Data Centres • Cloud Infrastructure • Digital Sovereignty
Published by GreyRadius Consulting | 27 July–2 August 2026
Executive Highlights
Global Amazon lifts 2026 capex to ~US$220B; AWS Q2 revenue reaches US$42.2B, +37% YoY, while capacity remains constrained.
Global Schneider raises 2026 revenue growth outlook to 10–13% and adjusted EBITA growth to 14–19% on data-centre infrastructure demand.
United States Meta spends US$31.08B in Q2 and raises the floor of 2026 capex guidance to US$130B–US$145B.
Israel A 140-day grid-processing freeze follows ~19GW of new data-centre connection requests in two months for projects ≥8MW.
France Orange–Morrison plan a 50:50 sovereign platform: five data centres, four campuses, ~€3B investment and a 400MW target.
United States NOAA selects Google Cloud H4D VMs with 5th-gen AMD EPYC as the primary infrastructure for operational weather HPC.
United States Google files a ~US$1B, 470,000-sq-ft second building at LaGrange; wider Project Pegasus is estimated at ~US$8B.
United States Stillwater proposes up to US$13B and 1.25GW near Atlanta across ~400 acres, as many as 15 buildings and 3.465M sq ft.
Hyperscale Investment | Cloud Infrastructure | AI Infrastructure
Amazon raises 2026 investment plan to US$220B as AWS and AI infrastructure demand accelerates
30 July 2026 | United States / Global | Amazon / AWS
Amazon reported Q2 2026 AWS revenue of US$42.2 billion, up 37% year on year, its fastest AWS growth in 18 quarters and equivalent to a US$169 billion annualised revenue run rate. Amazon said both its AI business and custom-chip business had exceeded US$25 billion annual revenue run rates, while Anthropic and OpenAI have made multi-year, multi-gigawatt commitments to Trainium infrastructure. Trailing-12-month free cash flow moved to a US$7.6 billion outflow, driven primarily by a US$66.1 billion year-on-year increase in property and equipment purchases that Amazon attributed primarily to AI investment. CEO Andy Jassy raised expected 2026 capital expenditure to approximately US$220 billion from an earlier expectation of roughly US$200 billion, while saying AWS continues to face capacity constraints. Amazon also disclosed that its global data centres are more than 7x more water-efficient than the industry average and that it has reached 75% of its target to become water-positive across global data-centre operations by 2030.
Track the conversion of the additional ~US$20B capex uplift into actually energised capacity rather than announced projects. The critical indicators are time-to-power, Trainium deployment, network availability and the share of multi-gigawatt AI commitments that becomes revenue-bearing infrastructure. With trailing-12-month free cash flow already at a US$7.6B outflow, execution quality will increasingly be measured by utilisation and speed-to-revenue, not capex scale alone.
GreyRadius Insight
AWS growing 37% while management still cites capacity constraints changes the strategic question from “is AI demand durable?” to “who can add powered compute fastest?” Amazon’s ~US$220B plan gives suppliers visibility, but it also raises the cost of delays: every bottleneck in power, cooling, chips or construction can defer monetisation of committed demand. For operators and investors, powered capacity and utilisation should now carry more weight than headline investment announcements.
Data Centre Power & Cooling | Infrastructure Supply Chain | Sustainability
Schneider Electric raises 2026 outlook as data-centre demand drives infrastructure growth
30 July 2026 | France / Global | Schneider Electric
Schneider Electric raised its 2026 adjusted EBITA growth outlook to 14–19%, from 10–15%, as demand for data-centre power, cooling, racks and electrical-distribution equipment continued to accelerate. It also lifted expected revenue growth to 10–13%, from 7–10%. First-half adjusted EBITA reached €4.09 billion, ahead of the approximately €3.8 billion market expectation, while Q2 revenue reached €11.46 billion. North America remains the largest data-centre market for Schneider, with the company also reporting strengthening demand in Southeast Asia. Schneider estimated a €400 million–€500 million full-year foreign-exchange headwind, underscoring that underlying operational growth is stronger than reported currency-adjusted figures suggest.
Watch order intake, backlog conversion and lead times across medium-voltage equipment, UPS, switchgear, racks and cooling. Schneider’s raised 10–13% revenue-growth and 14–19% adjusted-EBITA-growth outlook should be tested against manufacturing capacity and regional delivery constraints, especially in North America and fast-growing Southeast Asia. Supplier lead times can become an early-warning indicator for hyperscale commissioning schedules.
GreyRadius Insight
The AI infrastructure cycle is broadening from accelerators into the electrical and thermal systems required to make compute usable. Schneider’s €4.09B first-half adjusted EBITA and upgraded outlook suggest that value is migrating toward the “picks-and-shovels” layer of the data-centre stack. Leaders planning capacity should secure critical power and cooling packages earlier, because equipment availability can now determine time-to-revenue as much as server availability.
