SpaceX Targets 100,000 Satellites With AI Compute

Brighter with Herbert highlighted that Starship's next flight is scheduled for Monday, while Starlink now supports direct connections for regular iPhones across four major carriers.

The FCC is currently reviewing SpaceX's proposal for a 100,000-satellite network, each potentially housing an Nvidia supercomputer. Concurrently, the production of Starlink V5 terminals and V2 mobile hardware has commenced, indicating a significant scale-up in ground and space infrastructure.

SpaceX recently completed a wet dress rehearsal for Starship, showcasing remarkable efficiency in propellant loading. According to Larry Goldberg, the ship and booster were fully loaded with liquid oxygen and methane in approximately 30 minutes flat.

This rapid refueling time is notable as it is faster than the process required to fill a Falcon 9, indicating significant operational improvements for future Starship missions.

Flight 14 represents a pivotal moment for SpaceX as its first commercial mission carrying Starlink V3 satellites, shifting from a focus on testing to operational deployment. Larry Goldberg pointed out that this milestone arrived earlier than many industry predictions, which had anticipated such launches toward the end of the year or early next year.

Internal scheduling further reinforces this transition, with Flights 15 and 16 tentatively planned for October, underscoring SpaceX's accelerated push into revenue-generating space operations.

SpaceX is planning an aggressive launch schedule, with tentative plans for Flight 15 on October 19 and Flight 16 on October 30. Larry Goldberg noted that the October 30 flight is slated to carry Starlink 30-1 to orbit.

This rapid cadence illustrates an unprecedented speed in commercial aerospace operations, pushing toward SpaceX's ambitious goal of achieving one launch per day by the end of the year.

Such a schedule, if realized, would signify a transformative shift in the frequency and scale of orbital deployments, enabling rapid expansion of the Starlink constellation and other payloads.

SpaceX is rapidly accelerating Starship production and testing schedules, with hardware for upcoming flights nearing completion. The ship designated for Flight 15 has already undergone its initial rounds of testing, indicating a streamlined development process.

Larry Goldberg confirmed that the current production rate is exceeding projections for this year, with Flight 16 tentatively scheduled for no earlier than October 30, underscoring the company's commitment to swift progress.

Flight 16 will mark the start of scheduled commercial missions tasked with carrying full payloads of Starlink satellites, a milestone the company CFO highlighted while confirming a $100 million revenue run rate for the service. Larry Goldberg highlighted that Flight 16 will officially mark the beginning of scheduled commercial flights, with an objective to carry full payloads of Starlink satellites.

The shift underscores a move from experimental flights to operational deployment, with the SpaceX CFO confirming that Starlink is already generating a $100 million revenue run rate, demonstrating its commercial viability.

Humanoid robots and autonomous vehicles will soon outpace human users in data consumption, the SpaceX CFO argued, necessitating the high-bandwidth orbital compute constellations enabled by Starship. The increasing reliance on AI, particularly with humanoid robots, autonomous vehicles, and machine-to-machine connections, necessitates unprecedented levels of data flow.

Starship’s substantial payload capacity is critical for deploying next-generation direct-to-cell services and orbital compute constellations, providing the necessary infrastructure for these advanced applications.

SpaceX plans to configure satellites with enhanced solar power and radiators to support the massive energy and cooling requirements of orbital computing, enabling the deployment of AI in space.

Constant, reliable internet via direct-to-cell technology is increasingly vital, as existing cellular coverage often fails to support the data-intensive AI models and autonomous systems integrated into modern vehicles. Larry Goldberg pointed out that current cellular coverage is inconsistent, with vehicles frequently driving in and out of serviced areas.

Direct-to-cell satellite technology addresses this by ensuring constant, reliable internet for mobile devices, which is critical for the integration and functionality of AI models like Grok, as they depend on robust, always-on data connections.

Standard mobile phones gain connectivity in remote regions where traditional cell towers cannot reach, though it remains unclear if existing bandwidth can support traditional voice calls over the current data-focused infrastructure. The service is poised to extend network access to previously underserved regions, enhancing global communication.

However, questions remain regarding the service's capabilities, specifically whether it will support traditional voice calls or be limited to data. It is currently unclear if dial-up style voice capabilities can function effectively over the existing bandwidth, raising discussions about the full scope of the direct-to-cell feature.

