“Hyperscaler” has become one of the most-used words in AI investing coverage, but there’s no official textbook definition. The general industry consensus, per Cisco and IBM, is that a true hyperscale facility runs at least 10,000 square feet and houses a minimum of 5,000 servers — though the largest modern AI data centers dwarf that baseline many times over, some drawing enough electricity to power a few hundred thousand homes.
What matters more for investors than the technical threshold is who’s actually building and leasing out this infrastructure. Three companies dominate the space, together controlling roughly 63% of the public cloud market.
Amazon (AMZN)
Amazon Web Services remains the largest cloud infrastructure provider globally. Its hyperscale data centers increasingly house AI-capable hardware built to handle the massive processing workloads modern AI models require, and the company continues to expand its footprint to meet enterprise and AI-customer demand.
What to watch: AWS revenue growth, cloud operating margins, and capital spending trends relative to enterprise AI adoption.
Microsoft (MSFT)
Microsoft’s Azure platform is one of the three anchor tenants of the hyperscaler race. Its data center buildout has increasingly shifted toward higher-performance, AI-specific infrastructure, positioning it to capture enterprise AI workloads running on top of its cloud services.
What to watch: Azure growth rates, AI-related revenue disclosures, and the pace of new data center capacity coming online.
Alphabet (GOOGL)
Google Cloud rounds out the big three. What differentiates Alphabet is its proprietary Tensor Processing Units (TPUs) — custom AI chips that give it a hardware advantage some industry watchers believe could eventually help it close the market-share gap with Amazon.
What to watch: Google Cloud revenue and margin trends, and how quickly TPU-based infrastructure scales relative to competitors using third-party chips.
The next tier
A handful of other companies are building hyperscale-caliber infrastructure without necessarily competing head-on in the public cloud market:
- Meta Platforms — Building a $50 billion, 10-million-square-foot “Hyperion” data center in Louisiana, expected to be one of the largest AI facilities in the world once complete.
- Oracle — A smaller player by total footprint, but operates facilities comparable in scale to the big three.
- Apple — Doesn’t run a public cloud business, but its internal-use data centers are massive.
- Alibaba and Baidu — China’s leading hyperscaler names; Chinese infrastructure projects are increasingly built around domestically designed AI chips rather than U.S. hardware.
Why this matters for investors
As AI adoption grows, the volume of data being processed daily is measured in petabytes — a single petabyte equals roughly a million gigabytes. The companies that own and operate the infrastructure capable of that scale are positioned to capture a meaningful share of AI-driven spending, regardless of which specific AI applications or models end up winning in the market. Tracking cloud revenue growth, capital expenditure trends, and margin trajectory across these names is a useful proxy for the health of AI infrastructure spending broadly.
