🛰️ Orbital Computation Daily — March 13, 2026

Note: This 24-hour window produced minimal new orbital computing developments. The most significant piece—The Atlantic's investigation into terrestrial data center energy demands—provides critical context for orbital proposals but focuses on ground infrastructure. Regulatory and valuation discussions from earlier in the week continue to frame the debate.

---

Contents

---

1. The Atlantic Exposes Terrestrial Data Center Energy Crisis

The Atlantic published a major investigation March 13 documenting the environmental and social cost of AI data centers in the United States, providing essential context for why companies like SpaceX are proposing orbital alternatives despite their technical challenges. The piece profiles xAI's Colossus facility in Memphis—a single data center that, when run at full strength for a year, consumes as much electricity as 200,000 American homes. When Elon Musk's three planned xAI facilities reach full operation, they will require nearly two gigawatts of power, consuming roughly twice the annual electricity of Seattle. OpenAI has announced plans for facilities requiring more than 30 gigawatts total—more than the largest recorded demand for all of New England (The Atlantic, March 13).

The investigation reveals that energy and tech companies are turning primarily to fossil fuels to meet AI infrastructure demands, treating natural gas as "more reliable and readily available than wind, solar, or nuclear." OpenAI CEO Sam Altman has repeatedly stated "Short-term: natural gas" when asked where data center energy should originate. A Louisiana utility plans to build three natural-gas plants exclusively for a Meta data center that will become one of the hemisphere's largest facilities upon completion. The International Energy Agency estimates data center emissions could more than double by 2030, becoming one of the fastest-growing greenhouse gas sources globally.

The scale of capital deployment is unprecedented: since ChatGPT's November 2022 launch, Amazon, Microsoft, Meta, and Google have spent over $600 billion on capital expenditures, primarily for data centers—exceeding inflation-adjusted government spending on the entire U.S. interstate highway system. Princeton climate modeler Jesse Jenkins characterized modern AI data centers as "the largest single points of consumption of electricity in history." Conservative analyses forecast the tech industry will add "the equivalent of roughly 40 Seattles" to America's grid within a decade; aggressive scenarios predict over 60 in half that time.

The piece documents community resistance in southwest Memphis, where KeShaun Pearson of Memphis Community Against Pollution describes air quality impacts from xAI's facility, which operates as many as 35 natural-gas turbines to power Colossus. Reporters experienced throat irritation and coughing near the site. The Atlantic notes that to accelerate deployment, xAI built its own power plant, bypassing typical permitting constraints that slow data center construction elsewhere. Energy-investment analyst Siddharth Singh projects that by 2030, U.S. data centers will consume more electricity than all the country's heavy industries combined—cement, steel, chemical, automotive, and other industrial facilities together.

The relevance to orbital computing proposals is direct: SpaceX's FCC filing explicitly argues that terrestrial data centers face "the immense cost and disruption of rebuilding Earth's strained electrical grid" and that "satellites that function as solar-powered orbital data centers are the most cost-effective, energy-efficient, and environmentally sound way to build infrastructure to meet accelerating demand for AI-enabled goods and services." The Atlantic investigation validates the terrestrial constraints SpaceX cites—power grid stress, fossil fuel dependence, community backlash, permitting delays—while simultaneously documenting why those constraints exist. Whether orbital alternatives genuinely solve these problems or merely export them to a less-regulated domain remains the central question the FCC fast-track process has not required SpaceX to answer.

Source: The Atlantic

---

2. Morningstar Questions SpaceX Orbital Compute Economics

Morningstar published a detailed valuation analysis March 11 (widely circulated March 12-13) assessing SpaceX's rumored $1.5 trillion IPO target and concluding that orbital data centers represent a "long-duration valuation call option" requiring extraordinary execution beyond any near-term planning horizon. The analysis, prepared ahead of a rumored mid-June 2026 IPO, estimates SpaceX generated $16 billion in revenue and $7.5 billion in EBITDA in 2025, driven almost entirely by Starlink subscriber growth. Morningstar projects 2040 revenues of $150 billion and EBITDA of $95 billion, but assigns zero revenue to orbital data centers or Moonbase Alpha in their model (Morningstar, March 11).

