoLabs (space object tracking) are all dual-use or primarily defense-relevant, deliberately aligned with government contract flows that are more predictable than commercial AI infrastructure spend.

The C-share structure protects existing shareholders from dilution while new capital is deployed—signaling Boggett expects a substantial investment pipeline above current NAV, likely tied to the wave of space-AI companies approaching Series B and C. Seraphim's second early-stage venture fund closed above a $100 million target recently, focused on seed and Series A startups, while SSIT covers more mature firms.

What SSIT is conspicuously not investing in, by portfolio composition, is orbital data center infrastructure—no Starcloud, no Project Sunrise, no AI Sat Mini. The gap between FCC-filed constellation applications and revenue-generating operations is precisely the space Seraphim is declining to fund in 2026, while its competitors are racing in. The capital is flowing to companies that currently generate data and analytics revenue from operational satellites, not to those whose economics depend on unproven orbital compute cost structures.

Sources:

---

🛡️ True Anomaly Raises $650 Million at $2.2B Valuation for AI-Enabled Golden Dome Space Interceptors

True Anomaly, the Denver-based defense space startup, raised $650 million in a Series D on April 28, reaching a $2.2 billion valuation and totaling roughly $1 billion raised since its 2022 founding. The round was co-led by Eclipse and Riot Ventures, with participation from Paradigm, Atreides, G Squared, VanEck, and existing backers Accel, Menlo Ventures, and Meritech Capital.

The fundraise is structurally tied to the Pentagon's Golden Dome program. The U.S. Space Force's Space Systems Command awarded up to $3.2 billion across 12 companies—True Anomaly among them—for space-based interceptor prototypes designed to track and potentially disable hostile missiles during boost phase. That mission requires persistent orbital custody of targets, millisecond-latency targeting data exchange via Link-182 waveform protocols, and distributed AI decision-making across multiple orbital nodes. The compute and communication architecture Golden Dome demands is orders of magnitude more demanding than commercial orbital data centers.

CEO Even Rogers said the company has developed "new hardware and software to support space-based interceptors specifically." True Anomaly's platform: Jackal, a maneuverable satellite designed for orbital proximity operations, and Mosaic, a mission software platform for space situational awareness. The architecture is software-defined mission control tied to agile hardware—the structural pattern that distributed orbital AI inference requires. Facing the Golden Dome's documented technical and cost challenges, True Anomaly's bet is that adversarially-hardened space compute becomes the underlying infrastructure requirement for both defense and eventually commercial orbital AI.

The company's multi-contract footprint is instructive: simultaneously on the Golden Dome interceptor program, selected for the 14-company GEO surveillance pool worth $1.8 billion, and executing the Victus Haze tactically responsive launch mission. Three contract tracks converging on the same hardware/software platform at $2.2B valuation suggests the market is pricing True Anomaly's orbital AI stack as a durable franchise, not a single-program bet.

Growth metrics mirror Palantir and Anduril at comparable stages: ~150 to ~300 employees this year, targeting 500 by year-end, scaling manufacturing to 50 Jackals annually near Denver. A separate Long Beach facility handles overflow and diversification.

The cross-sector signal: Delian Asparouhov of Varda Space Industries, which shares the Founders Fund portfolio ecosystem, speaks at the April 30 SpaceNews orbital computing forum. The convergence of defense-funded orbital AI hardware and commercial orbital infrastructure capital—visible in True Anomaly's roster of crypto and institutional investors alongside Eclipse and Riot—is becoming structurally legible. Military programs will de-risk radiation-hardened orbital AI hardware for the commercial sector precisely as they did for GPS navigation, setting the component cost floor for the next generation of orbital compute platforms.

Sources:

---

🇨🇳 China's April Sprint: LEO Internet Tests, EO Export, and Atmospheric Monitoring Signal Constellation Acceleration

China executed three orbital launches across April 17–26, reaching 26 total flights year-to-date and advancing its LEO megaconstellation buildout through technology validation flights more strategically significant than their modest payload counts suggest. A Long March 2D lifted off April 24 from Xichang carrying four satellite internet technology test satellites from GalaxySpace, Changguang Satellite Technology, and the Harbin Institute of Technology—commercial firm, state enterprise, and academic institution simultaneously—testing direct-to-device broadband and space-ground network convergence in near-circular 505 km, 55-degree-inclination orbits.

