Land and Expand: NVIDIA, IREN, Coatue, Microsoft, Switch, Cerebras, Core Scientific

The latest wave of AI infrastructure expansion is no longer defined solely by hyperscale land grabs and new data center campuses. Increasingly, the industry’s “land and expand” strategy now encompasses the broader industrial ecosystem required to deploy AI at scale.

That ecosystem now spans power, optical manufacturing, accelerated compute architecture, supply chains, and vertically integrated infrastructure partnerships.

This latest edition of Land and Expand examines how the AI infrastructure buildout is evolving beyond traditional data center development into a far more integrated industrial systems effort, where the race to scale AI increasingly depends on coordinating land, energy, networking, manufacturing capacity, and deployment architecture simultaneously.

That shift came into sharper focus this month as NVIDIA unveiled two major partnerships aimed at accelerating the physical buildout of AI infrastructure at industrial scale: one centered on gigawatt-scale AI factory deployment with IREN, and another focused on dramatically expanding U.S.-based optical connectivity manufacturing with Corning Incorporated.

Taken together, the announcements illustrate how the AI buildout is rapidly expanding beyond a conventional data center construction cycle into something closer to a coordinated industrial infrastructure movement.

NVIDIA and IREN Target 5 Gigawatts of AI Infrastructure

NVIDIA and IREN announced a strategic partnership intended to support deployment of up to 5 gigawatts of NVIDIA DSX-aligned AI infrastructure across IREN’s global data center pipeline over time.

Future deployments are expected to focus heavily on IREN’s 2-gigawatt Sweetwater campus in Texas, which the companies described as a flagship deployment for NVIDIA’s DSX AI factory architecture.

The partnership is designed to combine NVIDIA’s AI systems and accelerated computing architecture with IREN’s expertise across power, land, data centers, GPU deployment, and infrastructure operations.

The agreement also signals NVIDIA’s increasingly direct involvement in infrastructure deployment strategy itself. As part of the partnership, IREN issued NVIDIA a five-year right to purchase up to 30 million shares of ordinary stock at an exercise price of $70 per share, representing a potential investment opportunity of up to $2.1 billion, subject to regulatory and other conditions.

“AI factories are becoming foundational infrastructure for the global economy,” said Jensen Huang, founder and CEO of NVIDIA. “Deploying these systems at scale requires deep integration across the full stack — compute, networking, software, power and operations.”

Huang added that IREN brings “the scale and infrastructure expertise to help accelerate the buildout of next-generation AI infrastructure globally.”

For IREN, the partnership reflects the growing strategic value of operators capable of securing not only GPU capacity, but also large-scale power access and deployable land positions suitable for AI factory development.

“This partnership combines NVIDIA’s AI systems and architecture leadership with IREN’s expertise across power, land, data centers, GPU deployment and infrastructure operations,” said Daniel Roberts, cofounder and co-CEO of IREN. “Together, we believe we can accelerate deployment of AI infrastructure and expand access to compute for AI-native and enterprise customers globally.”

Manufacturing the AI Buildout

One day earlier, NVIDIA announced a separate long-term partnership with Corning Incorporated aimed at dramatically expanding U.S.-based optical connectivity manufacturing capacity for AI infrastructure.

Under the agreement, Corning plans to increase its U.S.-based optical connectivity manufacturing capacity by 10x while expanding domestic fiber production capacity by more than 50%.

The expansion includes construction of three new advanced manufacturing facilities in North Carolina and Texas and is expected to create more than 3,000 jobs.

The announcement underscores how AI infrastructure growth is now extending deep into industrial supply chains traditionally viewed as adjacent to the data center sector itself.

As AI factories scale into deployments requiring thousands of GPUs operating within tightly integrated accelerated computing environments, optical connectivity has emerged as a critical infrastructure layer supporting the movement of massive volumes of data across AI systems.