AI Infrastructure | Hyperscale Investment | Data Centre Expansion
Meta lifts the floor of 2026 capex guidance after spending US$31.08B in Q2
29 July 2026 | United States / Global | Meta
Meta reported US$31.08 billion of Q2 capital expenditure, including finance-lease principal payments, as it continued to expand AI compute and data-centre infrastructure. The company narrowed full-year 2026 capex guidance to US$130 billion–US$145 billion, raising the lower bound from its previous US$125 billion–US$145 billion range. Q2 free cash flow fell to just US$784 million, versus US$8.55 billion a year earlier, illustrating the cash intensity of the infrastructure build-out. Meta ended June with US$225.7 billion of net property and equipment on its balance sheet, up from US$176.4 billion at the end of 2025. Infrastructure spending across chips, servers, energy and data centres is the central driver of the company’s accelerating investment profile.
Track how the US$130B–US$145B 2026 capex envelope translates into operational compute, power contracts and owned-versus-leased capacity. Q2 capex of US$31.08B alongside only US$784M of free cash flow makes utilisation, deployment timing and AI monetisation increasingly important. Delays in energisation or accelerator deployment would carry a larger cash-flow penalty at this spending level.
GreyRadius Insight
Meta’s rising capex floor shows that the hyperscale demand pool is no longer defined only by public-cloud providers. Its net property and equipment rose to US$225.7B from US$176.4B at end-2025, indicating a structural shift toward infrastructure-heavy economics. For data-centre markets, this widens competition for power, land and supply-chain capacity; for Meta, the strategic test is whether compute productivity rises fast enough to justify the balance-sheet intensity.
Data Centre Energy | Grid Regulation | Infrastructure Policy
Israel pauses new large data-centre grid applications for 140 days after connection requests reach 19GW
28 July 2026 | Israel | Electricity Authority / NOGA
Israel’s Electricity Authority imposed a 140-day freeze on processing new data-centre grid-connection applications while the power system assesses whether requested capacity can be accommodated. The measure applies to proposed facilities of 8MW or more and is expected to remain in force until early December. Israel Independent System Operator NOGA reportedly received approximately 19GW of new data-centre connection requests in only two months; DatacenterDynamics reported roughly 27GW of pending data-centre grid applications in total. The authority characterised the measure as an emergency brake intended to prevent the system from making connection commitments it may later be unable to deliver. The scale of the queue illustrates how AI and cloud infrastructure demand is beginning to exceed near-term power-system planning assumptions.
Watch whether the 140-day pause produces new connection criteria, deposits, milestones or prioritisation rules designed to separate executable projects from speculative queue positions. With ~19GW of requests arriving in two months for projects of 8MW or more, developers should treat grid entitlement as a core development asset. Financing and land acquisition decisions may increasingly need to follow—not precede—credible power allocation.
GreyRadius Insight
Israel illustrates a wider transition from grid connection as an engineering process to grid capacity as a scarce, policy-allocated resource. A 19GW surge can overwhelm planning assumptions long before facilities are built, prompting regulators to ration access. For developers, “power-secured MW” is becoming a more meaningful measure of pipeline quality than announced MW; for governments, queue reform and generation/transmission planning are becoming digital-infrastructure policy.
Digital Sovereignty | Colocation | Data Centre Investment | Strategic Partnership
Orange and Morrison plan €3B sovereign French data-centre platform targeting 400MW
27 July 2026 | France / Europe | Orange / Morrison
Orange and infrastructure investor Morrison entered exclusivity to form a 50:50 jointly controlled French data-centre company positioned explicitly around European digital sovereignty. Orange would seed the venture with five data centres across four campuses at Chevilly-Larue, Aubervilliers, Chartres and Val-de-Reuil. The partners plan to increase capacity to 400MW, nearly 10 times the portfolio’s current scale, supported by an approximately €3 billion investment programme. Funding is expected to combine Orange assets, Morrison equity and debt, while Orange and Orange Business would remain major customers. Completion is targeted for Q1 2027, subject to customary approvals and transaction conditions.
Track Q1 2027 transaction completion, financing structure, power procurement and the phasing required to move five existing facilities across four campuses toward 400MW. The key commercial signal will be how much expansion is anchored by Orange/Orange Business versus sovereign-cloud and external AI customers. Capacity should be evaluated campus by campus rather than assuming the full ~€3B programme converts uniformly into commissioned MW.
GreyRadius Insight
The 50:50 structure offers a practical European sovereignty model: combine trusted domestic telecom assets and customer relationships with specialist infrastructure capital. The targeted 400MW—nearly 10x current scale—shows that sovereignty is moving from policy language into physical capacity investment. For European operators, partnerships like this can accelerate scale while limiting balance-sheet concentration, but sovereign positioning will only create durable advantage if matched by competitive power, connectivity and operating economics.
