Digital Core REIT to sell three North American data centres. The company will redeploy funds towards its Singapore and Japan assets. Singapore-listed Digital Core REIT has announced an agreement to sell its interests in three North American data centres to its sponsor, Digital Realty, and reallocate the funds towards assets in Singapore and Japan, debt repayment and unit buybacks. The restructuring is expected to increase Digital Core REIT's distribution per unit by 4.1% and lower its debt ratio to 36.3% from 39.2%. Under the agreement, the real estate investment trust will sell its 90% interest in 371 Gough in Toronto for C$180m (US$127m), its 90% interest in 200 N. Nash in Los Angeles for $102.6m (US$79m), and a 39% stake in 8217 Linton Hall in Northern Virginia for $142.9m (US$110m). It will retain a 51% majority stake in the Northern Virginia property. Gross proceeds from the North American sales will amount to about $410.4m (US$316m). Join Singapore Business Review community From these earnings, Digital Core REIT will spend roughly $228.6m (US$176m) to acquire a 2.5% stake in 11 Loyang Close in Singapore for $87m (US$68m) and raise its ownership in Digital Osaka 3 in Japan to 45% for ¥17.6b (US$108m) from 20%. The transactions will yield net proceeds of around $181.8m (US$140m). The trust intends to use roughly $152.0m (US$117m) to pay down debt and allocate up to $26.0m (US$20m) for open-market unit repurchases. The deal is subject to unitholder approval and standard closing conditions, with completion anticipated before the end of the year. ($1.00 = US$0.78)

Digital Core REIT to sell stakes in three US data centers back to Digital Realty, buys stake in two facilities in Asia. REIT sells stakes for $316 million, buys stakes for $176 million August 12, 2026 Digital Core REIT is selling its stake in several US data centers back to Digital Realty, and buying a stake in two more in Asia from the operator. The company, a pure-play data center REIT listed in Singapore and sponsored by Digital Realty, announced today that it has reached an agreement to sell interests in three North American assets back to Digital Realty. The firm will, in turn, use the proceeds to acquire partial interests in a data center in Singapore, up its stake in a data center in Osaka, pay down debt, and repurchase units on the open market. The deal would see Digital Core REIT enter the Singaporean market for the first time. The transaction would see Digital Core REIT sell its 90 percent interest in 371 Gough in Toronto for CA$180 million (US$127 million), sell its 90 percent interest in 200 N. Nash in Los Angeles for US$79 million; and the sale of a 39 percent interest in 8217 Linton Hall in Northern Virginia for US$110m. In return, Digital Core REIT has agreed to acquire a 2.5 percent interest in 11 Loyang Close in Singapore from Digital Realty for S$87m (US$68m) and acquire an additional 25 percent interest in Digital Osaka 3 from Digital Realty for ¥17.6 billion ($108m), increasing its ownership stake to 45 percent. The sales are expected to generate gross proceeds of approximately $316m. The purchases are expected to represent a total investment of approximately $176m. Around $117m of the $140m in net proceeds will be used to pay down debt, and up to $20m to repurchase units on the open market. "With this transaction, we expect to tactically enhance Digital Core REIT's portfolio mix, leverage and distribution per unit, in an effort to better position the REIT for the unprecedented opportunity we see ahead," said Digital Realty chief investment officer, Gregory S. Wright. "Digital Core REIT was formed to capitalize on the enormous growth potential within the data center sector, which is now materializing in the midst of the sector's ongoing investment cycle. John J. Stewart, CEO of Digital Core REIT Management Pte. Ltd., the manager of Digital Core REIT, added: "This transaction marks our entry into Singapore and strengthens our presence in Japan - a pivotal step in our strategy to expand in the Asia Pacific region. This multi-faceted transaction reflects our Sponsor's firm commitment to Digital Core REIT's near- and long-term success and our own commitment to creating durable value for unitholders." Singapore-listed Digital Core REIT was set up by Digital Realty in 2021 to hold a number of its stabilized data centers. The company was seeded with a portfolio of ten data centers from Digital Realty across the US and Canada in Northern Virginia, Northern California, Los Angeles, and Toronto, totaling more than 1.2 million sq ft (111,484 sqm) and 49MW. Located in Bristow, Virginia, the single-story 