Data Centres • Cloud Infrastructure • Digital Sovereignty
Published by GreyRadius Consulting | 10–16 August 2026
Executive Highlights
United States23 companies commit to Texas data-centre standards covering self-funded power, electricity demand, water and incentives.
MalaysiaSTT GDC secures up to US$1.37B of green financing for a Johor campus planned for up to 166MW of IT load.
IndiaL&T and Together AI announce a 10,000-GPU NVIDIA B300 AI factory; Chennai Phase 1 is designed for 250MW.
United StatesBlack Box wins a US$131M, roughly three-year data-centre infrastructure order from a Tier-1 hyperscaler.
GlobalCoreWeave reaches 1.5GW active power, ~3.7GW contracted power and lifts 2026 capex guidance to US$35–39B.
United StatesAlpha Compute targets an initial 200MW Pennsylvania AI campus with potential expansion to 1GW and onsite gas generation.
North AmericaKeel advances a 2.2GW HPC/data-centre pipeline, including a conditional 96MW capacity transfer in Québec.
Data Centre Policy | Power & Grid | Water & Community
Texas secures data-centre standards commitments from 23 major AI, cloud and infrastructure companies
14 Aug 2026 | United States — Texas | 23 AI, cloud and infrastructure companies
Twenty-three companies committed during 10–14 August to comply with Texas Governor Greg Abbott’s data-centre standards, including OpenAI, Meta, Microsoft, Google, Amazon, Oracle, Equinix, Vantage Data Centers, STACK Infrastructure, Anthropic, Compass Datacenters and Hut 8. Participating developments are expected to fund their own electrical infrastructure; disclose projected electricity demand, onsite-generation or grid-relief plans; identify water sources and projected consumption; explain water-reuse and cooling approaches; and disclose taxpayer incentives. The framework also brings PUCT/ERCOT scrutiny to ownership, electricity, water and community impacts. The policy itself predates this reporting period; the new strategic development is the wave of corporate commitments.
Texas is shifting the data-centre approval conversation from how much capacity will be built to who carries the infrastructure burden. Watch whether self-funded electrical infrastructure, water disclosure and onsite-generation commitments become practical requirements for securing large-scale developments. If replicated elsewhere, hyperscalers may increasingly need to underwrite energy and water infrastructure alongside the data centre itself.
GreyRadius Insight
The competitive advantage of a data-centre developer is expanding beyond access to land and grid connections. Developers able to arrive with credible power, water and community-impact solutions could move through constrained markets faster than projects dependent on public infrastructure expansion. For hyperscalers, infrastructure strategy increasingly begins outside the data-centre fence—with generation, transmission, water and local stakeholder economics becoming part of site selection.
Data Centre Financing | Hyperscale Capacity | Sustainability
STT GDC secures up to US$1.37B green financing for 166MW Johor data-centre campus
13 Aug 2026 | Malaysia — Johor | ST Telemedia Global Data Centres
ST Telemedia Global Data Centres secured up to US$1.37 billion (MYR5.6 billion) of green financing for its flagship STT Johor campus at Nusa Cemerlang Industrial Park. The package includes a Green Loan Facility supporting Phase 1 and involves UOB Malaysia, OCBC Malaysia, Standard Chartered Malaysia and CIMB. The campus is planned for up to 166MW of IT load and is intended for cloud, AI, HPC and other high-density workloads. STT GDC also reported that renewables represented 83.2% of group electricity consumption in 2025 and that carbon intensity had fallen 70.5% from its 2021 baseline.
Johor's next test is no longer attracting announcements—it is converting its rapidly growing development pipeline into energized, occupied capacity. Track power availability, renewable-energy sourcing, construction timelines and hyperscaler/AI tenant commitments as the 166MW campus develops. These indicators will show whether Johor can sustain hyperscale growth without infrastructure constraints eroding its cost advantage.
GreyRadius Insight
The US$1.37B financing demonstrates that capital is beginning to follow executable capacity rather than established data-centre geography. Johor is benefiting from a combination of available land, regional connectivity and proximity to Singapore demand. The strategic implication is important: Southeast Asia's infrastructure hierarchy could increasingly be determined by which markets can deliver large blocks of financeable power fastest—not simply which markets historically hosted the most cloud capacity.