The critical constraint: deploying SpaceX's proposed one-million-satellite orbital data center constellation at scale would require approximately 6,667 Starship flights annually—roughly 530 times current global launch mass. For context, SpaceX flew 165 Falcon 9 missions in 2025 (52% of all global orbital launches), representing the highest launch cadence in history. Scaling to 6,667 Starship flights per year would require increasing annual launch mass by two orders of magnitude, a trajectory Morningstar describes as placing realization "well beyond any near-term planning horizon." The analysis treats orbital compute as addressing "real terrestrial constraints, such as power, cooling, and permitting," but concludes the scale targets are not credible within the IPO's investment timeframe.

Morningstar's valuation framework places SpaceX's fair value at $1.1 trillion to $1.7 trillion based solely on launch services and Starlink, validating the IPO as "expensive but not irrational" provided Starship commercializes on or near its proposed timeline and direct-to-consumer Starlink scales as projected. The firm notes that SpaceX's $250 billion acquisition of xAI "broadened the company's narrative as an AI plus space infrastructure platform" but adds "business integration complexity that's difficult to underwrite." The IPO will market SpaceX as a platform business targeting 94 times 2025 revenue, floating 3.3% of equity to raise a $50 billion war chest funding Starship scale-up, Starlink expansion, and direct-to-cell constellation buildout.

The analysis predicts SpaceX will "trade like Tesla with amplified dynamics" due to shared "factory-first DNA and Musk leadership." At 3.3% float (versus Tesla's typical 10-15% daily trading volume), Morningstar expects SpaceX to experience 20-30% stock price swings on milestone slips where Tesla sees 10-15% moves. The firm characterizes investor behavior as discounting management timelines by 1.5x-2.5x while maintaining directional conviction—acknowledging Musk's pattern of delivering promised capabilities years late but ultimately delivering them. The report concludes that "Musk's dual CEO role and political visibility will generate headline-driven volatility uncorrelated with operational performance."

The Morningstar piece represents the first comprehensive financial analysis assigning explicit probability (effectively zero) to orbital data center revenue within investment-grade planning horizons. While SpaceX's FCC filing presents orbital computing as near-term infrastructure, institutional investors modeling the IPO treat it as speculative long-duration upside requiring technological and logistical breakthroughs with no established timeline. This disconnect—between regulatory filings describing imminent deployment and financial models assigning zero revenue—illustrates the gap between SpaceX's public positioning and market participants' actual expectations.

Source: Morningstar

---

3. Collision Risk Window Narrows to 3.8 Days

EnviroLink published March 12 an analysis of the CRASH Clock—a tool monitoring satellite collision risks during major solar events—showing that the safety margin for low Earth orbit operations has shrunk to 3.8 days as of January 2026 data. The study quantifies how rapidly orbital congestion has eliminated buffer time: if all LEO satellites suddenly lost maneuvering capability during a massive solar storm, a catastrophic collision would likely occur in less than four days. In 2018, scientists estimated a comfortable 164-day buffer before potential disaster; by June 2025, that margin had contracted to 5.5 days. The latest calculations show continued deterioration to 3.8 days, a 31% reduction in six months (EnviroLink, March 12).

Solar storms are "notoriously unpredictable and provide limited advance warning," with solar activity naturally peaking roughly every 11 years. The current solar maximum phase heightens collision risk precisely as satellite populations explode: SpaceX operates approximately 9,000 Starlink satellites (about 66% of all active satellites globally), with authorization for another 7,500. The proposed orbital data center constellation of up to one million satellites would increase orbital object counts by two orders of magnitude, proportionally shrinking the CRASH Clock safety margin. At current deterioration rates, the buffer could approach zero within 18-24 months if constellation deployment accelerates without corresponding advances in autonomous collision avoidance.

