Project Sunrise: Blue Origin's Orbital Data Centre Ambition

The AI race is blasting off with orbital velocity – this time, out into space. 

Moving the world's most demanding AI infrastructure beyond the confines of Earth could soon become a reality with Jeff Bezos' space technology company, Blue Origin, unveiling plans for an orbital data centre network designed specifically to handle AI workloads.

Blue Origin has filed an official request with the US Federal Communications Commission (FCC) to launch and operate a constellation of up to 51,600 AI satellites, known as Project Sunrise.

The network, designed to host orbital data infrastructure, is marking a major expansion of Blue Origin's focus beyond its core launch services and positioning the company at the forefront of space-based AI computing.

Project Sunrise is intended to meet the ever-growing demand for AI workloads that are increasingly straining terrestrial infrastructure.

The FCC filing positions orbiting servers as "a complement to terrestrial infrastructure by introducing a new compute tier that operates independently of Earth-based constraints".

Addressing AI infrastructure limitations

The argument underpinning Project Sunrise is that Earth's infrastructure is struggling to keep pace with the power demands of AI training and inference workloads.

Blue Origin positions space-based data centres as the solution to overcome the limitations of Earth's AI infrastructure.

The company argues that moving computing capacity into orbit offers key advantages over land-based data centres, which contend with constraints such as land availability, massive energy consumption and extensive cooling requirements.

These limitations have become particularly acute as large language models and other AI systems demand increasingly powerful computational resources.

Blue Origin also highlights environmental benefits, stating: "The Project Sunrise system will ease mounting pressure on US communities and natural resources by shifting energy- and water-intensive compute away from terrestrial data centres, reducing demand on land, water supplies and electrical grids."

From a financial perspective, the company argues that orbital infrastructure could lower the marginal cost of compute capacity for AI applications.

"The built-in efficiencies of solar-powered satellites, always-on solar energy, lack of land or displacement costs and non-existent grid infrastructure disparities fundamentally lower the marginal cost of compute capacity compared to terrestrial alternatives," adds Blue Origin in its FCC filing.

Solar panels orbiting the planet can generate electricity continuously, dodging the interruptions caused by night-time and weather that hamper ground-based solar farms.

"These giant training clusters, those will be better built in space, because we have solar power there, 24/7. There are no clouds and no rain, no weather," Jeff Bezos had elaborated in the past, when discussing his ambitions for data centres in space. 

"We will be able to beat the cost of terrestrial data centres in space in the next couple of decades." Jeff says.

Technical design and deployment

Blue Origin plans to deploy up to 51,600 satellites into sun-synchronous orbit, which will keep them in constant sunlight. This will be crucial for keeping the AI computing services running continuously without interruption.

The satellites would operate at altitudes between 500 and 1,800 kilometres, with each orbital plane hosting between 300 and 1,000 satellites.

Communication between the satellites would primarily rely on optical intersatellite links (laser links), enabling high-speed data transfer essential for distributed AI training and inference tasks. These links will be supported by a mesh network architecture for efficient data routing.

Another of Blue Origin's planned projects, the TeraWave space broadband service, will manage the connectivity to Earth. The company is seeking FCC authorisation to use Ka-band radio frequencies for telemetry, tracking and control operations.

Blue Origin has acknowledged the environmental concerns associated with deploying such a vast number of satellites, stating its intention to comply with guidelines to minimise orbital debris and deorbit its satellites within five years of the end of their operational lives.

It also committed to working with the astronomy community to minimise the satellites' brightness and impact on astronomical observations.

The challenges facing the project are numbered, this includes the fact that Blue Origin has only flown its New Glenn rocket – which it plans to use for Project Sunrise launches – twice.

Moreover, the company has yet to launch a single TeraWave satellite, with plans to launch the first of its 5,000-plus TeraWave orbiters before the end of 2027.

Competition in orbital AI infrastructure

Orbital data centre aspirations are not limited to Blue Origin. In fact, this proposal comes amid a broader industry push to expand space-based AI capabilities and supports the exponential growth of machine learning applications globally.

According to FCC filings, SpaceX has filed its own application for up to one million orbital data centre satellites, which Amazon called on the FCC to deny, claiming it would take "centuries" to deploy.

Elsewhere, a startup named Starcloud has proposed a constellation of up to 88,000 satellites.

These proposals share common design elements, such as the use of sun-synchronous orbits and optical intersatellite links, suggesting a potential standardisation of space-based AI infrastructure architecture.