Fusion Energy and the Next Era of Data Center Real Estate

Commercial fusion is no longer a 2050 story. The first contracted fusion electrons are now scheduled to reach a Microsoft data center in 2028, and Google’s 200 megawatt offtake agreement with Commonwealth Fusion Systems is already reshaping land-use decisions in Chesterfield County, Virginia. For net lease investors, 1031 exchangers, and DST sponsors focused on industrial and data center real estate, fusion has quietly become a near-term pricing variable rather than a distant thesis.

This pillar connects the dots that most CRE investors are missing: how hyperscaler fusion power purchase agreements (PPAs) are reshaping data center site selection, why those site decisions create a new industrial real estate category, and where the 1031 and DST opportunities are opening up over the next five to ten years.

The Power Bottleneck Behind the AI Buildout

To understand why hyperscalers are contracting for fusion power years before the first reactor delivers a single commercial megawatt, start with the numbers behind the AI compute cycle.

US utility power provided to data centers reached roughly 62 gigawatts in 2025 and is projected to climb to approximately 76 gigawatts by the end of 2026. Globally, critical power supporting data centers is expected to approach 96 gigawatts in 2026, with AI workloads accounting for more than 40 percent of that demand. The International Energy Agency projects worldwide AI data center electricity consumption will reach 90 terawatt hours annually by 2026, roughly a tenfold increase from 2022.

Hyperscaler capital expenditure is scaling to match. The Big Five (Amazon, Microsoft, Google, Meta, Oracle) are projected to spend more than $600 billion on infrastructure in 2026, a 36 percent increase over 2025, with roughly 75 percent of that pointed at AI buildouts. Amazon alone guided to approximately $200 billion in 2026 capex. Google is targeting around $180 billion. Data center capex is forecast to cross the $1 trillion threshold for the first time this year.

The constraint is not chips. It is not capital. It is power. Sightline Climate’s 2026 Data Center Outlook estimates that 30 to 50 percent of announced 2026 data center pipeline may not come online before year end, and the primary bottleneck is interconnection and generation capacity. Hyperscalers are responding by moving upstream: contracting directly with generators, acquiring renewable pipelines outright (Google acquired Intersect Power’s 10.8 gigawatt portfolio), and now reaching further into the technology stack by signing PPAs for power sources that do not yet commercially exist.

That is the context in which fusion PPAs make sense. They are not speculation. They are option value on grid-scale firm carbon-free power at a moment when traditional options (natural gas peakers, new nuclear, utility-scale renewables plus storage) all face permitting, supply chain, or interconnection constraints that make their delivery timeline as uncertain as fusion’s.

Why Hyperscalers Are Signing Fusion PPAs

Three hyperscalers have now publicly committed to fusion offtake. A fourth, Meta, has issued a request for proposals. The structure of each deal signals where fusion sits in their energy procurement stack.

Microsoft and Helion: The First Fusion Offtake

Microsoft signed the world’s first commercial fusion PPA with Helion Energy in May 2023. The agreement calls for Helion to begin delivering at least 50 megawatts to Microsoft starting in 2028 from its Orion plant in Malaga, Washington, with Constellation Energy serving as power marketer. Helion broke ground on the Malaga site in July 2025 and received a Conditional Use Permit from Chelan County in October 2025. The facility sits on land leased from the Chelan County Public Utility District and will interconnect upstream of Microsoft’s Washington state data center footprint on the Columbia River corridor.

Helion has raised more than $1 billion through a January 2025 Series F at a $5.4 billion valuation, with Sam Altman as a major backer. Its seventh-generation prototype Polaris is expected to demonstrate the first privately produced electricity from fusion, a milestone that underwrites the commercial path to Orion.

For real estate investors, the Microsoft Helion deal matters because it pre-commits the buyer. The power does not need to clear the wholesale market. Microsoft has obligated itself to take the electrons, which means the Malaga site is effectively a captive generating asset for a single hyperscaler tenant base.