T-Mobile has previously expressed skepticism regarding the technical feasibility of satellite-to-phone connectivity. Larry Goldberg noted that T-Mobile specifically raised concerns about signal penetration through walls, roofs, and interior environments, highlighting significant technical hurdles.

Despite this industry skepticism, SpaceX leadership remains confident in its mobile integration plans, with the company anticipating the launch of direct-to-cell mobile services next year, aiming to overcome the challenges posed by traditional cellular limitations.

Gen 3 satellites may feature onboard Nvidia Blackwell NVL72 computer racks to achieve 100+ terabits of connectivity, a strategy Elon Musk underscored as a method to combine Starlink bandwidth with Star Mind AI compute power. Elon Musk indicated that Gen 3 satellites could achieve over 100 terabits of connectivity, and might feature onboard Nvidia Blackwell NVL72 computer racks.

This integration suggests a potential fusion of Starlink's connectivity capabilities with a 'Star Mind' AI compute power, creating a hybrid platform. The approach aligns with Musk's philosophy of maximizing resource efficiency by consolidating multiple high-function capabilities onto a single satellite platform.

This combination aims to establish a robust and intelligent orbital network capable of both high-bandwidth communication and advanced data processing, directly from space.

Star Mind and Starlink architectures differ, with the former requiring specific sun-synchronous orbits for constant energy and the latter maintaining less critical altitude requirements for its communication tasks. Larry Goldberg clarified that while the original announcement for Star Mind envisioned a full rack of compute hardware, a single Vera Rubin chip in Starlink provides significantly less power.

Star Mind requires specific sun-synchronous orbits for continuous energy supply, unlike Starlink, which prioritizes communication latency over precise orbital height for its compute requirements. These differences highlight the specialized nature of each system, catering to unique functional demands within the space infrastructure.

SpaceX is exploring the potential for onboard AI in Starlink satellites to enhance various functions. Larry Goldberg noted that AI could handle navigation, junk avoidance, and orbital station-keeping, which are currently managed by multiple chips and legacy code.

However, combining Star Mind's high energy and cooling requirements with Starlink satellites remains impractical for now. This indicates a phased approach to AI integration, prioritizing foundational tasks before more complex orbital computing applications.

Performing significant inference via a single chip network remains complex because satellites only stay in view for 10-15 minutes, forcing constant data handoffs. Larry Goldberg expressed skepticism, explaining that distributing a compute job across satellites passing overhead is highly inefficient.

Satellites are typically in view for only 10-15 minutes, complicating continuous processing. This limited visibility makes it difficult to perform significant inference tasks effectively with a single chip network, raising questions about the feasibility of distributed orbital computation at scale.

The Starlink V5 terminal delivers speeds up to 375 Mbps in a smaller, more power-efficient form factor currently in production alongside a new, portable mini version. Larry Goldberg detailed its smaller form factor, lightweight design, and enhanced power efficiency.

The V5 terminal delivers speeds up to 375 megabits per second, providing reliable home internet. It is now in production alongside a new, more portable mini version, expanding deployment options for users.

Early satellite technology involved garage-sized hardware with poor performance, whereas modern internet solutions emphasize smaller units with significantly higher bandwidth. Larry Goldberg recounted that early satellite technology involved garage-sized hardware that delivered poor performance, sometimes as low as 8 kilobits per second with high latency.

Modern satellite internet emphasizes smaller, more compact hardware that offers substantially higher bandwidth. The key factor for achieving mass adoption, according to Goldberg, is the price point of these new user terminals, making advanced connectivity more accessible.

Larry Goldberg explained that V2 offers a hundred times the data density and bandwidth compared to the V1 generation.

This upgrade is designed to enable native voice calls, high-speed applications, and streaming directly to mobile phones. The enhanced capabilities are expected to transform how users access the internet in remote and underserved areas.

Market participants are closely evaluating whether the service will rely on voice-over-data technology or a native voice architecture, which could impact the quality and reliability of direct phone communications.

The improved bandwidth and data density are crucial for supporting a wide range of mobile services, making satellite internet a more viable and competitive option for everyday use.

Larry Goldberg expressed uncertainty, stating he believes the current V2 satellites do not contain a Vera Rubin chip.

Analysts face challenges in confirming specific hardware components, such as AI chips, integrated into newer iterations. Speculation also exists about potential 'V2-1' versions involving component consolidation, indicating a dynamic and evolving design process within SpaceX's satellite development.