To The Moon

xAI's CapEx Is Set to Join the Top Tier. Who Are the Suppliers?

In February 2026, xAI was officially merged into $Space Exploration Technologies Corp(SPCX)$   , triggering a rapid expansion cycle for its Colossus supercomputing cluster. AI has now become the primary focus of the group's CapEx. 

Based on the latest quarterly earnings report and management commentary, while no absolute dollar guidance was provided for 2027, management confirmed that single-quarter CapEx in the second half of the year will remain at levels similar to Q2. The market projects that the SpaceX group's overall CapEx next year could exceed $100 billion, with the vast majority of resources allocated to the AI compute segment corresponding to xAI. As Colossus-1 and Colossus-2 continue to expand, xAI's compute capacity will rapidly catch up to cloud giants like Microsoft, Google, and Amazon, officially joining the ranks of top-tier global AI compute buyers.

This massive upfront capital investment will ripple outward along the entire supply chain—including chips, server OEMs, high-speed interconnects, data center infrastructure, power and energy storage, and public cloud partnerships. 

Chips

$NVIDIA(NVDA)$   is the core supplier, providing the primary GPUs for the Colossus clusters, as well as the complete Spectrum-X networking platform with 800G Ethernet to support RDMA for GPU clusters with hundreds of thousands of units. xAI also received Series C funding from Nvidia.

$Broadcom(AVGO)$   : Broadcom supplies the internal switching chips for Spectrum-X switches; any deployment of NVIDIA Spectrum-X switches inherently relies on Broadcom’s underlying switching silicon.

$Intel(INTC)$   : In xAI's Colossus clusters, NVIDIA's in-house Grace CPUs (Arm architecture) are primarily used for collaborative computing with GPUs in AI compute nodes. However, for general-purpose nodes such as cluster management and storage, Intel Xeon CPUs are widely deployed to handle scheduling, operations, storage I/O, and other non-AI training workloads. Additionally, Intel has officially joined Elon Musk's Terafab chip foundry initiative, providing process and manufacturing support for the chip fabrication plans of Tesla, SpaceX, and xAI. This project is still in the plan phase and has not yet seen large-scale deployment in existing Colossus clusters in the short term.

Servers & Hardware Integration

$Super Micro Computer (SMCI.US)$ : The primary supplier for the initial phase of Colossus, providing liquid-cooled chassis and storage nodes, as evidenced by early public case study white papers. Some later orders have been shifted to Dell.

$Dell Technologies (DELL.US)$ : Began securing a significant volume of xAI server orders starting in late 2024, with some volume transferred from Supermicro.

Data Center Infrastructure

$Credo Technology (CRDO.US)$ : Industry insiders have confirmed the presence of Credo's signature purple Active Electrical Cables (AECs) in photos taken by Elon Musk at the Colossus-2 site in July 2025. These cables are being widely deployed for switch-to-switch interconnects between racks.

$Corning (GLW.US)$ : A top choice for fiber optic cabling in large-scale North American AI Data Centers. As a core supplier and officially announced Spectrum-X ecosystem partner at NVIDIA GTC, Corning is highly likely to be supplying MPO fiber patch cables for Colossus. While there is no official xAI order announcement, this is a strong ecosystem-driven inference.

$Vertiv Holdings (VRT.US)$ : An official NVIDIA DSX ecosystem partner, providing PDUs (Power Distribution Units), UPS systems, Coolant Distribution Units (CDUs) for liquid cooling, and rack infrastructure. Vertiv's infrastructure is widely adopted in gigawatt-scale AI supercomputing projects across North America.

Power & Energy Storage

$Tesla (TSLA.US)$ : Tesla's Megapacks are used for grid-side energy storage, peak shaving, and backup power at xAI data centers. SpaceX filings have disclosed significant procurement volumes. With xAI now merged into SpaceX, this constitutes a related-party transaction backed by formal filing data.

$Solaris Energy Infrastructure (SEI.US)$ : xAI and Solaris have formed a joint venture to build gas turbine power plants, providing over 1GW of backup generation capacity for Colossus-2.

Cloud

$Oracle (ORCL.US)$ : The entire Grok 3/4 model family is officially hosted on the OCI Generative AI Marketplace, allowing enterprise customers to access them directly. Furthermore, OCI provides the underlying GPU compute for training and inference for xAI's next-generation Grok models, with xAI heavily leveraging OCI's clusters. 

Note: While Grok models are also listed on $Microsoft (MSFT.US)$ Azure and $Amazon (AMZN.US)$ AWS, there is virtually no large-scale collaboration on training compute.

@TigerStars  @CaptainTiger  @TigerWire  @Daily_Discussion  @Tiger_chat  @Tiger_comments  @MillionaireTiger