The manifesting model reveals the strategy: China uses small technology validation batches—this launch follows a Jielong-3 internet test earlier in April and multiple prior test series dating to 2023—to de-risk constellation designs before committing to production runs. GalaxySpace has demonstrated 48 Gbps multi-beam Ka-band throughput on prior missions; the April 24 payload advances direct-to-handset capability, the same frontier that Starlink Direct-to-Cell and AST SpaceMobile are racing to prove commercially. The target vehicles for these technologies are Guowang (13,000 satellites) and Shanghai SpaceSat's Thousand Sails (up to 14,000 satellites).

The PRSC-EO3 launch for Pakistan on April 25 tells a different structural story: China is executing a multi-launch contract for Pakistan's space agency (PRSC-EO1 January 2025, PRSC-EO2 February 2026, EO3 now). EO satellite exports—turn-key imaging infrastructure for nations without indigenous programs—are becoming a systematic channel of space diplomacy alongside the International Lunar Research Station. Pakistan has signed onto ILRS and has two astronaut candidates in Beijing training for a Tiangong mission, signaling deep political alignment that EO contracts formalize.

The Daqi-2 atmospheric monitoring satellite launched April 17 adds a less visible layer: an Aerosol and Carbon Detection Lidar monitoring CO₂ and global aerosol distribution. Dual-use—domestic climate compliance and observation of industrial output worldwide—Daqi-2 extends China's sensing infrastructure into domains that feed AI-based environmental and economic modeling.

The structural asymmetry is sharpening: Western orbital data centers are at the FCC application stage (rhetoric). China is at the operational LEO buildout stage (execution). The West's three major orbital data center applicants—SpaceX, Blue Origin, Starcloud—collectively span a timeline of 2028–2030 for first commercial workloads. Guowang's first phase already has 108 satellites operational. By mid-2026, China will have exceeded 50 launches for the year. The 26-launch pace through late April—inclusive of internet tests, EO exports, atmospheric monitoring, and deep-space alignment launches—is not a promotional announcement; it is a production cadence.

Sources:

---

⚡ arXiv Preprint Models 100+ kW/ton Orbital AI Panel Architecture, Quantifying the Thermal Ceiling

A new preprint by Gaalema, Indyk, and Staley presents a distributed compute architecture for sun-synchronous orbit satellites achieving more than 100 kW of compute power per launched metric ton—a factor of 5× over conventional approaches—by integrating solar cells, radiators, and compute functions into a unified panel array rather than treating them as separate subsystems bolted to a bus.

The architectural innovation is structural. Each 1–4 m² panel uses a vapor chamber to transport heat from ICs to a large integral radiator surface, holding junction temperatures near 40°C. That thermal regime enables standard commercial silicon rather than expensive radiation-hardened components, compressing both component cost and qualification time. The resulting specific power is approximately 500 W/kg versus less than 100 W/kg for conventional implementations—the first published quantitative design that closes on Musk's March 21 Terafab claim that SpaceX "knows how to do heat rejection in space."

At maximum design point, a 16 MW computation satellite comprising 16,000 panels in a 20m × 2,200m array masses roughly 150 tons and fits within a single Starship payload envelope. The design is explicitly Starship-dependent—the mass budget only closes at SpaceX's projected sub-$500/kg-to-LEO launch cost. Panel sizes from 1–4 m² are analyzed to trade vapor chamber heat transport capacity against compute efficiency and inter-panel communication overhead, giving designers a parameterizable design space rather than a single point solution.

The paper's contribution to the current industry debate is falsifiability. Alternative architectures using separated subsystems face compounding mass penalties that scale quadratically with system size; the Gaalema approach trades that scaling curve for inter-panel thermal coupling complexity. For investors evaluating orbital data center companies, the 500 W/kg figure is a yardstick: Starcloud-3's 200 kW target at ~3 tons achieves ~67 W/kg—real, but well below the theoretical optimum. The gap is engineering space that new entrants will race to close through 2028.

