AI/HPC colocation stocks: UBS launches Buy coverage of five AI colocation operators, arguing that scarce power access creates substantial pipeline optionality beyond contracted capacity.

Overview

UBS initiates coverage of five US-listed AI/HPC colocation operators transitioning from bitcoin-mining infrastructure. Its central thesis is that pre-existing access to land, grid interconnections and power gives these companies a scarce asset as AI demand accelerates, while current valuations largely recognize contracted leases rather than future pipeline conversion.

Core views

UBS argues that AI infrastructure demand is increasingly constrained by access to power rather than by demand for compute. Its AI Power Model projects incremental global data-center IT load of about 25GW in 2026, 70GW in 2028 and 100GW by 2030, with roughly 260GW of incremental capacity required during 2028-30, around three times the estimated 90GW required in 2025-27. Hyperscale capital expenditure is projected at about $5.5 trillion from 2026-29E. In North America, UBS estimates that about 43GW of new capacity has been initiated over the prior 12 months, including approximately 25GW pre-leased through third parties and 17GW of hyperscale self-builds. Third-party leasing has risen to about 60% of new demand from about 40% in prior years, in UBS's view, as tenants seek speed to power. The report's supply-side argument centers on grid scarcity. Traditional US data-center markets have vacancy below 2% and grid interconnection timelines exceeding five years. The five covered companies—Hut 8, TeraWulf, Applied Digital, Core Scientific and Cipher Digital—originated in bitcoin mining and therefore secured power blocks, interconnection positions, land and utility relationships before the AI buildout. UBS estimates that this has translated into about 4.8GW of leased AI/HPC colocation capacity, with roughly 12GW of additional development pipeline and expansion opportunities. These sites can offer speed to market, though AI/HPC facilities require far higher construction spending and technical specifications than mining facilities: roughly $10-15M/MW of build cost, 99.98%+ uptime, redundancy and advanced cooling, versus roughly $0.5-1.0M/MW for bitcoin-mining facilities. UBS expects the tenant base to broaden from neoclouds toward hyperscalers, AI labs and semiconductor companies. It notes that miner-backed operators have already signed leases with Amazon, Meta, Oracle, NVIDIA and AMD. Higher-quality counterparties can reduce financing and cash-flow risk, although ultimate AI demand can remain concentrated across a smaller set of AI labs. The institution sees current compute pricing as evidence of capacity scarcity: it estimates near-immediate compute deals at $30-50M/MW annually, while colocation rates have remained around $1.5-2.0M/MW annually. This difference suggests cloud providers capture substantial margins and may leave room for colocation pricing to rise. For the covered group, UBS forecasts average EBITDA growth of about 80% annually from 2027E through 2029E, about 10% above Street estimates in 2029E. It argues that shares trade at 12-18x 2028E EBITDA and largely value already contracted assets, assigning limited value to uncontracted pipeline capacity. The report identifies six catalysts: on-time delivery execution, clearer power pipelines, new lease signings, declining risk premiums as assets begin producing contracted cash flow, REIT conversion, and strategic interest arising from the valuation gap between physical data-center infrastructure and the compute deployed within it. The group’s share prices were down about 45% from recent highs despite being up about 45% over the prior 12 months; UBS attributes the pullback to AI-demand concerns, NIMBY and regulatory uncertainty, Texas power risk, higher yields and weaker appetite for leveraged AI stocks. UBS favors companies with demonstrated pipeline conversion, strong credit support and lease protections, and diversified power portfolios. Hut 8 is highlighted for its high-quality leases, roughly 2.8GW IT pipeline and estimated $2.3B of 2029E revenue. UBS assigns a Buy rating and $143 target, implying 45% upside. TeraWulf is favored for its brownfield development strategy, 840MW of signed leases and potential to increase contracted capacity by about 50% through Muskie; UBS assigns a Buy and $24 target. Applied Digital has the strongest leasing momentum, with 1.41GW of signed IT leases and a 2.7GW IT pipeline; UBS expects tenant quality to improve materially and assigns a Buy and $38 target. Cipher Digital has about 493MW of signed leases and near-term sites that could raise contracted capacity by about 70%, but about 95% of its portfolio is in Texas; UBS assigns a Buy and $23 target. Core Scientific has about 1.1GW of signed leases, is furthest along in its AI/HPC transition, and has 1.9GW of AMD expansion options; UBS assigns a Buy and $24 target. Execution is the immediate gating issue. Only about 720MW, or roughly 15% of group contracted capacity, had been delivered at the time of the report; UBS expects another 1.6GW by mid-2027, taking delivered capacity to about 2.3GW or 50% of leased capacity. The report expects 45-60 day ready-for-service delays to remain common because of labor constraints. Longer delays can produce escalating penalties, rent abatements, termination rights or tenant step-in rights. This is especially material where investment-grade lease backstops begin only after ready-for-service. Texas is another major uncertainty: ERCOT's Batch Zero process may affect allocation and energization timing, with final designations expected December 10, 2026 and preliminary Studied Load allocation results expected in April 2027. UBS values the group using EV/EBITDA and a sum-of-the-parts framework. It uses lower-to-mid historical data-center REIT multiples, generally 15-22x EBITDA, to reflect delivery risk, tenant concentration, non-traditional locations and early lease cash-conversion differences. The SOTP gives roughly 50-60% weight to stabilized contracted assets and 40-50% to probability-weighted pipeline optionality. Contracted lease portfolios are valued using cash NOI and cap rates generally between 6.5% and 10%, with higher cap rates for weaker credit, shorter duration, renewal risk and current yield conditions. Pipeline opportunities are probability weighted at about 70% for high-visibility sites within pre-leasing windows, about 30% for opportunities two to three years away, and 10-20% for longer-dated, less-certain projects.

