News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise ExpansionDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise Expansion
EveryNeocloud
Colossus

Colossus

An AI supercomputer cluster in Tennessee and Mississippi built by xAI.

Colossus is a massive artificial intelligence training cluster developed by xAI in the Memphis and Southaven area. It provides over one gigawatt of compute power using Nvidia accelerators to train models like Grok and serve external clients such as Google and Anthropic.

Share on WhatsApp

  • Built by xAI in Memphis and Southaven.
  • Provides over 1 GW of AI compute power.
  • Uses Nvidia H100, H200, B200, and GB300 chips.
  • Leased to Anthropic and Google.
  • Faced lawsuits over air quality and permits.

What it is

Colossus is an artificial intelligence supercomputer cluster located in the Memphis, Tennessee and Southaven, Mississippi region. Developed by xAI, now known as SpaceX Super Intelligence, it serves as the primary infrastructure for training the company's Grok chatbot. The facility spans multiple buildings, including Colossus 1, Colossus 2 (branded Macrohard), Minihard, and Macrohardrr. By mid-2026, the combined sites delivered approximately one gigawatt of computing power. The cluster utilizes hundreds of thousands of Nvidia graphics processing units to handle intensive machine learning workloads for internal development and external commercial leases.

How it came about

Construction on the initial site, Colossus 1, started in 2024 at a former Electrolux factory in South Memphis. This first phase became operational between July and September 2024 after a rapid twelve-day build period. A second major facility, Colossus 2, began development in early 2025 in the Whitehaven district of Memphis. Additional structures were added throughout 2025 and 2026 to expand capacity. The project evolved from a single data centre into a sprawling industrial complex involving significant land acquisitions and infrastructure upgrades across state lines.

The compute

The cluster relies heavily on Nvidia hardware, initially deploying around 100,000 H100 GPUs. Subsequent expansions introduced H200, B200, GB200, and GB300 processors. Colossus 2 alone housed roughly 110,000 GB200-class chips upon launch, with total counts reaching over 440,000 GB300 units by late 2026. Networking and storage support these massive arrays. The system is designed for high-throughput model training, shifting newer Grok iterations away from the older Colossus 1 hardware to the more powerful Colossus 2 environment.

Where the GPUs are

The primary operations are split between Memphis, Tennessee, and Southaven, Mississippi. Colossus 1 occupies a repurposed factory in South Memphis. Colossus 2 is situated in the Whitehaven area of Memphis, with associated power generation facilities located just across the border in Southaven. The cluster includes additional buildings at specific addresses like Tulane Road and Stateline Road West. Power is supplied through a mix of grid connections, temporary gas turbines, and a permanent plant under construction in Southaven.

Customers and deals

While xAI uses the majority of capacity for its own Grok models, significant portions are leased to third parties. In May 2026, Anthropic agreed to rent nearly all compute resources at Colossus 1. Google secured access to approximately 110,000 GPUs starting in October 2026 through June 2029, paying $920 million monthly. Other users include Reflection AI and various internal teams handling different stages of model development and inference tasks.

Money and ownership

Specific contracts reveal substantial revenue streams, such as Google's $920 million per month agreement for GPU access. Land purchases for expansion included an $80 million acquisition in Whitehaven. The project involves heavy capital expenditure on hardware, power infrastructure, and legal compliance measures related to environmental permits and community agreements.

Its place in AI

Colossus represents a significant shift in AI infrastructure deployment, moving towards dedicated, large-scale supercomputing clusters rather than shared cloud resources. It positions xAI as a major player in the sovereign and national AI compute landscape within the United States. The scale of the operation, exceeding one gigawatt, highlights the intensifying energy demands of modern artificial intelligence training. It serves as a central hub for developing frontier language models.

Questions and criticism

The rapid expansion raised serious environmental and community concerns. Temporary natural gas turbines caused air quality issues, leading to lawsuits from the NAACP and Southern Environmental Law Center under the Clean Air Act. Residents protested noise, emissions, and water usage. Legal battles involved the US Department of Justice intervening to dismiss cases citing national security. Plans for wastewater recycling were paused while focus remained on stabilizing the massive power requirements and managing regulatory compliance.

Questions people ask about Colossus

What is Colossus?

Colossus is an artificial intelligence supercomputer cluster developed by xAI. Located in Tennessee and Mississippi, it provides massive computing power for training models like Grok and hosting external client workloads.

Who are the main customers?

xAI uses the cluster for its own models. Major external customers include Anthropic, which leases Colossus 1, and Google, which rents 110,000 GPUs starting in late 2026. Reflection AI also accesses capacity.

How much does Google pay?

Google agreed to pay $920 million per month for access to approximately 110,000 GPUs. The contract runs from October 2026 through June 2029.

Why was there legal controversy?

Environmental groups sued over air pollution from temporary gas turbines used during construction. Lawsuits alleged violations of the Clean Air Act and disproportionate impacts on local communities.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Colossus (data center)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryNeocloud rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-16.

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.