Metas orbitaler Solar-Wette: Drahtlose Energie für KI-Rechenzentren aus 35.000 km Entfernung

Hacker News April 2026
Source: Hacker NewsArchive: April 2026
Meta hat einen Vertrag über die Beschaffung von 1 GW weltraumgestützter Solarenergie und 100 GWh Langzeitspeicherung unterzeichnet, mit dem Ziel, Energie aus 35.000 km Höhe über der Erde drahtlos an seine KI-Rechenzentren zu senden. Dieser beispiellose Schritt könnte die Energiebasis der KI-Industrie neu definieren.
The article body is currently shown in English by default. You can generate the full version in this language on demand.

In a move that sounds like science fiction, Meta has committed to purchasing 1 gigawatt of orbital solar generation capacity paired with 100 gigawatt-hours of long-duration energy storage. The plan involves deploying solar collectors in geostationary orbit (GEO), approximately 35,000 kilometers above the Earth, where they can harvest sunlight 24/7 without atmospheric interference. This energy would then be converted into microwave or laser beams and transmitted wirelessly to ground-based rectennas connected to Meta's AI data centers. The core motivation is clear: the exponential growth in compute demand for training and inference of large language models is outpacing the availability of reliable, clean baseload power from terrestrial renewables, which suffer from intermittency and land-use constraints. By vertically integrating energy production into space, Meta aims to secure a stable, carbon-free power supply that can operate continuously, effectively bypassing grid limitations and fossil fuel backup. This deal represents the first large-scale commercial commitment to space-based solar power (SBSP) from a major technology company, signaling a potential paradigm shift in how the AI industry thinks about energy infrastructure. While the technical hurdles—including wireless transmission efficiency, orbital maintenance costs, and regulatory spectrum allocation—are immense, the strategic logic is compelling: in the AI arms race, energy sovereignty may be the ultimate competitive advantage.

Technical Deep Dive

Meta's orbital solar strategy is not a single technology but a complex system of three interdependent innovations: space-based photovoltaic generation, wireless power transmission (WPT), and long-duration energy storage on the ground.

Space-Based Generation: The solar arrays will be placed in geostationary orbit (GEO), at an altitude of ~35,786 km. Unlike low Earth orbit (LEO) satellites that experience day-night cycles and require tracking, GEO satellites remain fixed relative to a ground station, enabling continuous solar exposure for all but ~22 days per year (during equinox eclipse periods, lasting up to 72 minutes). The solar flux in GEO is approximately 1.36 kW/m², about 30% higher than the best terrestrial locations due to the absence of atmospheric absorption and cloud cover. To achieve 1 GW of continuous power output to the grid, the space segment would need to generate roughly 2-3 GW of raw DC power, accounting for conversion and transmission losses. This implies a solar array area on the order of 2-3 square kilometers—a structure far larger than the International Space Station. Assembly would likely require in-orbit robotic construction or multiple launches of modular panels.

Wireless Power Transmission (WPT): The most critical technical challenge is beaming power to Earth. Two primary approaches exist:
- Microwave WPT: Uses phased-array antennas to transmit a focused beam at frequencies around 2.45 GHz or 5.8 GHz (ISM bands). The ground receiver, a rectenna (rectifying antenna), converts the microwave energy back to DC electricity. Theoretical end-to-end efficiency is around 50-60%, but practical demonstrations have achieved 30-40%. The beam is inherently safe—power density at the receiver is designed to be below 10 mW/cm², comparable to sunlight. The advantage is all-weather operation, as microwaves penetrate clouds and precipitation.
- Laser WPT: Uses high-power infrared lasers (e.g., 1064 nm) directed at photovoltaic receivers on the ground. Efficiency can be higher (up to 60% in lab settings), but the beam is more susceptible to atmospheric scattering, cloud cover, and requires precise pointing. Safety concerns are greater due to potential eye and aircraft hazards.

Meta has not publicly disclosed which WPT method it will use, but the scale (1 GW) and the need for reliability suggest a microwave approach. A relevant open-source project is the Space Solar Power Demonstrator (SSPD) repository on GitHub (though not directly affiliated with Meta), which contains simulation models for phased-array beamforming and rectenna design. The repo has seen increased activity since the announcement, with over 2,000 stars as of late April 2026.