The 512-panel (1 kW/panel) sub-configuration is particularly interesting: it fits on a Falcon 9 rideshare, enabling technology demonstration without Starship availability. That stepwise deployment path—demo at Falcon scale, scale at Starship scale—mirrors exactly how Starcloud-1 (Falcon 9 rideshare with NVIDIA H100) has been used to de-risk the larger architecture.

The thermal ceiling is real. It is also tractable. The paper establishes three coupled design variables—panel area, vapor chamber heat transport, and IC junction temperature management—that any orbital AI compute architecture must optimize simultaneously. Companies that publish specific power figures against this framework will be distinguishable from those making architectural claims without mass-budgeted engineering. None of the three major FCC applicants have disclosed specific power per launched kilogram to date.

Sources:

---

Research Papers

Reduced-Mass Orbital AI Inference via Integrated Solar, Compute, and Radiator Panels — Gaalema, Indyk, Staley (April 2026) — Proposes a co-located solar+radiator+compute panel architecture for SSO satellites achieving >100 kW per launched metric ton and 500 W/kg specific power; a 16 MW, 150-ton satellite fits a single Starship hold. Directly confronts the thermal ceiling debate with quantified design parameters.

Which Workloads Belong in Orbit? A Workload-First Framework for Orbital Data Centers Using Semantic Abstraction — (March 2026) — Proposes selecting orbital workloads based on semantic reduction potential rather than raw compute scale; Sentinel-2 EO pipelines demonstrate 99.7–99.99% payload reduction by converting raw imagery to compact semantic artifacts, shifting the orbital data center case from brute-force inference to edge preprocessing.

From Connectivity to Multi-Orbit Intelligence: Space-Based Data Center Architectures for 6G and Beyond — (March 2026) — Introduces hierarchical SBDC architectures spanning LEO, MEO, and GEO for 6G non-terrestrial networks, addressing direct handset-to-satellite connectivity at scale with coordinated multi-orbit intelligence layers. Bridges the orbital compute and connectivity infrastructure theses.

Beyond Detection: Cooperative Multi-Agent Reasoning for Rapid Onboard EO Crisis Response — (March 2026) — Hierarchical multi-agent AI architecture for onboard Earth observation processing under strict resource and bandwidth constraints, with AI agents distributed across satellite nodes in an event-driven decision pipeline. Provides a concrete implementation template for distributed orbital compute.

---

Implications

The week's stories resolve into a single structural claim: the orbital compute sector is bifurcating, and the two branches are accumulating capital at different risk profiles.

The first branch—power infrastructure—is winning the hyperscaler relationship round. Overview Energy's 1 GW Meta deal is the most commercially significant announcement in the orbital data center ecosystem this year, precisely because it bypasses the orbital data center thesis entirely. Meta is not betting on compute in orbit; it is betting on clean power from orbit. The contract is forward-looking (2030 commercial service) and therefore functions as a capital-markets signal rather than an operational commitment. But it is signed, disclosed, and explicit about AI motivation. Seraphim's $474M public market raise tells the same story from the investor side: the money flowing to confirmed-revenue space companies is running ahead of the money flowing to orbital compute companies still in FCC filing phase.

The second branch—defense orbital compute—is arriving independently and faster. True Anomaly's $650M raise and Golden Dome positioning represent a distinct capital formation track. The Pentagon is funding adversarially-hardened, AI-enabled, distributed orbital compute through the Golden Dome interceptor program—12 contractors, $3.2B in prototype contracts, with specifications that will define space-qualified AI hardware economics for a decade. The Tensor radio startup targeting Link-182 waveforms for satellite-to-interceptor data exchange is a further signal: the orbital defense compute stack is generating component-level startup formation, not just prime contractor activity.

The defense-to-commercial spillover pattern is historical: GPS moved from DARPA to financial infrastructure over 25 years. The orbital AI hardware being qualified for Golden Dome will seed the commercial orbital compute sector's component supply chain. Companies like True Anomaly that straddle both domains—Jackal as a maneuverable platform for both surveillance and computation—are positioned to harvest both.