AI Infrastructure | Sovereign AI | GPU Compute
L&T and Together AI announce 10,000-GPU NVIDIA B300 AI factory in Chennai
13 Aug 2026 | India — Chennai | Larsen & Toubro / Together AI
Larsen & Toubro, through Vyoma.AI and LTN Compute, announced an NVIDIA B300 AI factory for Together AI at its Chennai data-centre campus. The installation is designed around 10,000 NVIDIA B300 GPUs in an integrated cluster for inference, fine-tuning and large-scale model training. L&T describes Chennai as a gigawatt-scale AI infrastructure site; Phase 1 is designed for 250MW with 150MVA of power-infrastructure readiness. Reuters reported the contract value at up to ₹150 billion (approximately US$1.57 billion). L&T is positioning the facility within an Indian sovereign-AI stack spanning hyperscale colocation, GPU-as-a-Service, managed AI platforms and sovereign cloud.
The 10,000-GPU headline is significant, but the more important milestones are energized MW, installed GPUs and contracted utilization. Track conversion of the 250MW Phase 1 design and 150MVA power readiness into live AI capacity, alongside enterprise and sovereign workload commitments. These will determine whether Chennai develops into a genuine AI-compute cluster rather than simply another large infrastructure announcement.
GreyRadius Insight
India's AI-infrastructure market is moving beyond conventional hyperscale colocation toward an integrated compute + data centre + sovereign cloud + managed AI model. That potentially changes where value is captured: infrastructure providers capable of combining power, GPU capacity and AI platforms can move higher up the stack than traditional colocation operators. For enterprises, the emerging question becomes less ‘Where can we host AI?’ and more ‘Which domestic infrastructure platform can provide scalable compute with predictable economics, sovereignty and operational control?’
Hyperscale Delivery | Digital Infrastructure Services | Data Centres
Black Box wins US$131M U.S. Tier-1 hyperscaler data-centre infrastructure order
13 Aug 2026 | United States | Black Box / Unnamed Tier-1 hyperscaler
India-headquartered Black Box secured a US$131 million digital-infrastructure order from an unnamed global Tier-1 hyperscaler for a major U.S. data-centre project. The engagement is expected to run for approximately three years and covers digital-infrastructure services at a key U.S. location. Black Box indicated that successful delivery could create scope for additional expansion with the customer. The hyperscaler was not disclosed and no verified MW capacity, campus size or rack count was provided, so those metrics are not inferred here.
Watch the downstream beneficiaries of the hyperscale investment cycle. As data-centre capex expands, engineering, network integration, electrical systems and deployment services can become significant bottlenecks. Black Box's three-year engagement provides another signal that spending is spreading beyond developers and equipment vendors into the infrastructure-services layer.
GreyRadius Insight
The AI infrastructure opportunity is broader than ownership of data centres or GPUs. Every additional hyperscale campus creates demand across design, integration, networking, commissioning and lifecycle operations. This creates a second-order investment thesis: as hyperscalers race to deploy capacity, execution capability itself becomes scarce infrastructure. Providers able to deliver consistently across multiple campuses may therefore capture recurring value from the capex cycle without carrying the same asset intensity as data-centre owners.
AI Cloud | AI Infrastructure | Capital Expenditure
CoreWeave reaches 1.5GW active power and raises 2026 capex plan to US$35–39B
11 Aug 2026 | United States / Global | CoreWeave
CoreWeave reported that its purpose-built AI infrastructure footprint expanded by nearly 500MW, bringing active power to 1.5GW, while contracted power reached approximately 3.7GW. Revenue backlog stood at approximately US$104 billion at the end of Q2, excluding more than US$25 billion of additional customer commitments added in early Q3. The company also highlighted a US$3.1 billion HPC-backed term loan and a US$1 billion strategic investment from Jane Street. Reuters reported that CoreWeave raised expected 2026 capital expenditure to US$35–39 billion from US$31–35 billion after quarterly capex reached approximately US$9.4 billion.
CoreWeave's most important gap is now between 1.5GW active and approximately 3.7GW contracted power. Track how quickly that 2.2GW difference becomes operational capacity and whether its US$104B backlog converts into utilization and cash generation fast enough to support a US$35–39B annual capex programme. The critical metric is increasingly capex-to-revenue conversion, not contracted MW alone.
GreyRadius Insight
AI neoclouds are beginning to operate at infrastructure scale once associated primarily with hyperscalers. But rapid expansion creates a different strategic challenge: demand, financing, power and GPU deployment must arrive in sequence. A large backlog can justify aggressive capacity investment, but delays in any one layer can materially change returns. For investors and infrastructure partners, backlog quality, customer concentration and time-to-energization should therefore be evaluated alongside revenue growth.