Government Cloud | HPC | Cloud Infrastructure
Google Cloud selected as primary HPC infrastructure provider for NOAA weather supercomputing
27 July 2026 | United States | Google Cloud / NOAA
NOAA selected Google Cloud as the primary high-performance-computing infrastructure provider for its Weather and Climate Operational Supercomputing System, marking a significant migration from conventional on-premises supercomputing toward cloud-first infrastructure. Google Cloud H4D VMs, powered by fifth-generation AMD EPYC processors, will form the principal compute backbone. The system will support tightly coupled numerical-weather-prediction workloads that historically required dedicated supercomputing infrastructure. Google described NOAA as one of the first operational numerical-weather-prediction centres globally to undertake this type of public-cloud transition. Neither contract value nor total core/GPU count was disclosed, so no capacity estimate has been made.
Watch migration milestones, application performance, resilience architecture and the balance between elastic cloud resources and any retained dedicated infrastructure. Because contract value and compute count were not disclosed, the most important evidence will be operational: forecast throughput, reliability, scaling behaviour and cost efficiency on H4D infrastructure. Other public-sector HPC buyers will closely scrutinise those outcomes.
GreyRadius Insight
NOAA’s move is strategically more important than its undisclosed contract size because it tests public cloud against a tightly coupled, mission-critical scientific workload. If Google Cloud can meet operational weather-forecasting requirements, the addressable cloud market expands beyond enterprise IT into workloads historically reserved for purpose-built supercomputers. The implication for government CIOs is a new sourcing choice: procure HPC as elastic infrastructure rather than refresh dedicated systems on long hardware cycles.
Hyperscale Data Centres | Capacity Expansion | Investment
Google files US$1B second-building expansion for LaGrange, Georgia data-centre campus
27 July 2026 | United States | Google
A development filing submitted on behalf of the property owner details a second data-centre building at Google’s developing LaGrange campus southwest of Atlanta. The planned building would total approximately 470,000 sq ft / 43,665 sq m and is valued at roughly US$1 billion. Construction could continue through 2029, while the filing leaves open the possibility of additional phases but describes those as speculative. Local reporting places the wider Project Pegasus development at an estimated US$8 billion when fully built. No verified MW figure for this specific expansion was disclosed in the filing material surfaced during this run.
Track development approvals, utility interconnection, construction milestones through 2029 and any later disclosure of MW capacity. The ~US$1B, 470,000-sq-ft second building is tangible, while additional phases remain speculative; separate committed construction from long-range campus potential. Power delivery will be the key variable determining whether the wider ~US$8B Project Pegasus estimate becomes executable capacity.
GreyRadius Insight
The LaGrange filing shows why headline campus values should be decomposed into phase-level commitments. A US$1B second building is a meaningful near-term signal, but without verified MW the better indicators are construction status, interconnection progress and energisation dates. For market sizing, committed and power-backed phases should be weighted more heavily than total master-plan value.
Hyperscale Infrastructure | Data Centre Investment | Energy & Power
Stillwater files plans for up to US$13B, 1.25GW data-centre campus outside Atlanta
27 July 2026 | United States | Stillwater Development
Stillwater Development filed a development-of-regional-impact application for a large data-centre campus in Lovejoy, Georgia. The proposal covers roughly 400 acres and could ultimately contain as many as 15 buildings, with reported total floor area of approximately 3.465 million sq ft. The planned grid connection could reach 1.25GW, supplied by Central Georgia EMC, while two on-site substations are contemplated. Full project value could reach approximately US$13 billion, with the first phase potentially completing around 2031. The site sits within the wider Atlanta data-centre market, although Lovejoy itself is outside Clayton County’s moratorium applying to unincorporated county land.
Track regional review, customer commitments, financing and—above all—the 1.25GW interconnection pathway with Central Georgia EMC and delivery of two proposed substations. The ~400-acre, up-to-15-building master plan and potential 2031 first phase create substantial schedule exposure to transmission, permitting and long-lead electrical equipment. Treat the US$13B value as development potential until power and phased construction commitments become firmer.
GreyRadius Insight
At 1.25GW, the project’s strategic significance is less the number of buildings than the amount of power it seeks to aggregate at one location. Gigawatt-scale campuses are turning utility capacity, substations and transmission schedules into the gating assets of hyperscale development. For investors and economic-development agencies, the key distinction is increasingly between “planned GW” and “deliverable GW”—capacity backed by interconnection, infrastructure, financing and an executable timeline.
Introduces a gigawatt-scale, 15-building development pipeline outside Atlanta.
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.