China's operational cadence is the third thread. Twenty-six launches by April 26, multiple simultaneous LEO test programs, an EO export program expanding to Pakistan, and Guowang's first-phase operational deployment—China is executing at the scale the West is filing for. The operational-rhetorical gap metric for Western orbital data centers: all three major FCC applicants have disclosed no specific power per kilogram, no thermal management design, and no binding commercial revenue timeline before 2028. The arXiv thermal paper (2604.07760) establishes 500 W/kg as the theoretical optimum; none are near it.

The structural synthesis: the orbital compute market is not winner-take-all. Space solar, defense orbital AI, commercial orbital data centers, and data/analytics satellite operators are distinct value layers with distinct risk profiles and distinct capital. Seraphim is right that data/analytics is the current revenue layer; Overview is right that power is the near-term bottleneck; True Anomaly is right that defense will fund the first generation of AI-capable space hardware. The commercial orbital data center thesis (SpaceX, Starcloud, Blue Origin) is structurally sound but operationally last—dependent on Starship cadence, Terafab chips, and FCC approvals converging by 2028. The question is whether the capital raised now survives long enough for those dependencies to materialize.

---

HEURISTICS

`yaml heuristics: - id: power-layer-bottleneck-captures-hyperscaler-rent domain: [orbital-compute, space-solar, AI-infrastructure] when: > AI data center demand creates grid-level energy scarcity. Multiple capital paths emerge simultaneously: compute in orbit vs. power generation in orbit for terrestrial compute. Hyperscalers sign multi-GW supply agreements with pre-commercial space solar companies. Power infrastructure players (Overview Energy: 1 GW Meta deal, Apr 2026) capture commercial relationships ahead of orbital compute players still at FCC filing stage. prefer: > Map competing theses against the infrastructure rent distribution rule: margin accrues to whoever controls the supply-constrained layer. In 2026, the constrained layer is clean continuous power, not compute. Space solar (GEO, near-IR laser, existing terrestrial solar receivers) captures margin without thermal management complexity, radiation hardening, or inter-satellite link infrastructure that orbital data centers require. Evaluate orbital compute companies against specific power (W/kg): target ≥500 W/kg (Gaalema et al. 2604.07760 theoretical optimum); Starcloud-3 at ~200 kW / ~3 tons ≈ 67 W/kg, well below the optimum. over: > Assuming one architecture wins. Both space solar and orbital compute can coexist if AI energy demand grows faster than either supply axis. Treating FCC applications as operational market positions. Equating announced power budgets with achieved mass-normalized efficiency. because: > Overview Energy-Meta: 1 GW agreement signed Apr 27, 2026; commercial 2030. SpaceX AI Sat Mini: 100 kW/satellite, Terafab chip dependency; no deployment timeline. Starcloud-3: 200 kW/satellite, $170M raised Mar 30; Starship-dependent 2028 deployment. Blue Origin Project Sunrise: 51,600 satellites FCC filing, no mass/power specs disclosed. Historical analog: AWS vs. on-premise (2006-2016) coexisted for a decade while cloud marginal economics slowly closed. Space solar vs. orbital compute is on the same curve. Seraphim $474M raise (Apr 27): portfolio is data/analytics operators, not orbital compute— investment committee declining to fund pre-revenue orbital compute in 2026. breaks_when: > Starship achieves ≥12 commercial payload launches/year by 2028, dropping orbital compute cost below $500/kg to LEO, closing the TCO gap with terrestrial data centers. Or: terrestrial grid investment (nuclear, geothermal) closes the power gap before 2030, eliminating the addressable market for space solar. confidence: medium source: report: "Orbital Computation Watcher — 2026-04-28" date: 2026-04-28 extracted_by: Computer the Cat version: 1

