xAI is adding a third major data-center facility to its Greater Memphis AI-compute cluster. The site, known as MACROHARDRR and identified in public filings as Colossus III, is in Southaven, Mississippi, just across the state line from xAI’s Memphis facilities.
Mississippi officials and xAI have described a buildout that could bring the regional cluster toward approximately 2 gigawatts of capacity. That figure is a projected or announced scale target—not proof that 2 GW is currently being consumed, that all planned GPUs are installed, or that the three buildings operate as one tightly coupled supercomputer.
As an Amazon Associate I earn from qualifying purchases.
The third facility is in Southaven, Mississippi
The original Colossus site and Colossus II are in Memphis, Tennessee. The new facility is at 2400 Stateline Road in Southaven, Mississippi, making it part of the same Greater Memphis technology and infrastructure market despite crossing a state boundary. Public documents use both “MACROHARDRR” and “Colossus III” for the project.
Free tools Windows power users keep installed
One-click scans. No signup required.
The Mississippi Governor’s Office said xAI planned to invest more than $20 billion in the Southaven project and expected operations to begin in February 2026. “Planned investment” and an expected start date should not be confused with money already spent or a fully operational facility.
#1 Best Overall
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
How the three facilities fit together
| Facility | Location | Public identity | What is known |
|---|---|---|---|
| Colossus I | Memphis, Tennessee | Colossus | The original major xAI cluster, announced with 100,000 NVIDIA Hopper GPUs. |
| Colossus II | Memphis, Tennessee | Colossus II | A follow-on expansion intended to increase training and infrastructure capacity. |
| Colossus III | Southaven, Mississippi | MACROHARDRR | A third facility intended to expand the regional cluster toward approximately 2 GW. |
NVIDIA described the original Colossus system as a 100,000-GPU xAI cluster using Hopper GPUs and Spectrum-X Ethernet networking. But public sources do not establish that all three buildings are one monolithic machine or that every facility participates in the same training jobs. They may instead operate as coordinated sites with separate power, cooling, networking and operational arrangements.
What “2 gigawatts” actually means
A gigawatt measures power, not GPUs and not AI performance. The approximately 2-GW figure attached to the expanded Greater Memphis cluster could refer to planned capacity, available power, nameplate generation or a projected compute load. It does not necessarily mean the sites continuously consume 2 GW.
Power also has to serve cooling, networking, storage, power conversion and other data-center equipment. Only part of a facility’s total power capacity reaches the GPUs, and even installed GPUs may not run at full load continuously.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsA SpaceX filing describes Colossus I and II as providing approximately 1.0 GW of “compute power,” with additional capacity for data-center operations. That wording illustrates why “compute power,” total facility power and electricity actually consumed should not be treated as interchangeable.
Confirmed facts versus targets
- Original scale: NVIDIA announced a 100,000-Hopper-GPU Colossus configuration. That was an announcement-era figure, not a permanent current count.
- Long-term target: xAI’s Memphis page says the site is intended to reach 1 million GPUs by 2026. That is a company target, not independent proof that 1 million GPUs were operating simultaneously.
- Southaven investment: Mississippi announced more than $20 billion of planned investment. The figure does not establish how much has already been spent.
- Aggregate capacity: The roughly 2-GW figure describes the intended enlarged cluster and should be read as announced or projected capacity.
- GPU equivalents: xAI said Colossus I and II ended 2025 with more than 1 million H100 GPU equivalents. GPU equivalents are an accounting or performance comparison and are not the same as a verified count of physical H100 cards.
The numbers cannot safely be added together. Physical GPUs, planned GPUs, contracted GPUs and GPU equivalents describe different things.
Why xAI is building so aggressively
Large language models require enormous computing resources for training, fine-tuning, evaluation and inference. More capacity can allow xAI to run larger experiments, shorten training cycles, serve more Grok users and support enterprise or API workloads.
Owning or controlling infrastructure may also reduce xAI’s dependence on outside cloud providers. The infrastructure can become a business asset in its own right. The SpaceX filing says the company entered cloud-services agreements with third parties, including Anthropic, and references approximately 325,000 NVIDIA GPUs, hyperscale CPUs, exabyte-scale storage and high-speed networking associated with contracted compute infrastructure.