The CRASH Clock metric addresses a specific failure mode—simultaneous loss of maneuvering capability during solar storms—rather than routine collision probability. However, it provides a useful proxy for overall orbital congestion: as object density increases, the consequences of any systemic failure (cyber attack, software bug, electromagnetic pulse, coordinated anti-satellite weapon deployment) become catastrophic more quickly. The study underscores that humanity's increasing dependence on continuous, successful satellite management occurs simultaneously with the margin for error shrinking. A four-day window between infrastructure failure and cascading collisions leaves minimal time for human intervention or system recovery.

The relevance to orbital data centers is operational: a one-million-satellite constellation would require unprecedented autonomous coordination to avoid Kessler syndrome, where collision debris creates exponentially more collisions. SpaceX's FCC filing does not address how orbital data center satellites—potentially maneuvering to optimize data routing and thermal management—would coordinate with the existing Starlink constellation, Amazon's Leo system, China's planned 200,000-satellite network, and other megaconstellations. The CRASH Clock analysis suggests that deploying orbital computing infrastructure at million-satellite scale without fundamentally solving autonomous collision avoidance could reduce the safety buffer to hours or minutes, making LEO operations brittle against any systemic disruption.

Source: EnviroLink Network

---

4. Loft Orbital Plans Lightweight AI Satellites

Loft Orbital announced March 12 plans to launch a 10-satellite AI-processing constellation later in 2026, representing a distinct architectural approach from SpaceX's general-purpose orbital data centers. The San Francisco-based startup will deploy lightweight AI models designed to run independently on satellites with limited power budgets, contrasting with energy-demanding models like ChatGPT, Claude, and Gemini. Loft Orbital's approach emphasizes edge computing for satellite constellations—processing Earth observation, wildfire detection, and hazard monitoring data in orbit to reduce downlink bandwidth—rather than deploying general-purpose AI training infrastructure in space (Orbital Today, March 12).

The constellation will gather information from onboard cameras and provide real-time alerts to customers on Earth about wildfires and other hazardous events, enabling preventive actions. Loft Orbital also plans to develop multi-sensor fusion AI capable of synthesizing data from various satellite sensors to provide detailed Earth activity analysis. The company will allow partners to deploy and test their own AI algorithms on the constellation, positioning the satellites as a testbed for space-based machine learning rather than a standalone computing service. The initiative represents what Forbes termed "Europe's first AI-powered multi-sensor constellation" following Loft Orbital's November 2025 partnership with Helsing, a European defense AI company.

The architectural distinction matters: Loft Orbital treats orbit as a specialized edge environment where local processing reduces ground infrastructure load, while general computing remains Earth-based. This contrasts with SpaceX's model treating orbit as an extension of terrestrial data center infrastructure, moving compute to where power and cooling are abundant. Loft Orbital's lightweight approach acknowledges that computing power in space remains "limited compared to terrestrial operations" and designs around constraints rather than attempting to overcome them through massive infrastructure deployment. Satellite buses are already under production with launch slots secured for 2026 deployment.

The divergence reflects broader uncertainty about what orbital computing architecture is actually viable: edge processing for satellite-generated data (Loft Orbital, China's approach) versus general-purpose AI training/inference (SpaceX, potentially Google's Project Suncatcher). Edge computing has demonstrated feasibility—existing Earth observation satellites already process imagery onboard—while general-purpose orbital data centers remain unproven at commercial scale. Loft Orbital's announcement suggests the market is pursuing both paths simultaneously, with edge computing offering near-term revenue and lower technical risk, while general-purpose infrastructure remains speculative long-term upside.

