Huawei Data Center Rivals 7% of xAI Colossus
Epoch AI estimates Huawei's Horinger facility in Inner Mongolia holds about 7% of xAI's Colossus 2 computing capacity.
Updated on Nov. 6, 2026 in Data Centers
Epoch AI estimated that Huawei's Horinger data center in Inner Mongolia holds about 7% of the computing capacity of xAI's Colossus 2 facility. Horinger is operational with roughly 81,000 H100-equivalent units of AI compute, while Colossus 2 is estimated at about 1,112,000 H100-equivalent units.
Why it matters
The comparison highlights the scale gap between China's state-backed AI infrastructure and a leading U.S. AI campus. Both facilities are growing, with projections pointing to substantially higher compute in the near future.
Epoch AI's estimate puts Horinger at about 81,000 H100-equivalent units on 170 MW of IT power, against Colossus 2's roughly 1,112,000 units on 946 MW. The projected expansion targets for both sites remain estimates and have not been independently confirmed.
The players
Epoch AI
An independent research organization that tracks and estimates the computing capacity of artificial intelligence systems and data centers.
xAI
An artificial intelligence company founded by Elon Musk that operates the Colossus supercomputer campus in Memphis.
Huawei
A Chinese multinational technology company that designs the Ascend chips used at the Horinger data center.
The details
Horinger, located in Inner Mongolia, runs mainly on Huawei's Ascend 910B and 910C accelerators and is part of China's Eastern Data, Western Computing initiative. Colossus 2, spread across multiple buildings on the Memphis, Tennessee campus, leans on a large deployment of Nvidia GB200 and GB300 chips.
Timeline
End of 2026: Colossus 2 targets over 1.2 million H100-equivalent units.
Near future: Horinger is projected to reach about 141,000 H100-equivalent units.
Big Picture
Horinger is part of China's Eastern Data, Western Computing initiative, which aims to move computing capacity to western provinces where power and land are cheaper. Its scale relative to xAI's Colossus 2 shows how that policy is being measured against the largest U.S. AI campuses.
For readers following AI costs and access, the gap in compute between these facilities affects the pace at which new AI models can be trained and offered. Changes in capacity at these campuses can influence the availability and pricing of AI services over time.
The takeaway
The scale of AI infrastructure is becoming a measurable competitive factor between nations and companies, and tracking estimates from independent groups like Epoch AI helps readers compare these facilities on consistent terms.
Further reading
Explore more on Data Centers to see how AI compute capacity is being built out around the world.