Google and Commonwealth Fusion Systems: The Virginia Play

Google signed a 200 megawatt PPA with Commonwealth Fusion Systems in mid 2025 and expanded its equity stake in the company at the same time. The offtake will come from CFS’s ARC plant at the James River Industrial Park in Chesterfield County, Virginia, a 94 acre site leased from Dominion Energy. Chesterfield County’s Board of Supervisors unanimously approved the conditional use permit in September 2025. Total plant capacity is planned at 400 megawatts, with first power targeted for the early 2030s. Google’s 200 megawatt slice represents half the plant’s output.

In parallel, Virginia Governor Glenn Youngkin announced in August 2025 that Google would invest an additional $9 billion in Virginia through the end of 2026, with much of that funding directed at a new data center in Chesterfield near Meadowville Technology Park. That is not coincidence. Google is anchoring data center development proximate to the fusion plant it has contracted to take power from, and doing it inside the Dominion Energy service territory that hosts the ARC site.

CFS has raised more than $2 billion privately, received $16.5 million in Department of Energy grants, and has guided plant construction cost to more than $2.5 billion. Its demonstration machine SPARC, housed in Devens, Massachusetts, is expected to achieve first plasma in 2026 (this year) and net fusion energy shortly after.

The Second Wave

Amazon’s fusion positioning is still developing, with the company focused for now on small modular reactor and traditional nuclear offtake (including the 960 megawatt Talen Energy PPA in Pennsylvania). Meta has signaled fusion interest through a 2025 RFP. Oracle, the fifth hyperscaler, has prioritized gas and nuclear to date.

Expect fusion PPA announcements to compound through 2026 and 2027 as SPARC and Polaris deliver (or fail to deliver) on their near-term technical milestones. Every successful test accelerates offtake commitments, and every offtake commitment narrows the near-term siting map.

How Fusion PPAs Reshape Data Center Site Selection

Hyperscaler site selection historically optimized for five variables: fiber connectivity, latency to users, power cost, power availability, and tax incentives. AI workloads have rewritten the weighting. Power availability now dominates. Latency matters less for training workloads. And power availability increasingly means contracted firm clean power, because corporate net-zero commitments and investor pressure make uncontracted grid power strategically fragile.

Fusion PPAs introduce a sixth variable: proximity to a contracted generating site. If Google has committed to 200 megawatts from Chesterfield, Virginia, the economics strongly favor locating compute load nearby to minimize transmission losses, interconnection risk, and the accounting complexity of matching hourly carbon free energy. The same logic applies to Microsoft in central Washington, where the Columbia River data center corridor is already one of the densest hyperscaler footprints in North America and where Helion’s Orion site sits directly on existing transmission assets.

For CRE investors, this creates a new type of industrial land play. The traditional data center submarket map (Northern Virginia, Silicon Valley, Dallas, Phoenix, Columbus, Chicago, Atlanta) is now being supplemented by a second map defined by fusion and advanced nuclear offtake sites. Chesterfield County and Chelan County are the first two pins. Others will follow as the next wave of fusion PPAs gets signed, most likely in PJM (Virginia, Pennsylvania, Ohio), ERCOT (Texas), and the Pacific Northwest.

Credit Tenant Analysis: Hyperscaler Counterparty Strength

From a net lease credit perspective, hyperscalers with contracted fusion offtake are arguably strengthening their long-term tenancy profile, not weakening it. Here is why.

The worst-case scenario for a 20 year NNN lease on a hyperscaler-anchored data center is that the tenant’s operational footprint becomes obsolete before lease expiration. The likeliest cause of that obsolescence is a power constraint that forces workload migration. A hyperscaler with a contracted fusion PPA at a specific site has signaled multi-decade commitment to that geography. The PPA is a lease-lengthening instrument even though it sits on a different balance sheet line item.

From a credit rating perspective, all four publicly contracted fusion hyperscalers (Microsoft, Google, Meta via RFP, and Amazon via nuclear adjacency) carry investment grade ratings with deep capital market access. Microsoft and Google specifically remain among the highest-rated corporate credits in the world. Counterparty risk on a 20 year net lease to Microsoft or Google at a data center tied to a fusion offtake site is as close to sovereign-equivalent risk as commercial real estate offers.