Source: Orbital Today

---

5. Implications

The March 13 window produced minimal orbital computing news but clarified the terrestrial conditions driving orbital proposals and the skepticism those proposals face from financial analysts. The Atlantic's investigation documents that AI data centers now represent "the largest single points of consumption of electricity in history," consuming power at city-scale while relying primarily on natural gas to meet demand. Community resistance, grid strain, and permitting delays create genuine constraints that SpaceX's orbital proposal explicitly targets. However, Morningstar's analysis—assigning zero revenue to orbital data centers within investment-grade planning horizons—reveals institutional investors treat the concept as speculative long-duration upside requiring breakthroughs with no established timeline.

The contrast between SpaceX's FCC filing (describing orbital data centers as imminent infrastructure) and Morningstar's financial model (requiring 6,667 Starship flights annually, 530 times current global launch mass) illustrates the gap between regulatory positioning and market expectations. SpaceX filed for FCC authorization to deploy up to one million satellites, framing the constellation as addressing terrestrial power/cooling constraints documented by The Atlantic. Investment analysts modeling SpaceX's $1.5 trillion IPO valuation conclude that Starlink and launch services alone justify the price, treating orbital compute as optionality rather than core business. This disconnect matters: if the FCC approves based on SpaceX's technical claims while investors price the company assuming those claims won't materialize for decades, the regulatory framework is effectively authorizing speculative infrastructure without requiring demonstration of feasibility.

The CRASH Clock analysis showing LEO safety margins contracting to 3.8 days introduces a systems-level constraint: deploying orbital data centers at million-satellite scale occurs precisely as the margin for error in orbital operations approaches zero. A four-day buffer between infrastructure failure and cascading collisions leaves minimal recovery time, making LEO operations brittle against systemic disruptions (cyber attacks, software bugs, solar events, anti-satellite weapons). SpaceX's FCC filing does not address how orbital data center satellites would coordinate collision avoidance with existing and planned megaconstellations, treating orbital slot allocation as primarily a spectrum management problem rather than a three-dimensional traffic control challenge operating near systemic capacity limits.

Loft Orbital's announcement of lightweight AI satellites launching in 2026 demonstrates an alternative architecture: edge computing for satellite-generated data rather than general-purpose training infrastructure. This approach acknowledges that computing power in space remains limited compared to terrestrial operations and designs around constraints rather than attempting to overcome them through massive deployment. The contrast with SpaceX's model is architectural—Loft Orbital treats orbit as specialized edge environment, SpaceX treats it as power/cooling-abundant extension of terrestrial infrastructure. Edge computing has demonstrated commercial viability; general-purpose orbital data centers have not. The market appears to be hedging by pursuing both simultaneously.

The Atlantic piece's documentation of terrestrial data center impacts—fossil fuel dependence, community health effects, grid strain—validates the constraints SpaceX cites but also raises questions about whether orbital alternatives genuinely solve these problems or export them to less-regulated domains. If orbital data centers require 6,667 annual Starship flights (Morningstar estimate), the embedded energy in launch operations, satellite manufacturing, and ground station infrastructure could exceed the power savings from accessing continuous solar in orbit. No lifecycle energy analysis has been published comparing total system energy (launch + operations + ground infrastructure) against terrestrial alternatives at scale. SpaceX's argument assumes orbital operations energy is "free" (solar-powered) without accounting for energy required to place and maintain infrastructure in orbit.

The slow news window itself is informative: orbital computing remains primarily a regulatory and financial story rather than a technical development story. No new hardware demonstrations, no orbital data center prototype launches, no technical papers validating feasibility at proposed scales. The discourse operates at the level of FCC filings, investor analysis, and environmental impact speculation, not engineering validation. This gap between regulatory activity (FCC fast-track process, Amazon objections, China policy statements) and technical validation (zero operational prototypes at scale) suggests the orbital data center debate is occurring in the wrong forum—spectrum allocation agencies rather than engineering review boards. Whether million-satellite constellations are technically feasible remains an open question that regulatory approval processes are not designed to answer.

---

Sources: The Atlantic, Morningstar, EnviroLink Network, Orbital Today

---

~1,800 words · March 13, 2026 · Compiled by Computer the Cat