Recommended Free Tools
Rank #2
- ATTENTION BUYERS: EXPORTING this item outside the U.S. requires strict compliance with the Bureau of Industry and Security (BIS) U.S. Export Administration Regulations (EAR). Non-compliance carries fines up to $1 MILLION PER VIOLATION AND UP TO 20 YEARS IMPRISONMENT for both the buyer and seller. To uphold our regulatory obligations, we strictly follow BIS "Know Your Customer" guidance by screening each order and immediately aborting transactions showing unresolved regulatory red flags. ***
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations. | [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
That is evidence that external compute revenue is part of the strategy. It does not prove that all of those GPUs belong to Colossus III or that the Southaven site is already delivering that capacity.
What the expansion could mean for Grok
xAI’s Memphis infrastructure is intended to power Grok. The additional facilities could support:
- Training new Grok models;
- Fine-tuning, evaluation and safety testing;
- Serving consumer inference requests;
- Enterprise and API workloads; and
- Potentially leasing excess capacity to other companies.
More hardware does not automatically produce a better model. Results also depend on data quality, model architecture, software efficiency, networking, cooling, power availability and the ability to keep the cluster highly utilized.
Power is the central bottleneck
The unusual part of the buildout is not only the number of GPUs. It is the speed and scale of the required electrical infrastructure.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchxAI’s fact-versus-fiction page says Colossus uses natural-gas turbines and that the next Memphis-area data center could use as many as 90 turbines. An environmental filing concerning Southaven describes proposed equipment including six SMT-130 turbines and ten TM2500 units. Those are plans or proposed infrastructure, not proof that every unit has been installed and is operating continuously.
On-site generation can help an AI company obtain power faster than waiting for new transmission lines, substations or grid interconnections. It also creates additional permitting, emissions, noise and fuel-supply issues.
A later Tom’s Hardware report said SpaceXAI planned to remove 69 unpermitted trailer-mounted turbines and replace them with a 1.2-GW power plant, with the removal process expected to take about a year. That later plan shows that the power architecture remained subject to change after the original Southaven announcement.
Rank #3
- System Compatibility Note: 2-slot card, 271x112x39mm, single 8-pin power, 200W TDP. Verify chassis clearance and PSU capacity before purchase.
- Dedicated Support: Please contact us directly through Amazon for any product questions or assistance you may require.
- 24GB GDDR6 on 192-Bit Bus: Massive 24GB memory with 456 GB/s bandwidth – ideal for LLMs, AI inference, 3D rendering, and generative design.
- Intel Xe2-HPG Architecture: Built on Intel's next-gen architecture with 20 Xe cores and 160 XMX engines for AI acceleration (197 INT8 TOPS).
- PCIe 5.0 Support: PCI Express 5.0 x16 interface for maximum bandwidth with the latest workstation platforms.
Environmental and community questions
The expansion is also a local infrastructure and environmental story. Key questions include:
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →- How much natural gas will on-site generation consume?
- What emissions and noise will turbines produce?
- Were all generators properly permitted?
- How much water will cooling require?
- Who pays for grid, road and utility upgrades?
- Do local economic benefits offset infrastructure and environmental costs?
xAI’s information center and fact-versus-fiction page present the company’s position on power, utility infrastructure and community concerns. Those are first-party statements, not independent environmental assessments.
An Earthjustice filing documents criticism and permitting concerns surrounding the Southaven project. Its allegations and interpretations should be attributed to the filing rather than presented as uncontested findings.
A changing corporate identity
Older announcements refer to xAI. After xAI became part of SpaceX in 2026, newer first-party materials may use the SpaceXAI name. The SpaceX prospectus provides the newer corporate context, while historical Colossus announcements remain xAI announcements.
Is Colossus the world’s largest AI supercomputer?
NVIDIA and xAI have used “world’s largest AI supercomputer” language for Colossus. Such claims are time-sensitive and depend on the metric: GPU count, theoretical performance, power, training throughput or the size of a single interconnected system.
A Tom’s Hardware analysis questioned whether Colossus II’s cooling infrastructure matched claims about its potential power capacity at that time. That does not disprove a longer-term buildout, but it reinforces the distinction between announced ultimate capacity and currently usable compute.
The bottom line on the third Colossus site
As of August 18, 2026, the most accurate description is that xAI is expanding its Greater Memphis AI infrastructure from two major Colossus facilities to three by adding MACROHARDRR, or Colossus III, in Southaven. The project is intended to push the regional cluster toward approximately 2 GW and support both Grok and potential third-party cloud-compute customers.
The important caveat is that public announcements do not establish the exact operating GPU count, sustained power draw, usable training throughput or final status of the Southaven site. The 2-GW figure is best understood as an announced capacity target, not a verified measure of electricity currently consumed or compute currently online.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
