The new arms race in AI compute: Who’s building the fastest hyperscale data centers?

Why the AI Infrastructure Race Is Reshaping Hyperscale Strategy in 2025

The global race to build AI-specific data center infrastructure is entering an accelerated phase, with leading firms including Applied Digital Corporation (NASDAQ: APLD), Amazon.com, Inc. (NASDAQ: AMZN), Equinix, Inc. (NASDAQ: EQIX), Digital Realty Trust, Inc. (NYSE: DLR), and privately held CoreWeave each executing multi-billion-dollar expansion plans. Driven by surging demand for GPU capacity to power generative AI, machine learning, and large-scale inference, the industry is witnessing a strategic pivot: from general-purpose colocation to purpose-built, high-density hyperscale campuses. This AI-first design philosophy is expected to redefine data center economics, geography, and investor expectations over the next decade.

Why Did CoreWeave Choose Applied Digital for Its $7 Billion AI Hosting Deal?

On June 2, 2025, Applied Digital announced two 15-year lease agreements with CoreWeave, securing 250 megawatts (MW) of IT load at the Ellendale, North Dakota campus. The deal, estimated at $7 billion in total contract value, gives CoreWeave a scalable site for AI workloads, with an option to expand by another 150 MW in the coming years. Applied Digital expects the first 100 MW to go live in Q4 2025, followed by a 150 MW facility by mid-2026. A third phase—an optional 150 MW building—could take CoreWeave’s total footprint to 400 MW. This makes Ellendale one of the largest single-tenant AI campuses currently under development in the U.S.

With energy-efficient design, proximity to low-cost power, and modular deployment capabilities, the Ellendale campus has positioned Applied Digital as a credible alternative to legacy colocation operators. Analysts view the CoreWeave deal as a defining moment for the company, which had a history rooted in blockchain hosting but is now firmly pivoting toward AI and HPC. The market responded swiftly. On the day of the announcement, Applied Digital shares surged 23.1 percent to close at $9.56. The company now holds approximately 65.7 percent institutional ownership, with major positions held by BlackRock, Vanguard, and Hood River Capital. Analyst consensus has shifted to a “Strong Buy,” with a price target range of $10.91 and bullish projections on long-term revenue anchoring.

How Fast Is Digital Realty Scaling Hyperscale Capacity in the U.S.?

Digital Realty Trust, Inc. has long been regarded as a foundational name in the REIT-backed data center industry. Now, as hyperscale AI demand accelerates, it is moving to stay relevant amid rising competition. In early 2025, the company filed for a 400 MW development in Charlotte, North Carolina, and has expanded its land holdings in Northern Virginia and Dallas. The firm reported Q4 2024 revenues of $1.4 billion, marking a 5 percent year-on-year increase. A notable uptick in leasing backlog, driven by AI-specific deployments, has signaled renewed investor confidence.

While Digital Realty maintains one of the most diversified and globally interconnected footprints, the pivot to power-dense AI workloads introduces new design and efficiency pressures. Analysts remain neutral-to-positive on the stock, citing its stable dividend performance and expanding hyperscale tenant mix. However, some caution persists regarding construction lead times and the strain of high capex cycles in an increasingly competitive environment. The company has recently reaffirmed guidance on adjusted funds from operations, with strategic commentary signaling a stronger focus on AI verticals in both U.S. and EMEA markets.

Can Equinix’s $15B Plan Keep Up With AI Demand?

Equinix, Inc. has committed to investing over $15 billion through its xScale joint venture to build AI-optimized hyperscale data centers in major U.S. metros and international hubs. The company’s Jakarta launch in May 2025 marked the beginning of a global push to deliver “AI-ready” data centers tailored for hyperscalers and cloud-native AI startups alike. These facilities support high-rack density and power availability designed for GPUs and custom silicon used in training and inference.

The Equinix xScale initiative is backed by sovereign wealth capital from GIC and CPP Investments, reinforcing the long-term view investors are taking toward AI infrastructure. Equinix has continued to generate stable revenue growth and free cash flows, maintaining a dividend yield favored by institutional investors. EQIX stock has outperformed the broader REIT sector this year, riding momentum from AI-related expansion announcements and cross-market demand in Asia and Europe. Analyst sentiment is bullish, with most forecasts pointing to continued strength in enterprise interconnection demand, even as the company juggles expansion and margin pressures.