AI Data Centres | Onsite Power | Infrastructure Development
Alpha Compute signs US$55M deal for 200MW Pennsylvania AI campus with potential expansion to 1GW
11 Aug 2026 | United States — Pennsylvania | Alpha Compute
Alpha Compute signed a binding term sheet to acquire Pennsylvania land and natural-gas rights for US$55 million to develop its first U.S. data-centre campus. The initial development is planned at 200MW, supported by onsite natural-gas generation, with potential expansion to as much as 1GW of power. The proposed transaction covers approximately 1,800 mineral acres of undeveloped Marcellus shale gas rights. Alpha Compute said operations could begin in Q3 2027, while CEO Brittany Kaiser estimated subsequent campus and electrical development at approximately US$500 million over several years. The transaction remains conditional on financing, permitting, due diligence, local approvals and definitive agreements.
The key development is not the potential 1GW campus—it is Alpha Compute's attempt to combine land, natural-gas rights and onsite generation into one infrastructure strategy. Watch permitting, financing and the first 200MW deployment closely. Successful execution would strengthen the case for vertically integrated energy-and-compute campuses in markets where utility interconnection timelines cannot keep pace with AI demand.
GreyRadius Insight
AI infrastructure is starting to reshape site-selection economics. Historically, developers selected locations and then secured electricity. Increasingly, projects may begin with energy resources first and compute second. Acquiring fuel, generation capability and land together can reduce exposure to grid queues and give developers greater control over time-to-power. The implication for future AI campuses is significant: energy strategy may become the primary determinant of geography.
HPC Infrastructure | Site Conversion | Power & Grid
Keel advances 2.2GW North American HPC/data-centre development pipeline
10 Aug 2026 | North America | Keel Infrastructure
Keel Infrastructure reported progress across a North American HPC/data-centre development portfolio totaling approximately 2.2GW of power capacity. Priority sites include Panther Creek, Sharon and Moses Lake, with zoning and land-development progress at the first two and initial Vertiv modules delivered at Moses Lake. Keel also disclosed an agreement with Hydro-Sherbrooke for the conditional transfer and operation of 96MW of existing capacity for its Sherbrooke, Québec project. As of 7 August, Keel reported approximately US$819 million of liquidity—around US$698 million in cash plus US$121 million of unencumbered Bitcoin—following a US$458 million convertible-note raise during Q2. The company has decommissioned its U.S. Bitcoin-mining operations in preparation for HPC construction.
Keel's transition from Bitcoin mining to HPC should be measured by how much of its 2.2GW development pipeline converts into contracted AI workloads. The 96MW Sherbrooke capacity and progress at Panther Creek, Sharon and Moses Lake provide useful milestones, but anchor tenants and construction commitments will determine whether former mining infrastructure can compete effectively for hyperscale AI demand.
GreyRadius Insight
The AI infrastructure boom is creating a new value proposition for former crypto assets. Sites with secured power, substations, land and existing electrical infrastructure can potentially bypass years of greenfield development. But power access alone does not create an AI data centre—network connectivity, cooling architecture, redundancy and customer-grade operating standards remain critical. The winners will be operators that can convert legacy power assets into institutional-quality compute infrastructure faster and cheaper than greenfield alternatives.
Market Data & Intelligence
Metric
Latest Value
Reporting Period
Region
Executive Implication
Texas standards commitments
23 companies
10–14 Aug 2026
United States — Texas
Shifts more grid, water and community infrastructure responsibility onto large data-centre developers.
STT Johor financing / capacity
Up to US$1.37B / 166MW IT load
13 Aug 2026
Malaysia
Shows large-scale green financing supporting Johor’s hyperscale and AI build-out.
200MW initial / up to 1GW / US$55M land & gas rights
11 Aug 2026
United States
Illustrates integrated land, fuel and onsite-generation strategies for faster time-to-power.
Keel HPC pipeline
~2.2GW / 96MW conditional Québec capacity
10 Aug 2026
North America
Highlights conversion of power-rich crypto infrastructure into AI/HPC capacity.
Turn Infrastructure Strategy Into Execution
GreyRadius helps leadership teams translate cloud, data centre and AI infrastructure shifts into market priorities, investment choices and execution-ready growth plans.