Long-Duration Storage: The 100 GWh of storage is a crucial buffer. Even with GEO's near-continuous sunlight, there are eclipse periods and potential transmission interruptions. This storage likely combines grid-scale lithium-ion batteries (for short-duration smoothing) with flow batteries or compressed air energy storage for multi-hour backup. The scale is immense—100 GWh is roughly equivalent to 10,000 Tesla Megapacks, representing a capital cost of $3-5 billion at current prices.

Data Table: Wireless Power Transmission Efficiency Comparison

| Method | Theoretical Efficiency | Demonstrated Efficiency | Atmospheric Loss | Safety Risk | Maturity Level |
|---|---|---|---|---|---|
| Microwave (2.45 GHz) | 60% | 35-40% | Low (<5%) | Low | TRL 6-7 (prototype) |
| Microwave (5.8 GHz) | 65% | 40-45% | Low (<5%) | Low | TRL 5-6 |
| Laser (1064 nm) | 70% | 50-60% | High (20-50% in clouds) | High | TRL 4-5 |
| Laser (1550 nm, eye-safe) | 60% | 40-50% | High (30-60% in clouds) | Moderate | TRL 4 |

Data Takeaway: Microwave WPT at 2.45 GHz offers the best balance of efficiency, reliability, and safety for baseload power delivery, but even at 40% efficiency, the space segment must generate 2.5 GW to deliver 1 GW to the data center. Laser systems are more efficient in clear conditions but are impractical for continuous operation in most climates.

Key Players & Case Studies

Meta is not operating in a vacuum. The SBSP ecosystem includes several key players:

- Emrod (New Zealand): A private company developing microwave WPT systems. They have demonstrated 2 kW transmission over 36 meters at 70% efficiency in a controlled environment. Emrod's technology uses a proprietary metamaterial-based rectenna that could scale to megawatt levels. Meta may be partnering with or licensing from Emrod.
- SpaceX (USA): As the dominant launch provider, SpaceX's Starship is the only vehicle capable of lifting the massive payloads required for GEO solar arrays at a reasonable cost ($10-20 million per launch for 100+ tons). Starship's fully reusable design reduces the cost per kilogram to orbit to under $100, a critical enabler for SBSP economics.
- Caltech's Space Solar Power Project (SSPP): In 2023, Caltech launched the MAPLE (Microwave Array for Power-transfer Low-orbit Experiment) prototype, which successfully beamed power in space and to Earth. The experiment demonstrated phased-array beam steering and rectenna operation. The project's lead, Dr. Ali Hajimiri, has published extensively on the technology. Caltech's work is foundational, though not directly commercial.
- Japan Aerospace Exploration Agency (JAXA): JAXA has been researching SBSP since the 1990s and demonstrated a 1.8 kW wireless power transmission over 55 meters in 2015. Their roadmap targets a 1 GW commercial system by 2050.
- China's BISSE (Beijing Institute of Space Science and Engineering): China has announced plans for a 1 MW orbital solar test station by 2030, with a 1 GW system by 2050. This represents a state-backed competitor to Meta's private initiative.

Data Table: SBSP Initiatives Comparison

| Entity | Target Capacity | Timeline | Investment (Est.) | Approach | Status |
|---|---|---|---|---|---|
| Meta | 1 GW | 2030-2035 (implied) | $10-20 billion | Private procurement | Deal signed, no hardware yet |
| China (BISSE) | 1 GW | 2050 | $5-10 billion (state) | Government-led | 1 MW test by 2030 |
| JAXA | 1 GW | 2050 | $2-5 billion (state) | Government-led | 1.8 kW demo completed |
| Caltech/SSPP | N/A (research) | N/A | $100 million (donor) | Academic | MAPLE demo in orbit |
| Emrod | Commercial systems | 2025-2030 | $50 million (VC) | Private | 2 kW demo completed |

Data Takeaway: Meta's commitment dwarfs all other private and public initiatives in scale and speed. While China and Japan have longer-term government roadmaps, Meta's willingness to sign a commercial offtake agreement for 1 GW signals a belief that the technology can be deployed within a decade, not by 2050. This is a high-risk, high-reward bet.

Industry Impact & Market Dynamics

Meta's move has immediate and profound implications for the AI, energy, and space industries.