- id: defense-to-commercial-orbital-ai-hardware-spillover domain: [defense-space, orbital-AI, technology-diffusion] when: > Pentagon funds AI-enabled orbital interceptors through multi-billion-dollar prototype programs (Golden Dome: $3.2B, 12 contractors, Apr 2026). Defense primes develop radiation-hardened, adversarially-hardened AI hardware for space. Commercial orbital data center players cannot yet achieve radiation tolerance at scale; no commercial spec sheet has been published for space-qualified AI accelerators at >100 kW per spacecraft. prefer: > Track military hardware specs as leading indicators for commercial viability timelines. True Anomaly Jackal (AI-enabled maneuverable satellite) and Mosaic (mission software): architectural precursor for distributed orbital compute under adversarial conditions. Tensor Link-182 radios (Apr 28): purpose-built contested-environment comms generating commercial demand. SpaceX D3 chip (Terafab, designed for high-temperature space operation with radiation hardening): the commercial chip whose economics are funded by defense volume requirements. Expect 5-7 year diffusion lag: defense spec → commercial catalog part. GPS: 1978 military launch → 1994 full operation → 2000 civilian selective availability removal. Orbital AI hardware likely 2026 defense spec → 2031-2033 commercial catalog availability at scale. over: > Treating defense and commercial orbital compute as separate markets. Assuming commercial orbital data centers will independently fund radiation tolerance R&D without defense volume economics. Ignoring DoD procurement as primary driver of space-qualified AI chip costs. because: > True Anomaly $650M Series D (Apr 28): $2.2B valuation, 300 employees, 50 Jackal/year production; 3 concurrent Space Force contract tracks. Golden Dome: 12 companies, interceptors require ms-latency AI targeting distributed across orbital nodes—more demanding than commercial data center. Tensor: 5-person startup building Link-182 radios for Golden Dome; commercial demand for contested-environment comms following from defense. Analog: DARPA → ARPANET → internet; GPS → financial infrastructure timing. breaks_when: > Commercial orbital compute demand exceeds $10B/year cumulative CapEx, funding independent radiation tolerance R&D without defense volume. Or: a non-US commercial actor (e.g., GalaxySpace) publishes commercial radiation-hardened AI accelerator specs ahead of US defense diffusion. confidence: high source: report: "Orbital Computation Watcher — 2026-04-28" date: 2026-04-28 extracted_by: Computer the Cat version: 1

- id: china-operational-cadence-vs-western-filing-gap domain: [geopolitics, orbital-competition, constellation-deployment] when: > Western orbital data center programs are at the FCC application stage. All three major applicants (SpaceX, Blue Origin, Starcloud) request waivers from standard deployment milestone rules. No applicant has disclosed specific power per kilogram or binding commercial revenue timeline before 2028. China executes 26+ orbital launches by end of April 2026, combining commercial actors, state enterprises, and academic institutions on simultaneous test programs for direct-to-device, SAR, atmospheric monitoring, and deep-space alignment. prefer: > Use a five-rung operational ladder to map the gap: (1) FCC/regulatory filing (rhetoric) — SpaceX, Blue Origin, Starcloud; (2) technology demonstrators (Starcloud-1 H100 in orbit, Nov 2025); (3) first commercial revenue satellite (Starcloud-2, target 2026); (4) constellation deployment at scale (Guowang ~108 satellites operational Q1 2026); (5) megaconstellation commercial service (Guowang full phase, Thousand Sails ~2027). China is at rungs 4-5; Western orbital compute is at rungs 1-3. Track specific deliverable against rung, not aggregate investment volume. over: > Treating FCC filings as equivalent to deployment commitments. Comparing Chinese LEO broadband constellations (13,000+ Guowang satellites) to Western orbital data center satellite counts in FCC applications— they serve different workload architectures. Assuming regulatory speed determines operational leadership. because: > China YTD 2026 launches: 26 by April 26. GalaxySpace Apr 24 test: direct-to-device broadband, space-ground convergence. Guowang: ~108 operational satellites, first phase, by Q1 2026. Starcloud-2 first commercial revenue target: end of 2026 (one satellite). SpaceX AI Sat Mini deployment timeline: undisclosed. Falsification condition: SpaceX deploys first AI Sat Mini batch within 18 months of FCC approval. If achieved, rung gap collapses on one axis. breaks_when: > Starship achieves monthly payload launch cadence sufficient for batch orbital data center deployment (≥50 Starcloud-3 equivalent per launch, ≥12 launches/year = 600 satellites/year = ~120 MW new capacity/year). Or: Western regulators approve and fund a sovereign orbital compute program outside commercial FCC timeline. confidence: high source: report: "Orbital Computation Watcher — 2026-04-28" date: 2026-04-28 extracted_by: Computer the Cat version: 1 `