8217 Linton Hall Road facility was built in 2001 and totals 9MW across 207,000 sq ft (19,230 sqm). Previously occupied by a "Fortune 50 software company" from 2005 to the end of this year, Digital Core REIT recently secured a 10-year agreement with an unnamed "investment grade global cloud service provider" for the whole facility. The Linton Hall Road facility was acquired by Digital Realty in 2017 when it bought DuPont Fabros Technology. Previous reports suggest the site had been leased to Microsoft. The data center at 371 Gough Road in Toronto, Ontario, totals 120,040 sq ft (11,152 sqm) and 6.75 MW. Set on 7.6-acres, the multi-tenant site was built in 1980 and renovated into a data center in 2015. Digital Realty operates the single-story site as its YYZ10 data center. Meanwhile, 200 North Nash Street in El Segundo, California; 113,606 sq ft (10,554 sqm). Built in 1976 and sitting on 4 acres, Digital operates it as its LAX11 facility. The two-story site was previously set to be leased to Cyxtera until 2033, but the lease ended early after Cyxtera's bankruptcy and sale to Brookfield. First announced in 2019, Digital Realty launched its SIN12 facility at 11 Loyang Close in 2021. The five-story facility totals 50MW across 365,000 sq ft (33,910 sqm). According to Digital Core REIT, the site is 94 percent occupied. Digital Core REIT first took a 20 percent interest in Digital Osaka 3 (aka KIX12) from Mitsubishi Corporation last year. The facility was completed in July 2021. It offers 19.9MW of IT load across 193,535 sq ft and is fully leased. More in investment / M&A / financing.

Liquid cooling captures 34% of AI data centre builds as power density quadruples. Direct-to-chip immersion cooling adoption accelerates amid hyperscaler demands for 500+ kW/cabinet densities. Infrastructure constraints force rapid technology migration. Liquid cooling reaches critical mass. Liquid cooling systems now account for 34% of new AI-focused data centre builds globally, up from just 8% two years ago, according to Datacentres.com infrastructure tracking data. This represents a fundamental shift in how the industry approaches thermal management as power densities in hyperscaler facilities have quadrupled to 500-800 kW per cabinet in 2026, compared to 150-200 kW in 2022. The transition is driven by physics, not preference. Traditional air-cooled systems cannot efficiently dissipate heat generated by the latest NVIDIA H200 and upcoming Blackwell GPU clusters without operating at prohibitive Power Usage Effectiveness (PUE) levels above 1.4. Direct-to-chip immersion cooling systems achieve PUE figures of 1.05-1.15, translating to 25-30% energy cost savings across a facility's operational lifetime. Microsoft's reported $40B Aligned Data Centers consortium arrangement with NVIDIA and xAI includes specifications for immersion cooling across 60% of planned GPU cabinet deployments. Meta Platforms has mandated liquid cooling for all new Llama model training infrastructure. Amazon Web Services has trialled single-phase and two-phase cooling systems in its Northern Virginia and Oregon facilities, with rollout decisions expected by Q4 2026. Infrastructure bottlenecks drive adoption. Liquid cooling adoption paradoxically accelerates despite supply constraints. Immersion cooling tanks occupy 35% less floor space than traditional air-cooled cabinet configurations, allowing operators to achieve 8-10 MW per 100,000 square foot facility versus 5-6 MW previously. This space efficiency proves critical as construction costs hover at $11.3M per MW and land availability remains constrained in primary markets. Supply chains remain stressed. Lead times for specialised cooling pumps and dielectric fluids extend to 18-24 months. Takraf, Aspen Systems, and ExaScaling have expanded manufacturing capacity by 40% through 2026, yet orders exceed availability. Secondary-market cooling system costs have increased 12-15% year-on-year, creating margin pressure for smaller operators unable to negotiate volume discounts. Electrical infrastructure constraints reinforce the case for denser, liquid-cooled systems. Transformer procurement timelines remain at 36 months, the critical bottleneck identified by the Equipment Manufacturers Association in March 2026. Every kilowatt-per-square-foot gain through improved cooling efficiency reduces transformer capacity requirements, offsetting construction delays in power supply chains. AI workload requirements reshape architecture. Power density acceleration tracks directly with AI model training requirements. Transformer architecture training for large language models now requires sustained 600+ kW per cabinet for 90-day training cycles. This differs fundamentally from cloud computing workloads, which typically operate at 40-60% utilisation with variable thermal loads. Equinix, Digital Realty, and CoreWeave have all announced immersion cooling installations specific to AI training clusters. CoreWeave's specialised AI infrastructure platform currently operates 12 immersion-cooled facilities globally, with 8 additional projects under construction. The company has achieved average cabinet utilisation exceeding 94%, compared to 68% for traditional air-cooled enterprise data centres. Liquid cooling also enables geographic arbitrage. Facilities in cooler climates - Scandinavia, Finland, Iceland - benefit less from liquid cooling's thermal advantages but gain competitive positioning through renewable energy alignment. However, liquid-cooled systems reduce reliance on ambient conditions, theoretically allowing deployment in warmer regions with comparable operating costs. This geographic flexibility proves strategically valuable as $156B in projects face community opposition in traditional markets. Modular construction accelerates deployment. Liquid cooling integrates naturally with modular, pre-fabricated data centre construction. Standardised immersion cooling pods, manufactured off-site and deployed as complete units, reduce field installation time by 40-50% compared to traditional cabinet-by-cabinet assembly. Dell Technologies and Hewlett Packard Enterprise have jointly launched modular immersion-cooled systems designed for 48-72 hour deployment. These containerised units arrive fully integrated with cooling circuits, electrical distribution, and network infrastructure. Scale Computing and others have ordered 2,400 such modules through 2027, implying ~14-16 GW of modular AI infrastructure capacity in pipeline. Modular construction addresses labour constraints. Skilled data centre technicians remain scarce in primary markets; containerised systems reduce on-site skilled labour requirements by 35-40%. This proves particularly valuable in Nordic expansion markets where construction workforce availability constrains traditional build methodologies. Energy innovation and efficiency gains. Liquid cooling catalyses broader energy innovation. Waste heat recovery systems integrated with immersion-cooled facilities achieve secondary uses: district heating networks in Scandinavia, greenhouse agriculture in controlled environments, or secondary power generation through organic Rankine cycles. Hydrogen cooling research, previously theoretical, has moved to pilot stage. Commonwealth Fusion Systems and several regional utilities are exploring hydrogen-based cooling loops as ultra-efficient thermal transfer media. Pilot programmes operate in Massachusetts and Sweden, with commercial viability assessments due Q1 2027. Energy management software integration improves efficiency further. Real-time thermal monitoring on immersion-cooled systems enables predictive power distribution and workload shifting at sub-second intervals. Google has reported 3-5% additional efficiency gains through software optimisation combined with hardware cooling improvements. Market consolidation implications. Liquid cooling capital requirements favour scale. Facilities deploying immersion cooling require 18-24% higher capex per MW than traditional builds, but achieve 25-30% lower operational costs. This favours large operators - Equinix, Digital Realty, QTS Realty, Iron Mountain - capable of amortising technology investments across multiple geographies. Independent operators face margin compression. Smaller facilities cannot justify immersion cooling capex without hyperscaler anchor tenants providing 80+ MW commitments. This drives consolidation: three independent data centre operators have exited markets entirely in 2026, with portfolios acquired by tier-one operators at 0.7-0.85x replacement cost. Outlook. Liquid cooling transitions from emerging technology to industry standard within 18 months. By 2028, immersion cooling will likely account for 55-65% of new AI infrastructure builds, approaching 40% of total data centre capacity additions. Energy efficiency gains alone justify the technology shift, whilst space optimisation and thermal performance address infrastructure constraints that would otherwise limit AI model training deployment. Operators unable to adopt liquid cooling risk competitive disadvantage in securing hyperscaler contracts worth $2-3B annually per major market. Need bespoke market analysis? Its advisory team delivers in-depth research tailored to your investment and operational requirements.