AI Industry: The primary driver is the insatiable energy appetite of AI. Training a model like GPT-4 is estimated to consume 50-100 GWh, and inference costs are growing even faster. By 2030, AI could consume 10-20% of global electricity. Terrestrial renewables cannot provide the 99.999% uptime required for hyperscale data centers without massive overbuilding and storage. SBSP offers a path to true baseload renewable power. This could trigger a wave of similar deals from other hyperscalers—Google, Microsoft, Amazon, and Oracle—each of which has made net-zero pledges but faces the same intermittency problem. The competitive dynamic shifts: companies that secure orbital energy capacity gain a structural cost advantage in compute.

Energy Market: If successful, SBSP could disrupt the traditional utility model. Meta is effectively bypassing the grid, building its own power plant in space. This vertical integration could reduce reliance on natural gas peaker plants and grid interconnection queues, which currently delay data center construction by 3-5 years. The cost of SBSP is currently estimated at $0.10-0.20 per kWh, compared to $0.03-0.05 for terrestrial solar plus storage. However, as launch costs fall with Starship, SBSP could reach parity by 2035. The 100 GWh storage requirement also creates a massive market for long-duration energy storage technologies, benefiting companies like Form Energy (iron-air batteries) and Malta (pumped heat storage).

Space Industry: This is a catalyst for the in-space servicing and manufacturing sector. Building a 2-3 km² solar array in GEO requires autonomous robotic assembly, orbital tugs, and heavy-lift launch. Companies like SpaceX (Starship), Blue Origin (New Glenn), and Relativity Space (Terran R) will compete for launch contracts. The demand for launch capacity could double the current global launch market within a decade.

Data Table: Projected Cost Comparison for AI Data Center Power (2035)

| Power Source | Levelized Cost ($/MWh) | Uptime | Carbon Intensity | Land Use (acres/MW) |
|---|---|---|---|---|
| Terrestrial Solar + Storage | $40-60 | 70-80% | Zero | 5-10 |
| Onshore Wind + Storage | $30-50 | 40-50% | Zero | 10-20 |
| Natural Gas (combined cycle) | $40-60 | 95% | High | 0.5 |
| Nuclear (SMR) | $100-150 | 95% | Zero | 0.1 |
| Space-Based Solar (SBSP) | $80-120 | 99% | Zero | 0.01 (ground) |

Data Takeaway: SBSP is projected to be more expensive than terrestrial renewables but cheaper than nuclear and competitive with natural gas when carbon costs are factored in. Its key advantage is 99% uptime, which is critical for AI training workloads that cannot tolerate interruptions. The land use advantage is also significant for data center site selection.

Risks, Limitations & Open Questions

Despite the bold vision, the path to operational SBSP is fraught with challenges:

1. Wireless Power Efficiency: The end-to-end efficiency from sunlight to data center plug is likely 15-25% after accounting for solar panel efficiency (30%), WPT efficiency (40%), rectenna conversion (80%), and storage cycling (90%). This means the space array must be 4-6 times larger than the delivered power suggests, driving up launch costs.

2. Orbital Debris and Space Traffic: A 2 km² structure in GEO is a massive target for micrometeoroids and space debris. While GEO is less congested than LEO, the risk of catastrophic damage from a single impact event is non-trivial. The structure must be modular and repairable, requiring a fleet of orbital service vehicles.

3. Regulatory and Spectrum Allocation: Wireless power transmission at high power levels requires allocation of radio frequencies by the International Telecommunication Union (ITU). Interference with existing satellite communications, radar, and radio astronomy is a real concern. Meta will need to negotiate frequency bands and power limits, a process that could take years and face opposition from incumbent users.

4. Geopolitical Risks: A space-based power station is a strategic asset. It could be a target for cyberattack or physical attack (e.g., anti-satellite weapons). The technology could also be dual-use—the same beam could theoretically be used as a directed-energy weapon. International treaties on the peaceful use of outer space may need reinterpretation.

5. Economic Viability: The upfront capital cost is staggering—estimated at $10-20 billion for the first 1 GW system. Meta's market cap is over $1 trillion, so it can afford the bet, but the return on investment depends on launch costs falling faster than expected. If Starship fails to achieve its cost targets, the economics collapse.

6. Environmental Concerns: The microwave beam, while low-intensity, could affect bird migration, aircraft avionics, and local ecosystems near the rectenna site. Long-term health effects of chronic low-level microwave exposure are not fully understood.