How Much Is Amazon Really Spending on AI Data Centers in 2025?

Amazon Web Services, the cloud infrastructure powerhouse under Amazon.com, Inc., announced plans to invest at least $11 billion in new AI-optimized data center infrastructure in Georgia alone. This development, which will bring hundreds of high-skill tech jobs to the region, is part of a broader AWS capital expansion strategy to support AI workloads at scale. Amazon’s use of proprietary silicon, including Inferentia and Trainium chips, places it at the forefront of custom AI compute among hyperscalers.

The $11 billion investment builds on Amazon’s multi-state infrastructure push, with sites in Virginia, Oregon, and Ohio also under development. While much of Amazon’s AI capability is directed toward internal products like Alexa and Amazon Bedrock, the cloud division continues to service a growing roster of enterprise AI clients. Analysts tracking AMZN stock remain highly bullish, given AWS’s continued double-digit revenue growth, its industry-leading operating margin, and its entrenched developer ecosystem. However, Amazon’s AI cloud dominance is increasingly challenged by Microsoft Azure’s OpenAI integrations and niche players offering specialized GPU-as-a-Service models.

Is CoreWeave the Most Aggressive GPU Cloud Builder Right Now?

CoreWeave has emerged as a standout disruptor in AI cloud compute, offering GPU-as-a-Service capacity at a pace unmatched by larger players. The company operates 32 data centers and controls more than 250,000 GPUs as of mid-2025, largely comprising Nvidia’s H100 and A100 chips. These resources power workloads for a growing list of AI-native startups, from generative video platforms to LLM training-as-a-service vendors.

Backed by a combination of Nvidia equity, Magnetar Capital’s strategic funding, and a recent $750 million data center JV led by Blue Owl Capital, CoreWeave’s momentum is rooted in its flexibility and scale. It has secured long-term leases across key U.S. markets like Texas and North Dakota, including its flagship partnership with Applied Digital. Analysts see CoreWeave as a credible contender for IPO consideration by late 2025 or early 2026, especially if ARR growth continues and major enterprises begin adopting its platform as an alternative to AWS or GCP.

Though still privately held, CoreWeave is seen by institutional investors as a signal for the next wave of hyperscale-oriented capital allocation. Its success could encourage more infrastructure startups to pursue vertically integrated strategies, anchored in GPU density, developer tooling, and modular deployment timelines.

What’s the Sentiment Around AI Infrastructure Stocks?

Investor enthusiasm around hyperscale AI infrastructure has been intensifying. Applied Digital’s stock (NASDAQ: APLD) jumped more than 23 percent following its CoreWeave announcement. The stock has maintained strong institutional backing, with 65.7 percent of shares held by funds, including major inflows from BlackRock and Hood River Capital. Analyst consensus leans “Strong Buy,” with price targets pointing to sustained momentum provided execution risk remains contained.

Digital Realty (NYSE: DLR) continues to attract long-horizon REIT-focused capital, despite more modest price action. Equinix (NASDAQ: EQIX), by contrast, is favored for its consistent dividend profile and its central role in AI ecosystem interconnection, particularly in emerging APAC markets. Amazon (NASDAQ: AMZN), while broader in business scope, is increasingly evaluated by investors based on AWS’s ability to convert capex into enterprise AI dominance.

While sentiment is broadly positive, analysts also note structural risks including chip supply bottlenecks, rising energy costs, and regulatory scrutiny over AI data sovereignty and power usage. These concerns may shape investor behavior as the hyperscale boom matures.

Where Is AI Infrastructure Headed Next?

Between 2025 and 2028, the AI infrastructure sector is expected to experience its most significant phase of investment since the cloud computing revolution of the early 2010s. However, unlike that era, today’s hyperscale builds are being tailored for GPU clusters, liquid cooling systems, and low-latency interconnects optimized for large model inference and fine-tuning. Experts anticipate that modular builds in low-energy-cost jurisdictions like North Dakota, Texas, and parts of Europe will continue to attract long-term capital.

Future winners are likely to be those who can combine physical infrastructure scale with software-level optimization. The next wave may see further consolidation or strategic M&A, especially as companies like CoreWeave, Lambda Labs, and Vultr rise in relevance. Policy factors such as U.S. federal incentives, CHIPS Act-linked funding, and data center energy legislation will also shape competitive dynamics in the years ahead.