Analysis framework

UBS starts with AI compute demand, power requirements, hyperscale spending and grid constraints. It then assesses each operator’s contracted leases, tenant quality, delivery progress, power pipeline and financing exposure. Valuation combines 2028E EV/EBITDA multiples with an SOTP that separately values stabilized contracted cash NOI and probability-weighted development pipelines. UBS also uses satellite imagery to assess construction progress at selected sites.

Methodology notes

• Industry AnalysisSupply-demand framework

  AI compute demand versus the availability of powered data-center capacity.

  UBS uses accelerating AI infrastructure demand, rising power density and constrained grid interconnection capacity to explain why time to power is the key industry bottleneck.

• Valuation methodsEV/EBITDA valuation

  Forward EV/EBITDA comparison against established data-center REITs.

  UBS applies 2028E EBITDA multiples of 15-22x across the covered stocks, adjusting for execution risk, tenant quality, pipeline optionality and lease maturity.

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation of contracted lease assets and development pipelines.

  Contracted assets are valued using stabilized cash NOI and cap rates, while pipelines receive probability-weighted values based on timing, site readiness and leasing visibility.

• OtherSpread analysis

  Project-financing spreads and tenant credit quality as inputs to valuation and risk.

  UBS uses financing yields and credit support to frame cost of capital, cap rates and relative risk across lease portfolios.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Hut 8 (HUT)

  UBS views HUT as its strongest overall combination of high-quality contracted leases, pipeline optionality and downside protections.

  Strengths

  949MW of signed IT leases; high-credit tenant support; estimated ~1.8GW IT development pipeline; strong lease economics and River Bend construction progress.

  Weaknesses

  No AI/HPC capacity delivered at the report date; Texas energization visibility remains relevant.

  Comparison

  UBS favors HUT for pipeline conversion and lease protection versus peers.

  Risks

  Delivery execution, construction cost overruns, Beacon Point energization and residual bitcoin exposure.

• TeraWulf (WULF)

  UBS sees a differentiated brownfield strategy and near-term pipeline conversion potential.

  Strengths

  840MW of signed IT leases; Hawesville validates brownfield conversion; Muskie could add about 50% to contracted capacity.

  Weaknesses

  Early-stage delivery ramp and exposure to NIMBY risk near population centers.

  Comparison

  UBS views WULF as attractive on contracted asset support and brownfield replicability.

  Risks

  Delivery delays, labor availability, financing needs and tenant/end-demand concentration.

• Applied Digital (APLD)

  UBS highlights leasing momentum, tenant-quality improvement and REIT-conversion optionality.

  Strengths

  1.41GW of signed IT leases; 175MW delivered; diversified tenants including CoreWeave, Oracle and a high-investment-grade hyperscaler; ~2.7GW IT pipeline.

  Weaknesses

  Still early in development and exposed to cost-of-capital uncertainty through the Macquarie funding structure.