AINews Verdict & Predictions

Meta's orbital solar deal is the most audacious energy bet in the history of the technology industry. It is not a sure thing—the technical and regulatory hurdles are immense, and the timeline is likely longer than Meta's public posture suggests. However, the strategic rationale is unassailable. AI compute demand is growing at 10x per year, and terrestrial renewable energy cannot scale to meet it without massive overbuilding and grid upgrades that are politically and physically constrained.

Our predictions:

1. By 2028, Meta will announce a 100 MW pilot system in GEO, likely using a partnership with SpaceX for launch and Emrod for WPT. This will serve as a proof-of-concept for a single data center cluster.

2. By 2032, Google and Microsoft will announce similar SBSP offtake agreements, triggering a gold rush in orbital energy. The market for SBSP will be valued at $50 billion annually by 2035.

3. The biggest bottleneck will not be technology but regulation. The ITU will struggle to allocate spectrum for multiple GW-scale power beams, leading to a new international framework for space-based energy transmission. Countries with equatorial ground station locations (e.g., Brazil, Indonesia, Kenya) will become strategic energy hubs.

4. The cost of SBSP will fall faster than most analysts predict because the launch cost curve is steeper than expected. Starship's fully reusable design will push launch costs below $50/kg by 2030, making SBSP cheaper than nuclear and competitive with natural gas.

5. The ultimate winner is not Meta, but the AI industry as a whole. By proving the concept, Meta will unlock a new energy paradigm that allows AI to scale without planetary boundaries. The era of terrestrial energy constraints on compute is ending. The next frontier is not just in the cloud—it is in the stars.

More from Hacker News

VibeLens: Das Open-Source-„Geistesmikroskop“, das KI-Agenten-Entscheidungen transparent machtThe rise of autonomous AI agents—systems that plan, use tools, and execute multi-step tasks—has introduced a critical prClaude Codes versteckter 'OpenClaw'-Auslöser: Ihr Git-Verlauf bestimmt jetzt die API-PreiseAn investigation by AINews has identified a secret trigger mechanism within Anthropic's Claude Code, an AI-powered codinAgent-Recall-AI: Der Checkpoint-Retter, der KI-Agenten unternehmenstauglich machen könnteThe promise of autonomous AI agents has long been overshadowed by their brittleness. When an agent is tasked with a multOpen source hub2705 indexed articles from Hacker News

Archive

April 20263011 published articles

Further Reading

Meta- und AWS-Graviton-Deal signalisiert das Ende der reinen GPU-basierten KI-InferenzMeta und AWS haben einen mehrjährigen Vertrag unterzeichnet, um Llama-Modelle und zukünftige agentische KI-Workloads aufVibeLens: Das Open-Source-„Geistesmikroskop“, das KI-Agenten-Entscheidungen transparent machtEin neues Open-Source-Tool namens VibeLens bietet eine interaktive Echtzeit-Visualisierung der KI-Argumentation und verwAgent-Recall-AI: Der Checkpoint-Retter, der KI-Agenten unternehmenstauglich machen könnteKI-Agenten haben einen fatalen Fehler: Sie sterben mitten in einer Aufgabe. Ein neues Open-Source-Tool, agent-recall-AI,Die Überengineering-Falle: Warum Einfachheit die ultimative Weisheit für KI-Backends istEin offener Rückblick eines Entwicklers zeigt, wie das Stapeln fortschrittlicher Komponenten wie Distillation, Routing u

常见问题

这次公司发布“Meta's Orbital Solar Bet: Wireless Power for AI Data Centers from 35,000km”主要讲了什么?

In a move that sounds like science fiction, Meta has committed to purchasing 1 gigawatt of orbital solar generation capacity paired with 100 gigawatt-hours of long-duration energy…

从“Meta orbital solar power deal details”看,这家公司的这次发布为什么值得关注?

Meta's orbital solar strategy is not a single technology but a complex system of three interdependent innovations: space-based photovoltaic generation, wireless power transmission (WPT), and long-duration energy storage…

围绕“How does space-based solar power work for AI data centers”,这次发布可能带来哪些后续影响?

后续通常要继续观察用户增长、产品渗透率、生态合作、竞品应对以及资本市场和开发者社区的反馈。