“AI is driving the largest infrastructure buildout of our time — and a once-in-a-generation opportunity to reinvigorate American manufacturing and supply chains,” Huang said. “Together with Corning, we are inventing the future of computing with advanced optical technologies.”

Corning Chairman and CEO Wendell P. Weeks framed the partnership as evidence that the AI boom is rapidly becoming an advanced manufacturing story as much as a computing story.

“This partnership is proof that AI is not just a technology story,” Weeks said. “It is a manufacturing story, and it is happening here in the United States.”

Capital Flows Toward Powered Land

The race to secure AI infrastructure capacity is also reshaping how investment firms approach digital infrastructure itself.

According to reporting from The Wall Street Journal, technology-focused investment firm Coatue Management has launched a new venture called Next Frontier focused on acquiring land positioned near large power sources for future AI data center development.

The effort signals how “land and expand” strategies are increasingly drawing participation not only from hyperscalers and traditional data center developers, but also from major AI investors seeking direct exposure to the physical infrastructure layer supporting next-generation AI systems.

The Wall Street Journal reported that tens of billions of dollars could ultimately flow into the initiative, which is being led by Robert Yin, a Coatue partner focused on AI infrastructure, alongside former Blackstone executive Peter Wallace, who recently joined Coatue as head of private investments.

Founded by prominent technology investor Philippe Laffont, Coatue has emerged as one of the industry’s more aggressive AI infrastructure investors, with positions spanning AI labs, neocloud providers, GPU infrastructure companies, power firms, and data center operators.

The firm has invested across companies including OpenAI, Anthropic, CoreWeave, Hut 8, GE Vernova, Vistra, and data center firms including Scala Data Centers and DayOne.

The new venture appears aimed at extending that strategy directly into the powered land and infrastructure development layer now emerging as one of AI’s most strategically constrained resources.

According to both the Wall Street Journal and reporting from Data Center Dynamics, Next Frontier has already formed a joint venture with neocloud provider Fluidstack tied to development of a large-scale AI campus in Indiana.

The project involves a planned 430 MW data center complex in New Lebanon, Indiana, backed by $5.7 billion in senior secured notes issued through a joint venture tied to Fluidstack and Frontier Holdings Indiana LLC, an affiliate of Next Frontier.

According to Fitch Ratings documentation cited by Data Center Dynamics, the campus is anchored by a 15-year lease with Fluidstack that includes three five-year extension options. The first 65 MW data hall is expected online in July 2027.

Power for the development is expected to come from Hoosier Energy and WIN Energy.

The broader campus buildout reflects the rapidly escalating scale of AI infrastructure development now moving into emerging markets positioned around large power availability.

Data Center Dynamics reported that the development includes plans for both a 245 MW turnkey data center and an additional 185 MW facility, along with construction of an onsite electrical substation across an approximately 140-acre property in Sullivan County, Indiana.

The site also appears connected to broader infrastructure activity unfolding around the region’s Heartland Industrial Park near the Merom Generating Station, where up to 2.1 GW of future development has been discussed.

The Fluidstack relationship also highlights the increasingly interconnected nature of the AI infrastructure ecosystem now forming between neocloud providers, hyperscalers, AI model developers, infrastructure investors, and power-backed land developers.

Several recent Fluidstack-related projects tied to Anthropic deployments have included financial backing or guarantees from Google, which has invested billions into Anthropic and continues supporting infrastructure deployments tied to the Claude AI platform.

Anthropic previously announced plans to invest $50 billion into U.S. AI infrastructure alongside Fluidstack, including deployments in Texas and New York.

Viewed together, the Coatue and Fluidstack developments further illustrate how the AI infrastructure race is increasingly pulling together capital markets, AI model developers, cloud platforms, energy infrastructure, and powered land acquisition into a single integrated development ecosystem.

Pennsylvania Shows Both Sides of the AI Campus Race

Pennsylvania is also emerging as a useful test case for the next phase of AI infrastructure development, where large powered sites are being positioned as regional growth engines even as other communities push back against the scale and local impact of proposed data center clusters.

