Author: Ye Zhen; Source: Wall Street News

The listing of SpaceX not only reshaped the market value pattern of global technology giants, but its strategy of entering the space AI field also directly triggered a new round of market trends in the hardware manufacturing and optical communications industry chain.

On the 12th local time, SpaceX officially landed on the Nasdaq exchange and successfully raised approximately US$75 billion. The closing price on the first day surged by nearly 20%, with the market value soaring to US$2.11 trillion. On the same day of listing, Musk responded to Nvidia’s congratulations on social platform X and made it clear that he would push the cooperation between the two parties to a “new level.”

Behind this statement is SpaceX’s huge space computing power blueprint. Wall Street News previously mentioned that Musk disclosed for the first time the core technical parameters of its first-generation dedicated AI computing satellite "AI1" on the eve of SpaceX's listing. He plans to deploy the AI ​​data center directly to space orbit and has signed large orders for cloud service computing power worth tens of billions of dollars with companies such as Google and Anthropic.

The huge computing power deployment plan was quickly transmitted to the capital market and supply chain. Industry analysts believe that as the demand for space server cabinets and inter-satellite laser communications becomes clear, major server assembly manufacturers such as Hon Hai and Quanta, as well as key optical communication component suppliers such as Lianya and Lianjun, are ushering in substantial order expectations and business increases.

The AI1 satellite is unveiled and 1GW of space computing power will be deployed next year

The first dedicated AI computing satellite launched by SpaceX this time, AI1, aims to break through the limitations of ground infrastructure.

The satellite supports a peak computing power of up to 150 kilowatts per satellite and is equipped with a liquid cooling radiator, meteoroid protective layer, centralized computing module and deployable solar panels. The equipment will be mass-produced at SpaceX's Gigasat facility in Texas.

According to SpaceX's plan, the company will fully enter the space AI market and achieve an annual deployment of 1GW (1 billion watts) of space AI computing power by the end of 2027, gradually promote commercialization in 2028, and finally increase the computing power scale to 100GW in 2030.

7400 cabinet annual demand: server OEM giant benefits

To realize a space computing power network of this scale, building an AI server cabinet suitable for the space environment is the first priority.

According to industry research cited by the media, every time SpaceX launches an AI1 satellite, it will send an NVIDIA GB300 AI server cabinet into space. It is estimated that the energy consumption of a single GB300 cabinet is about 135 kilowatts. The annual computing power deployment of 1GW is equivalent to the annual demand for approximately 7,400 new cabinets of GB300 AI servers.

Driven by huge hardware demand, the relevant supply chain has begun to operate. The industry expects that the early assembly of related servers will still be handled by ODM (original design manufacturer) partners such as Hon Hai, Quanta, and Wistron. After completing basic assembly, these foundries will deliver it to SpaceX, which will perform final adjustments on satellite body assembly and space adaptability.

1Tbps bandwidth interconnection: the optical communication industry chain welcomes new growth

In addition to the core server computing unit, data transmission between AI1 satellites also relies on high-end hardware support.

SpaceX plans to realize direct interconnection between satellites through advanced laser optical paths, thereby building a high-transmission, low-latency data transmission network in space.

This design enables large-scale AI computing tasks to achieve efficient collaboration and data exchange between different satellite nodes, speeding up computing. At present, the laser optical path used by SpaceX's new generation of Starlink satellites can support high-speed bandwidth of 1Tbps level in a single orbit.

According to industry analysis quoted by the media, the main component suppliers of this function are two international laser giants, Coherent and Lumentum. At the same time, this has also directly boosted the prosperity of the upstream supply chain, including Lianya, a supplier of indium phosphide (InP) epitaxial wafers, and Lianjun, a Coherent laser packaging foundry, which will all benefit from this expansion of space computing power.

Monetization of computing power: Transformation from rocket launch to cloud service provider

While advancing hardware deployment, SpaceX has begun to convert its computing power reserves into commercial orders, accelerating its transformation into an underlying AI computing power supplier.

On the eve of listing, SpaceX has signed a cloud service agreement with Google, providing it with computing power equivalent to 110,000 Nvidia GPUs and related components, charging a monthly fee of US$920 million.

In addition, SpaceX provides AI startup Anthropic with access to 220,000 NVIDIA GPUs, including H100, H200 and GB200. The contract charges US$1.25 billion per month, equivalent to US$15 billion per year.

In addition to computing power leasing, SpaceX is also expanding its ecological layout horizontally. The company is working with Tesla and Intel to rapidly advance the construction of a project called "Terafab" in an attempt to build a multi-layered business ecosystem spanning space, computing power and manufacturing.