The NNN investor’s caution is not credit. It is obsolescence. And contracted firm clean power is the strongest available hedge against obsolescence in the AI data center category.

Industrial Real Estate Plays Near Announced Fusion Sites

Two geographies now offer concrete investable theses. More are coming.

Chesterfield County, Virginia

Chesterfield sits south of Richmond in the Dominion Energy service territory, inside the PJM interconnection zone that already hosts Northern Virginia’s data center cluster in Loudoun County. The James River Industrial Park (the ARC site) is a 94 acre parcel, but the surrounding industrial inventory and Google’s planned Meadowville Technology Park data center investment suggest a broader submarket build-out.

Investable thesis for the submarket:

1. Industrial flex and light manufacturing within a 30 mile radius of the ARC site, targeting component suppliers, cooling equipment vendors, and fusion-adjacent service providers
2. Truck terminal and last-mile logistics assets serving the data center construction cycle
3. NNN-leased office and R&D assets serving fusion engineering firms and Dominion Energy’s expanded workforce

Cap rates in Chesterfield industrial have already compressed relative to broader Virginia industrial averages. The question for investors is whether to enter now on submarket momentum or wait for the next leg of compression as ARC moves from conditional use permit (2025) to construction (2026 to 2027) to operation (early 2030s).

Chelan County, Washington

Chelan County, and specifically the Malaga area outside Wenatchee, sits on the Columbia River corridor that already hosts hyperscaler operations for Microsoft, Yahoo, and several colocation operators. The area has been attractive for decades because of Bonneville Power Administration hydroelectric access. Helion’s Orion site adds a second pillar of firm clean generation.

Investable thesis for the submarket:

1. Raw industrial land with transmission access inside the Chelan County PUD footprint
2. Existing flex and warehouse product in Wenatchee and East Wenatchee serving data center operations
3. Workforce housing exposure, which benefits both from fusion plant construction jobs and expanded data center operations

The Central Washington market remains thinly traded compared to Northern Virginia, which creates both the opportunity (low competition) and the caution (less liquidity, smaller comp set).

What to Watch Next

The next fusion siting announcements will likely surface in:

• ERCOT Texas (the combination of permissive permitting, energy sector labor force, and hyperscaler data center demand makes it a natural second wave host)
• PJM Pennsylvania and Ohio (deep industrial land supply adjacent to existing nuclear and gas infrastructure)
• MISO Illinois and Indiana (Meta’s strong presence and existing nuclear operator relationships)

Investors watching the space should track conditional use permit applications in industrial parks within Dominion, PSE, PG&E, CenterPoint, and Ameren service territories. Fusion companies will not publicly announce sites until permitting is in motion, but local planning commission agendas reveal the early signal.

Lease Structure Implications

If a data center is contracted to take power from a fusion plant with a 20 year operating design life, the natural lease term for the anchor tenant extends. This matters for structuring NNN and DST investments.

Traditional hyperscaler data center leases have ranged from 10 to 15 years with multiple renewal options. Fusion-anchored sites will likely see base lease terms migrate toward 15 to 20 years with longer renewal options, because the tenant’s power infrastructure investment and the generator’s capital recovery schedule both favor longer commitments.

For NNN investors, longer base terms compress cap rates (good for sellers, compressing yield for buyers), reduce re-leasing risk, and improve loan terms on CMBS and agency debt. For DST sponsors, longer lease terms align well with the typical DST hold period and simplify the exit analysis because the residual value risk shifts further into the future.

The caution: longer lease terms also lock in any structural mistakes. A data center built to 2026 power density and cooling specifications may need capital-intensive retrofits to remain competitive by 2036. Tenant improvement allowances and reinvestment clauses matter more in a fusion-anchored lease than in a traditional big box NNN.

Cap Rate Implications for Hyperscaler NNN

Cap rate compression in hyperscaler NNN has been in motion since 2022. Fusion offtake is a new accelerant.

