The next great infrastructure race may not happen on Earth.

During a recent space industry panel at the Milken Institute Global Conference in Los Angeles, the conversation quickly moved beyond rockets and satellites into something more ambitious: orbital supercomputers, mobile military space networks, and industrial manufacturing in orbit.

What once sounded like science fiction is being discussed as a practical engineering roadmap.

“We are witnessing the industrialization of space,” said Bülent Altan, founding partner at Alpine Space Ventures and a former SpaceX executive. The challenge, he explained, is no longer invention but scaling. “The hunger for launch capacity is growing exponentially.”

Altan was joined on the panel by Bridgit Mendler, CEO and co-founder at Northwood, which is building ground infrastructure for space communications; Philip Johnston, CEO of Starcloud, which recently demonstrated AI compute capability in orbit; and John E. Shaw, former deputy commander of US Space Command and a Sierra Space board member who warned that the growing commercial dependence on space is rapidly becoming a national-security concern.

Their comments arrive at a pivotal moment for the industry.

Industry ready for blast off

SpaceX is preparing for a Nasdaq listing next month that could rank as the largest initial public offering in history, with a valuation approaching USD 2tn. Investors and executives across the aerospace sector increasingly see it as a watershed moment that has already started to pull massive new pools of capital into the broader space economy.

Aerospace and defense components maker Arxis successfully debuted on the Nasdaq in April, raising over USD 1.1bn. Satellite manufacturer York Space Systems raised USD 629m in a NYSE debut in January. Firefly Aerospace raised USD 868m on the Nasdaq in August. And others like Greenbriar-backed Applied Aerospace & Defense are lining up their own listings.

Yet SpaceX’s dominance creates a void in the market for a second full-stack launch company. RocketLab, despite its progress, is not yet profitable and operates at a much smaller scale. It may take an entire ecosystem to match the breadth of capabilities that SpaceX can deliver, Altan said.

He pointed to a need for greater geographic diversification, highlighting Europe and citing Isar Aerospace in Germany as an emerging launch provider that can help ease concentration risk.

Alongside startups and midmarket players, tech giants are now entering the fray.

On 12 May, Google disclosed discussions with SpaceX and other launch providers such as Planet Labs regarding future missions for an orbital data-center project designed to explore whether large-scale machine learning workloads can eventually operate in space. Google holds an estimated 6% stake in SpaceX. The initiative has drawn skepticism from engineers who question whether orbital compute can overcome launch costs and thermal constraints. Still, the concept is gaining traction as reusable launch systems reduce the cost of reaching orbit.

Unlocking more launch capacity would have a cascading impact across the sector, Altan said, but launch is only one constraint. Power and data transmission remain critical bottlenecks. Altan pointed to portfolio companies including Source Energy, which is developing power infrastructure, and Northwood, which focuses on moving data between space and Earth.

The discussion’s timing is not coincidental. The AI boom on Earth is colliding with growing resistance to large terrestrial data centers, which consume enormous quantities of electricity and cooling water. Orbital infrastructure offers a provocative alternative: effectively limitless solar power and the possibility of relocating energy-intensive compute away from population centers.

Starcloud is testing that thesis.

A supercomputer in orbit

Last year, Starcloud successfully operated an Nvidia H100 AI chip in space, a milestone many aerospace engineers considered implausible because of two longstanding problems: radiation exposure and heat dissipation. Roughly 70% of Starcloud’s engineering work is building a low-cost, low-mass deployable radiator, which is expected to launch early next year.

Johnston said the company’s first successful onboard experiments included training a language model on the complete works of Shakespeare and running a version of Google’s Gemini AI system in orbit. He recalled receiving the first successful outputs while alone in a hotel room in Helsinki during a conference. “We were very unsure this was even possible,” Johnston said.

Earlier this month, Johnston told Mergermarket that Starcloud closed a USD 200m funding round from a mix of new and existing investors to support the development of next-generation satellites, expanded manufacturing, and build-out of its space-based data center platform.

Johnston believes the first commercial orbital compute clusters could emerge as early as 2028, largely dependent on the arrival of fully reusable heavy-lift systems such as SpaceX’s Starship.

The objective is to make Starcloud’s platform lightweight and cheap enough to scale, he said.

Scaling off-world infrastructure

While rockets capture public imagination, the panel repeatedly returned to the less glamorous but increasingly important matter of infrastructure. The most valuable companies, according to Altan, may be those rebuilding the infrastructure stack beneath them. There are holes all across the stack, such as power systems, launch capacity, communications, and manufacturing, he said.

Networking infrastructure is another critical layer in what was once a fringe concept.

Mendler’s company, Northwood, is developing deployable ground stations designed to connect satellites and spacecraft. Without those systems, satellites cannot reliably transmit data, receive commands, or coordinate with one another. In January, Northwood raised USD 100m in Series B financing, nine months after securing USD 30m in a Series A. The startup has also won a USD 50m US Space Force contract to expand capacity across its satellite control network.

Speaking on the conference sidelines, Mendler told Mergermarket the company continues to see inbound investor interest and will be opportunistic about raising additional capital this year.

She argued that networking infrastructure should be shared, allowing space companies to focus on mission-level execution. “We believe networking infrastructure should be the boring layer of space,” Mendler said, likening it to the role Amazon Web Services plays in cloud computing.

Northwood’s hardware is intentionally portable. Mendler described units small enough to ship aboard commercial passenger flights and rugged enough to be forklifted into remote environments and activated within hours. That mobility, she said, is critical as military planners reassess the vulnerability of fixed ground infrastructure in future conflicts. Large static ground stations, such as dish antennas, are conspicuous military targets for adversaries, she noted.

The future is dynamic operations, she said.

Space becomes security theater

Addressing these military concerns, the discussion underscored a larger strategic shift in which outer space is becoming operational terrain rather than merely an enabling domain.

Shaw described a future where launch facilities, satellite networks, communications links, and orbital compute assets turn into contested infrastructure. “Space and cyber are inseparable,” the three-star general said. He outlined a range of threats affecting military planning, including cyberattacks, signal jamming, directed-energy weapons, and kinetic anti-satellite systems.

As governments and companies move activity into orbit, the pressure to defend those assets grows accordingly, he said. That change is driving new interest in resilient architectures: larger constellations, mobile infrastructure, and systems capable of rapid replacement after attack.

Johnston noted that Starcloud recently filed regulatory paperwork for a proposed constellation of roughly 88,000 satellites, partly to enhance redundancy in the face of potential military conflict.

In a perfect world

The discussion eventually turned to more radical ideas, including relocating entire categories of heavy industry and energy-intensive manufacturing off-world.

“Why use Earth for any dirty industry? We’re looking at more and more planets and not finding any that compare to the one we are on,” Altan said, noting that many industries depend on raw materials that may be even more abundant in space than on this blue and green biosphere.

In a perfect world, Earth may ultimately become primarily a residential planet, the panelists said.