Switch announced plans to develop a new 382-acre data center campus in Big Beaver Borough, Pennsylvania, expanding its Prime campus portfolio into the greater Pittsburgh area.

The Beaver County project is positioned around connectivity as well as power. Switch said the campus sits at the intersection of key East-West and North-South fiber routes, with low-latency access to enterprise, healthcare, finance, higher education and government users across the Eastern United States.

The company also framed the project around infrastructure responsibility, saying it is committed to funding the infrastructure required for its power needs in a way that does not pass costs onto consumers. Switch said its EVO data centers use a closed-loop system that recycles water, with zero water consumed to cool servers and GPUs and only minimal water required for office and warehouse functions.

“Beaver County helped build America’s industrial foundation, and this campus represents a new chapter for the Borough of Big Beaver,” said Natalie Stewart Mitchell, SVP of Government Affairs & Campus Development at Switch.

Mitchell said Switch plans to serve as “a long-term partner to Beaver County” through infrastructure investment and local initiatives designed to support the regional economy.

The project will join Switch’s existing Prime campus portfolio in Las Vegas, Tahoe Reno, Atlanta, Grand Rapids and Austin.

But elsewhere in Pennsylvania, the same fundamentals driving AI data center siting — power access, land availability and water — are producing sharper local conflict.

Tom’s Hardware, citing Washington Post reporting, detailed intense pushback in Archbald, Pennsylvania, where developers have proposed six data center campuses in a town of roughly 7,000 people. The proposals reportedly include 51 data warehouse buildings across a 17-square-mile community, with the total footprint compared to about 51 Walmart Supercenters.

Archbald’s appeal is clear from an infrastructure perspective. A major 500-kV transmission line, the Susquehanna-Roseland line, runs through the town, connecting the 2.5-GW Susquehanna nuclear plant to New Jersey. Combined with available land and water, that transmission access has made the former coal-region community attractive to AI data center developers.

The local response shows how quickly powered-land logic can collide with community politics. Tom’s Hardware reported that four of Archbald’s seven town council members have resigned, along with several planning board members, as the data center fight intensified. Residents have raised concerns over utilities, noise, light pollution, land use and broader quality-of-life impacts.

Together, Big Beaver and Archbald capture the two sides of the current Pennsylvania opportunity. In one case, a major campus is being presented as a managed infrastructure investment in a former industrial region. In the other, clustered data center proposals have triggered a local political backlash over scale, process and community control.

For developers, the lesson is increasingly hard to miss: powered land may open the door, but local legitimacy determines how far expansion can actually go.

AI Infrastructure Expands Across the U.S. South

The next layer of the AI infrastructure map is taking shape across the U.S. South, where smaller cities, power-rich industrial sites, and emerging data center markets are increasingly being pulled into the orbit of frontier AI compute.

That trend is visible in Alabama, Georgia, Texas and Oklahoma, where recent announcements show the range of what “land and expand” now means: purpose-built AI colocation, hyperscale cloud region growth, gigawatt-scale campuses, jurisdictional siting battles, and the continued conversion of former crypto infrastructure into AI capacity.

Cerebras Anchors 40 MW AI Campus in Alabama

Digi Power X announced a 10-year master services agreement with Cerebras Systems for a purpose-built 40 MW AI data center campus in Columbiana, Alabama, marking another example of AI compute demand moving into power-positioned secondary markets.

The agreement is valued at approximately $1.1 billion over its initial term, with total potential contract value of up to $2.5 billion including renewal terms.

Digi Power X said the facility will be developed in two phases, beginning with 15 MW of IT load followed by an additional 25 MW, for a combined 40 MW. The company said the campus will be built to Tier III infrastructure standards and optimized for the high-density thermal requirements of next-generation AI accelerator hardware.