The logic works in two ways. First, contracted firm clean power reduces operational risk at the tenant level, which reduces the credit premium the market demands. Second, fusion adjacency creates locational scarcity. There will be a small number of fusion-adjacent data center submarkets in 2030, and scarcity in real estate translates directly to pricing power.

Expect to see cap rates on hyperscaler-anchored NNN data centers in Chesterfield and Chelan submarkets trade at 50 to 100 basis points inside the broader hyperscaler NNN universe within 24 to 36 months of plant construction commencement. If you are a seller, this is a window. If you are a buyer, the question is whether to accept the compression now or wait for the next cycle.

Tax Strategy: 1031 Exchanges and DSTs Into Data Center Adjacent Industrial

For 1031 exchangers, fusion-adjacent industrial real estate offers a rare combination: growing submarket fundamentals, investment grade tenancy exposure, and long-duration lease structures. The tax strategy angle becomes compelling because the underlying real estate thesis stands on its own even before considering the tax benefits.

Three 1031 and DST pathways are opening up:

1. Direct acquisition of industrial or flex assets near announced fusion plant sites. This is the most capital intensive path but offers the most control and upside.
2. DST sponsorship and investment in data center adjacent industrial portfolios. Several national sponsors have begun assembling data center adjacent industrial DSTs, and fusion siting will likely accelerate that trend. For accredited investors, DSTs offer a way to deploy 1031 proceeds into the thesis without taking active management responsibility.
3. Build-to-suit development partnerships with hyperscaler suppliers. This is the most entrepreneurial path but offers the tightest alignment with the underlying demand driver.

For a deeper walk through the 1031 and DST mechanics specific to this thesis, see the cluster article on 1031 exchanges into fusion-adjacent industrial properties (coming soon).

What Could Go Wrong

Intellectual honesty requires flagging the risks. Fusion commercialization is a first-of-kind engineering project, and first-of-kind engineering projects run over budget and behind schedule. SPARC first plasma in 2026 is still an expectation, not a certainty. Helion’s Polaris prototype still needs to demonstrate commercial relevance. ARC’s early 2030s grid delivery date assumes permitting, construction, and operational milestones all land on schedule.

If SPARC or Polaris miss their near-term milestones by two years or more, fusion PPAs will still matter but the hyperscaler site selection logic may soften. The stronger risk is to the generator side (Helion, CFS) rather than the tenant side (Microsoft, Google), which is exactly the structural advantage of investing in real estate proximate to these sites rather than in the fusion developers themselves. You are taking land-use risk, not technology risk.

The second risk is transmission and interconnection. A fusion plant that delivers electrons on schedule but cannot clear PJM or WECC interconnection queues delivers less strategic value to its offtake partner. This is a grid infrastructure problem, not a fusion problem, but it affects the same CRE thesis.

The third risk is political. Fusion siting enjoys bipartisan support today. That could change, particularly at the state level where permitting authority sits. Virginia and Washington both remain friendly, but investors should watch state-level politics in any new siting geography.

The Investor Playbook for 2026 to 2030

If the thesis here is right, the investor playbook breaks into three phases.

Phase 1 (2026 to 2027): Accumulate. Build position in Chesterfield County, Virginia and Chelan County, Washington industrial, flex, and adjacent workforce housing. Watch for new fusion siting announcements in Texas, Pennsylvania, Ohio, Illinois, and Indiana. Build relationships with the fusion developers’ supplier ecosystem.

Phase 2 (2028 to 2030): Harvest. First fusion electrons reach Microsoft. SPARC moves to ARC construction. Cap rates compress meaningfully in fusion-adjacent submarkets. Sellers should evaluate exit timing. Buyers should transition to Phase 3 geographies.

Phase 3 (2030 and beyond): Scale. Second-generation fusion plants begin to enter permitting in new geographies. The thesis becomes mainstream. Competitive advantage shifts to operators who built relationships and expertise in Phases 1 and 2.

Frequently Asked Questions

This is a pillar article in our Data Center Net Lease cluster. For the companion deep dives on credit tenant analysis, submarket plays, lease structure, and 1031 and DST mechanics, see our related coverage.