Top 10 U.S. Data Center Operators by GPU Deployment: The Race to Build AI Infrastructure

The rapid expansion of artificial intelligence is reshaping computing infrastructure. Modern AI applications, including generative models, enterprise copilots, autonomous technologies, and advanced simulations, rely heavily on GPU-based parallel processing rather than traditional compute resources. As a result, the scale of GPU deployment has become a key indicator of competitiveness in the data center sector.

For infrastructure vendors, semiconductor companies, investors, and private equity firms, identifying which organizations operate the largest GPU clusters provides insight into capital investment, long-term AI strategies, supply chain positioning, and infrastructure dominance.

Below is an overview of the leading U.S. data center operators ranked by total GPU deployment and what their scale indicates about the evolving AI infrastructure market.

Explore BIS Market IQ

?

1. Amazon Web Services (AWS) – ~2.5 Million GPUs

Amazon Web Services currently leads the U.S. market with an estimated 2.5 million GPUs deployed across its data center network. As the largest hyperscale cloud provider globally, AWS has made extensive investments in GPU-based infrastructure to support large-scale AI model training, enterprise AI workloads, and high-performance computing.

Although AWS has developed custom AI chips such as Trainium and Inferentia, GPUs remain central to its AI cloud services. Its global data center footprint enables distributed AI training environments and provides significant procurement and operational advantages.

2. Meta – ~1 Million GPUs

Meta has deployed approximately 1 million GPUs, primarily to support internal AI operations. These clusters power a wide range of applications, including Llama foundation model training, recommendation systems, and immersive digital platforms.

Unlike public cloud providers, Meta’s infrastructure is designed mainly for internal use. Its large capital investments in AI infrastructure reflect a broader industry trend in which technology companies build dedicated AI superclusters instead of relying entirely on third-party cloud services.

3. Oracle – ~808,000 GPUs

Oracle Cloud Infrastructure (OCI) has rapidly expanded its AI capabilities, reaching roughly 808,000 GPUs in deployment. Oracle has positioned OCI as a high-performance alternative to traditional hyperscalers by offering dedicated AI clusters and competitive GPU pricing.

Its focus on low-latency networking, optimized performance environments, and partnerships with AI startups has helped OCI attract organizations seeking large GPU capacity without the complexity often associated with larger cloud ecosystems.

Drive Smarter Decisions – Get the Database Today - Request a Quote

4. Google – ~793,100 GPUs

Google operates an estimated 793,100 GPUs across its U.S. data centers. While the company is widely recognized for its proprietary Tensor Processing Units (TPUs), GPUs still play a significant role in supporting enterprise AI workloads and generative AI services such as Gemini.

Google’s strategy combines custom silicon with traditional GPU infrastructure, enabling flexible workload allocation and improved cost efficiency.

5. Microsoft – ~726,100 GPUs

Microsoft has deployed approximately 726,100 GPUs, largely driven by its collaboration with OpenAI and the rapid expansion of Azure’s AI capabilities. The company has developed specialized AI supercomputing environments designed for both training and inference workloads.

The widespread integration of AI copilots across Microsoft’s software platforms has further increased demand for GPU capacity within Azure’s data centers.

6. CoreWeave – ~611,600 GPUs (Neo-Cloud Provider)

CoreWeave represents one of the fastest-growing AI-focused cloud providers, with around 611,600 GPUs deployed. Originally built for graphics rendering workloads, the company shifted its focus toward AI infrastructure early and capitalized on growing demand for GPU-intensive computing.

Unlike hyperscalers offering broad cloud services, CoreWeave concentrates specifically on high-performance GPU environments optimized for AI training and inference.

7. xAI – Colossus 2 (Memphis) – ~550,000 GPUs

xAI’s Colossus 2 data center in Memphis houses roughly 550,000 GPUs, making it one of the largest dedicated AI training clusters in the United States. The facility is designed to support next-generation foundation model development.

This deployment highlights an increasing trend in which AI companies build proprietary “AI factories” to maintain full control over performance optimization, model training pipelines, and infrastructure scalability.

8. Apple – ~263,000 GPUs

Apple operates approximately 263,000 GPUs across its data center infrastructure. These resources primarily support AI-powered capabilities integrated across the company’s product ecosystem.

From backend AI processing to features enhancing on-device intelligence, Apple’s GPU investments are closely tied to improving user experiences rather than expanding public cloud services.

9. IBM – ~197,200 GPUs (Enterprise AI Focus)

IBM has deployed roughly 197,200 GPUs, largely dedicated to enterprise and hybrid cloud AI solutions. Through IBM Cloud and Watson-based platforms, the company provides AI infrastructure tailored to industries that require strict compliance and data security.

Although smaller in GPU count compared with hyperscalers, IBM maintains strong positioning in enterprise-grade AI environments.

10. Nscale – ~104,000 GPUs (Emerging AI Infrastructure Provider)

Nscale completes the top ten with approximately 104,000 GPUs deployed. The company represents a new category of specialized AI compute providers that focus on delivering dedicated GPU infrastructure for AI workloads.

The presence of firms like Nscale reflects a broader diversification of the AI infrastructure market, where niche providers complement hyperscale cloud platforms.

Explore BIS MarketIQ-  Request a Demo

Hyperscalers vs. AI-Native Infrastructure Providers

The leading GPU operators fall into several structural categories:

Hyperscale Cloud Providers
 AWS, Microsoft, Google, Oracle, Apple

Vertically Integrated AI Companies
 Meta, xAI

AI-Focused Neo-Cloud Providers
 CoreWeave, Nscale

Enterprise Cloud Platforms
 IBM

While hyperscalers still control the majority of GPU capacity, specialized AI infrastructure providers are expanding rapidly and attracting significant investment.