Construction on Phase 1 is expected to begin immediately, with ready-for-service targeted for Dec. 15, 2026. Full deployment is targeted by the end of Q1 2027.

The Columbiana site was selected for access to robust power infrastructure, a favorable regulatory environment, and proximity to major fiber corridors serving the Southeast. Digi Power X said it owns the underlying real property and has already completed construction of the dedicated onsite substation serving Phase 1, with grid interconnection finalized and a power delivery agreement in place with Alabama Power.

“This agreement is transformational for Digi Power X,” said Michel Amar, chairman and CEO of Digi Power X. “Signing a $1.1 billion anchor contract with a premier AI compute company is validation of everything we have built — our team, our sites, our infrastructure capabilities, and our vision for what a next-generation data center operator looks like.”

Alec Amar, president of Digi Power X, called the agreement “a statement,” adding that the contract “signals Digi Power X is a serious player operating at the highest level.”

The customer itself adds another layer of significance. Cerebras is entering the public markets at a moment when investor attention around AI semiconductor companies is intensifying. Fast Company reported that Cerebras priced its IPO at $185 per share, well above its earlier expected range of $115 to $125, raising approximately $5.5 billion and valuing the company at about $56.4 billion.

Fast Company noted that Cerebras has built its position around wafer-scale AI chips designed for high-throughput AI workloads, with customers including AstraZeneca, GlaxoSmithKline, G42, IBM, Meta, Mistral, Notion and Perplexity. The publication also reported that Cerebras recently inked a $20 billion deal with OpenAI.

For data center operators, the Cerebras-Digi Power X agreement underscores a broader shift: AI accelerator demand is no longer only a hyperscaler campus story. It is also creating opportunities for specialized operators that can deliver powered, thermally capable, fiber-connected AI capacity on accelerated timelines.

Microsoft Adds Tyrone to Its Georgia Data Center Map

Microsoft has confirmed plans for a new data center in Tyrone, Georgia, extending its already significant footprint around metro Atlanta.

According to Data Center Dynamics, Microsoft announced plans to hold a community information session on May 19 to discuss the project’s energy usage, water and sustainability, construction timeline, site design, community investment, workforce pathways, and jobs.

The Tyrone project had been known locally for several months before Microsoft publicly acknowledged it. DCD reported that Microsoft previously received zoning permission to develop a data center on a 147-acre parcel along North Highway 74 between Kirkley Road and the northern city limits. The project is expected to include two data center buildings totaling about 500,000 square feet, along with a substation.

The development adds to Microsoft’s expanding Georgia platform. DCD noted that Microsoft is already developing projects in Douglasville, Palmetto and East Point, all near Atlanta, and has announced plans for its East US 3 cloud region in Atlanta in 2027. The company also owns a 347-acre property in Rome, Georgia, purchased in 2023 for a future data center campus.

The Tyrone outreach also reflects a newer phase of hyperscale development politics. Microsoft’s own project update frames the site under its “Community-First Infrastructure” approach, with an emphasis on transparency, local engagement and long-term partnership.

That positioning matters as data center growth across the South increasingly intersects with questions around water, energy, noise, aesthetics, trees, emissions, jobs and local control.

Texas Sees More Gigawatt-Scale Campus Activity

Texas continues to function as one of the central theaters for AI-era land and power development.

Data Center Dynamics reported that CloudBurst Data Centers has received approval for a 706-acre data center campus outside San Antonio, after the Guadalupe Commissioners Court approved a development agreement and a tax abatement worth $500 million.

The project is planned for a site near San Marcos and is expected to include between 10 and 12 data center buildings totaling roughly 3 million square feet. CloudBurst plans to invest $14.5 billion in the campus, which is designed for 1.2 GW of capacity at completion.

According to DCD, the project is expected to be built in three 400 MW phases, with the first phase beginning in 2026 and finishing in 2027. An initial 50 MW phase has already broken ground and is expected to go live in Q4 2026.