  Comparison

  UBS identifies APLD as having the strongest leasing volume among the covered group and better geographic diversification, including no Texas exposure.

  Risks

  Neocloud concentration, financing, execution on newer projects and GPU-cloud separation uncertainty.

• Cipher Digital (CIFR)

  UBS sees the widest upside skew from near-term leasing and improved Texas clarity.

  Strengths

  493MW of signed IT leases; Amazon leases improve tenant quality and risk protection; ~340MW IT near-term lease candidates could lift contracted capacity by ~70%.

  Weaknesses

  About 95% of the portfolio is in Texas and much of the pipeline is subject to ERCOT review.

  Comparison

  UBS sees CIFR as offering the greatest upside potential but also among the highest power-timing risks.

  Risks

  ERCOT allocation and energization, Texas regulatory developments, aggressive delivery schedules and financing.

• Core Scientific (CORZ)

  UBS views CORZ as the most advanced AI/HPC transition among the five, with material AMD expansion optionality.

  Strengths

  ~1.1GW signed IT leases; ~440MW delivered; AMD expansion options of ~1.9GW; geographically diversified footprint.

  Weaknesses

  Current tenant concentration to CoreWeave and uncertainty over AMD financing and ultimate offtakers.

  Comparison

  CORZ trades at 12x 2028E EBITDA, below peers at 13-18x, according to UBS.

  Risks

  AMD delivery execution, tenant concentration, Texas regulatory exposure and project-financing clarity.

Key data

• Incremental global AI data-center capacity required~260GW from 2028-30UBS estimate, about 3x the ~90GW estimated for 2025-27.
• Covered companies' leased capacity~4.8GWAI/HPC colocation capacity already leased across the five companies.
• Covered companies' pipeline~12GWAdditional development pipeline and expansion opportunities.
• Average EBITDA growth~80% annually from 2027E-29EUBS forecast across the five covered companies.
• Group valuation12-18x 2028E EBITDACompared with established data-center REITs at about 19x.
• Contracted capacity delivered~720MW, or ~15%UBS expects a further 1.6GW delivered by mid-2027.
• Average price-target upside36%Across HUT, WULF, APLD, CORZ and CIFR.

Impact & implications

UBS believes that successful delivery and lease conversion could change the group’s market perception from volatile, development-stage mining-linked operators to owners of long-duration infrastructure cash flows. It argues that this transition could support earnings revisions, lower risk premiums and higher valuation multiples, but only if companies deliver capacity, secure financing and demonstrate reliable power availability.

Risks

• Tenant concentration remains high, with most covered companies having only two or three tenants across signed leases.
• Aggressive construction schedules, labor shortages and equipment constraints can delay ready-for-service milestones, trigger penalties or permit tenants to terminate or step in.
• Construction inflation and cost overruns can reduce development yields and increase equity funding requirements.
• Project financing availability and higher credit spreads can pressure cost of capital and equity valuations.
• Utility, interconnection and energization delays—particularly under ERCOT Batch Zero—can defer revenue and reduce pipeline value.
• Regulatory scrutiny, local opposition, permitting, water-use and emissions concerns may delay projects or pressure economics.
• AI demand could be uneven, while gains in accelerator efficiency, model architecture or inference optimization could reduce long-term power-intensity assumptions.
• Market rental rates and development yields could decline if supply rises faster than demand.
• Residual bitcoin-mining exposure leaves some companies sensitive to bitcoin price, network hashrate and power-cost changes.

What to watch

• On-time delivery and revenue commencement, especially group capacity ramping through mid-2027.
• Final ERCOT Batch Zero designations on December 10, 2026 and preliminary Studied Load allocation results in April 2027.
• New lease announcements at Cipher's Reveille, Ulysses and Odessa sites; TeraWulf's Muskie campus; and Applied Digital's tenant expansion options.
• Hut 8 Beacon Point energization and River Bend delivery progress.
• TeraWulf Lake Mariner ready-for-service milestones and Hawesville development progress.
• Applied Digital’s Oracle expansion option, Delta Forge construction progress, financing and tenant-mix improvement.
• Core Scientific's AMD delivery schedule, financing details and exercise of AMD expansion options.
• Progress toward lower risk premiums, REIT conversion and strategic-value catalysts as lease portfolios stabilize.