<!--
Source: https://scrutica.com/facilities/fac-xai-colossus-memphis
Format: native markdown twin rendered from the same data modules as the page (not an HTML conversion).
Machine access: https://scrutica.com/api/mcp (MCP) · https://scrutica.com/agents (all surfaces)
-->

# xAI Colossus Memphis

Facility record `fac-xai-colossus-memphis` — data source: press-reporting (https://siliconangle.com/2025/12/30/elon-musk-reveals-plan-expand-xais-colossus-data-center-2-gigawatts/); authority tier 3; contains estimated values (is_estimated: true)

| Field | Value |
| --- | --- |
| Type | ai training |
| Status | expanding |
| Location | Memphis, Tennessee, US (35.0571, -90.1132) |
| Power capacity | 2,000 MW |
| IT load | not published |
| GPU count | 80,000 |
| Peak PFLOPS (FP16) | not published |
| Owner | xAI |
| Operator | xAI |
| Announced | 2024-06-01 |
| Operational | 2024-09-01 |

Grid context: balancing authority Tennessee Valley Authority; grid carbon intensity 407 gCO2/kWh.

xAI's Colossus supercluster complex in Memphis, TN. World's largest single-site AI training installation as of Jan 2026. Colossus 1 (Bldg 1): 100K H100, operational Sep 2024 in 122 days. Colossus 2 (Bldg 2): ~455K cumulative GPUs including GB200 Blackwell chips, crossed 1 GW threshold Jan 2026. Building 3 (Southaven, MS, nicknamed 'MACROHARDRR') purchased Dec 2025, conversion to data center began Q1 2026. Total planned capacity: 2 GW.

## Data-quality flags

- flop_value_units_anomaly
- flop_value_units_anomaly

## Cite

> Scrutica, "xAI Colossus Memphis" (facility record), https://scrutica.com/facilities/fac-xai-colossus-memphis; data source: press-reporting; accessed 2026-07-27.

Full interactive record: https://scrutica.com/facilities/fac-xai-colossus-memphis · Methodology for derived values: https://scrutica.com/methodology#flop-estimation · All programmatic surfaces: https://scrutica.com/agents