CloudBurst has said the campus will use closed-loop cooling and behind-the-meter generation, and estimates the project could generate roughly $120 million in annual tax revenue for the community and county.

But the approval also illustrates the rising political friction around large-scale AI infrastructure development. DCD reported that both the tax abatement and development agreement passed by a one-vote margin, with many attendees speaking against the proposal. Concerns included environmental impact and potential health effects, and a petition opposing the data center had gathered more than 1,200 signatures.

A separate DCD report shows similar tension around another major Texas site. PowerHouse Data Centers is pursuing a facility in Caldwell Valley Technology Park near Austin, a Tract-backed powered land site that reportedly can support up to 4 GW of capacity.

PowerHouse has filed a petition to release the proposed facility from the jurisdiction of Uhland, Texas, preferring instead to bring the proposal before Caldwell County. DCD noted that Texas counties generally have less authority than cities to regulate and manage data center projects, particularly around planning and zoning.

That maneuver highlights one of the defining tensions of the current land rush: as power-ready sites become more valuable, jurisdictional strategy is becoming part of data center development strategy.

PowerHouse already has a growing Texas presence, including a 201 MW data center in Irving and a 768-acre campus with Provident Data Centers south of Dallas in Ellis County. DCD reported that the company’s broader development pipeline totals more than 8 GW across multiple states.

Core Scientific Extends the Crypto-to-AI Conversion Playbook

In Oklahoma, Core Scientific is expanding its Muskogee campus by acquiring adjacent crypto infrastructure and its associated grid capacity, continuing a broader strategy of converting bitcoin-era infrastructure into large-scale AI data center capacity.

According to reporting from Data Center Dynamics, Core Scientific has agreed to acquire Polaris DS LLC, a bitcoin mining firm with 440 MW of gross power contracted under an energy agreement with Oklahoma Gas & Electric.

The transaction includes 40 acres adjacent to Core Scientific’s existing Muskogee facility, an onsite substation, and electrical service agreements supporting up to 440 MW of capacity. The acquisition is valued at approximately $421 million in cash and is expected to close in Q3 2026.

Core Scientific is currently developing a 70 MW facility in Muskogee leased to CoreWeave. DCD reported the building is designed to support NVIDIA GB300 systems and is progressing through final testing and commissioning, with delivery expected in Q2 2026.

The company now aims to expand the Muskogee campus to 1.5 GW of gross power, representing approximately 1 GW of leasable capacity. Construction has already begun on a second, currently unleased 82.5 MW building, with delivery expected in Q4 2027.

But Muskogee is not the company’s only large-scale AI expansion effort.

In April, Core Scientific separately announced plans to scale its Pecos, Texas campus to approximately 1.5 GW of gross power through a combination of additional utility capacity and a scalable behind-the-meter power strategy.

Earlier this year, the company began transforming the Pecos site — where 300 MW of gross power capacity is currently used for bitcoin mining — into a campus designed for AI infrastructure and high-density colocation services.

Core Scientific said it has already secured an additional 300 MW of gross power capacity under contract with its utility provider while also developing a behind-the-meter expansion plan expected to provide substantial additional capacity beyond its previously announced leasable power pipeline.

To support the expansion, the company has secured more than 200 acres of land at the Pecos site.

“We continue to leverage our deep in-house expertise to differentiate how we build and scale next generation artificial intelligence infrastructure, further illustrated by our behind-the-meter solution at our Pecos campus,” said Adam Sullivan, CEO of Core Scientific.

“By expanding in a market where we already control power, infrastructure, and operations, we can execute with speed to meet market demand,” Sullivan added.

Taken together, Core Scientific's Muskogee and Pecos expansions reinforce one of the defining infrastructure adaptations of the AI era: crypto mining operators with existing substations, power contracts, operational expertise and industrial land positions are increasingly repositioning themselves as AI infrastructure developers.

Across multiple markets, former bitcoin infrastructure is rapidly becoming AI infrastructure.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.