Five AI Data Center Power Landlords Win UBS Buy Ratings on 4.8 GW Backlog
UBS has opened coverage on five North American AI data center power landlords, awarding Buy ratings across the group on the argument that energized, permitted megawatts now constrain artificial intelligence expansion more tightly than the chips that fill the racks.
The investment bank initiated on Hut 8, TeraWulf, Applied Digital, Core Scientific and Cipher Digital on September 22, 2026. Contracted IT capacity across the five companies totals roughly 4.8 GW, which UBS ties to about $125 billion in potential revenue.
Price targets diverge sharply, reflecting how far each operator has moved from bitcoin mining toward hyperscale colocation.
| Company | Ticker | UBS price target |
|---|---|---|
| Hut 8 | HUT | $143 |
| TeraWulf | WULF | $24 |
| Applied Digital | APLD | $38 |
| Core Scientific | CORZ | $24 |
| Cipher Digital | CIFR | $23 |
Applied Digital offers the most concrete near-term reference. Its shares trade near $28.54, so the $38 target implies roughly 33% upside. Hut 8 carries the highest absolute target at $143, while the other three cluster between $23 and $24.
The ratings land at a moment when the AI infrastructure trade has split in two. Hyperscalers and chipmakers trade on earnings growth, while the AI data center power landlords supplying land and power trade on the value of their permits and interconnects. UBS is underwriting the second group as an asset class rather than a technology bet.
The core of the UBS thesis is a change in what drives the value of these companies. Secured utility interconnections and long-term colocation leases with hyperscale tenants now set the equity story, displacing the hash-rate economics that defined the group's mining era. UBS treats time-to-market for energized clusters as the primary competitive advantage, because a signed lease generates nothing if the power arrives years late.
The re-rating has an arithmetic basis. Bitcoin miners were valued on hash rate and power cost per coin, both of which swing with a volatile commodity price. Colocation landlords are valued on contracted revenue from tenants with investment-grade balance sheets, which supports longer-duration cash flows and lower discount rates. Converting the same substation from mining to AI hosting changes the multiple applied to identical electrons.
Why AI Data Center Power Landlords Now Trade on Megawatts
The size of the prize explains the interest. Projections put global data center IT load at 104 GW annually by 2030, against the 4.8 GW this group has under contract. Broader forecasts place total data center capacity at 219 GW by 2030, a 14% compound annual growth rate.
Those figures also bound the bull case. Four-point-eight gigawatts amounts to less than 5% of the projected 2030 load, so the investment argument rests on scarcity pricing at the margin rather than market dominance. When the final megawatts in a constrained grid region decide whether a training cluster ships on schedule, whoever holds them sets the terms.
The electrical profile of AI workloads has changed what these buildings are worth. A conventional cloud or enterprise hall runs racks at 5 to 10 kW. A rack filled with AI accelerators draws 40 to 130 kW or more, and newer systems push higher still. Multiplied across thousands of racks, a single AI campus can demand 100 MW to more than 1 GW, the power footprint of a mid-sized city.
That intensity has repriced land. Site costs in Northern Virginia and the Northeast passed $8 million per acre, and a developer in Loudoun County, Virginia, offered $4.4 million per acre for one parcel. In some markets, entitled power-ready land has climbed nearly tenfold in under two years. Data center land deals rose 79%, drawing roughly $6 billion in early 2026.
The land banks are the quiet asset in this story. Parcels assembled for mining at industrial or agricultural prices now sit in markets where powered sites clear at multiples of their purchase cost, a revaluation that never shows up in contracted revenue figures.
The category itself is a conversion story. Each of these companies already owned substations, high-voltage interconnects and industrial land assembled during the crypto boom. What changed is the tenant: instead of mining bitcoin against a floating spot price, they lease energized shells to buyers who sign decade-long contracts.
Utility interconnection is the gating item. For a CIO signing a colocation contract, the delivery date is set by the utility rather than the landlord, which turns grid queues into leverage for whoever already holds an approved connection.
Demand on the tenant side follows the same physics. OpenAI plans to lease an Ohio data center built by SB Energy, a SoftBank subsidiary, securing 8 GW of capacity with the first 800 MW expected online in 2028.
The alternative for a hyperscaler is to build. That route is slower and more capital-intensive, and it competes for the same transformers, switchgear and skilled construction crews already in short supply. Equipment lead times, not land, now set many project schedules.
Efficiency is the counterweight. Every gain in accelerator performance per watt lowers the megawatts a given workload requires, and each algorithmic improvement that cuts inference cost does the same. The demand forecasts behind these price targets assume the current relationship between compute growth and power draw holds. If that relationship loosens, the scarcity premium attached to energized land compresses.
The Capital Gap Behind the Backlog
UBS flagged three risks alongside its Buy ratings: high leverage, construction delays, and roughly $24 billion of additional unfunded capital required to deliver the contracted pipeline.
That capital gap is the crux of the trade. A $125 billion revenue pool tied to 4.8 GW of capacity means little if the equity cannot be raised at a workable cost. The same operators carry debt from their mining years, and the lenders pricing that debt now watch construction schedules as closely as lease signings.
Some are assembling their own generation. The Cornerstone acquisition closed on June 15, 2026, adding about 2.6 GW of gas capacity at Waterford, Darby and Lawrenceburg, alongside a pipeline of roughly 4 GW in land development and data center contracting options.
Owning generation reduces exposure to interconnection delays. It also adds fuel price risk, emissions obligations and a longer permitting runway to the balance sheet.
Utilities are collecting part of the same value through regulated returns on the transmission and generation built to serve these loads, which gives investors a second route into the power bottleneck. The difference is where the risk sits. A regulated utility earns a set return on invested capital, while a landlord earns the spread between lease revenue and its cost of capital. When borrowing costs rise, that spread compresses before lease terms reset.
Tenant concentration cuts the other way. The counterparties are a handful of hyperscalers large enough to fund their own campuses, negotiate hard on lease terms and abandon projects that slip. That is a different risk from an office landlord's vacancy problem. It is not smaller: a single tenant's decision to self-build can erase years of backlog.
Execution risk compounds. A delayed substation pushes revenue into later years, raises the carrying cost of the land and hands negotiating leverage to the tenant. For a group that must return to capital markets repeatedly to fund 4.8 GW, each slip raises the price of the next raise.
Regional politics add another variable. Data centers have transformed rural land markets, and opposition to energy-intensive facilities has hardened in several jurisdictions, lengthening approval timelines for generation and transmission alike.
Wall Street's appetite for the build-out is not unlimited. Several companies tied to the data center industry have delayed initial public offerings amid rising public opposition to these facilities, which narrows an exit route for private developers and raises the cost of the next equity round.
The near-term catalysts are narrow. New lease signings, stronger tenant covenants and progress on energizing already-contracted capacity are the variables UBS identifies as capable of pushing the shares higher, and each is measurable quarter by quarter.
Why this matters
The five ratings mark a shift in how the AI build-out gets financed. The bottleneck has moved from silicon to substations, and the AI data center power landlords holding approved interconnects now carry both the upside of scarcity pricing and the balance-sheet risk of a pipeline that needs roughly $24 billion it has not yet raised.
For investors, the group is a leveraged bet on the same hyperscaler capital budgets that already drive the chip trade, with land and power standing in for the volatile commodity exposure of their mining past.
Related Articles
- The Vertiv UIG Acquisition Puts a $2.6B Price on AI's Power Problem
- Data Center Opposition Meets Wall Street: Why $198B in Stalled Projects Is Now an IPO Problem
- Phantom AI Data Center Demand: Why Only 28% of Requested Power Will Materialize
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.