2025 has been nothing short of astonishing for hyperscale data center investment. In just a few months, tens of billions of dollars in new capital commitments have been announced, reshaping the global digital infrastructure map. From multi-gigawatt AI campuses in Texas, Wyoming, and Arizona, to sovereign AI hubs in Europe and Asia, the momentum signals that the data center has become the defining industrial buildout of the decade.
The sheer size of these projects is unprecedented. Several single-campus sites are now measured in gigawatts of IT load, backed by multi-billion-dollar CAPEX plans that rival or exceed the annual GDP of smaller nations. Driving this surge is insatiable demand for AI compute, powered by GPUs and accelerators that require new levels of density, cooling, and interconnect capacity. The following table tracks 20 of the most significant hyperscale builds and expansions announced or advanced in Q2 2025.
| Company / Project | Location / Region | Scale / CAPEX | Status / Milestone | Target Timeline |
|---|---|---|---|---|
| Stargate — Abilene Phase Two | Abilene, Texas, USA | 1.2 GW campus | Phase two underway (6 new buildings) | H1 2025–mid-2026 (phased) |
| Stargate — Norway “AI Gigafactory” | Kvandal/Narvik, Norway | 230 MW initial; 100,000 GPUs by 2026 | Announced; site plan disclosed | 2026 (first phase) |
| Vermaland — 3-GW Arizona Campus | Phoenix–Tucson corridor, Arizona, USA | Up to 3 GW; 3,300 acres (1,336 ha) | Plan announced; QOZ incentives | TBD (multi-year) |
| Crusoe + Tallgrass — Wyoming AI Campus | Southeast Wyoming, USA | 1.8 GW initial; scalable to 10 GW | Announced; energy partnership secured | TBD (phased) |
| CoreWeave — Lancaster County | Pennsylvania, USA | $6B; 100 MW initial → 300 MW | Announced with co-developers | TBD (construction start not disclosed) |
| Vantage — NV1 Campus | Storey County (TRIC), Nevada, USA | $3B; 224 MW across 4 buildings | Announced; first two fully leased | Q2 2026 (first building) |
| Aligned — Conesville Campus | Coshocton County, Ohio, USA | Multi-billion; 197-acre (79.7 ha) site | New campus announced | Mid-2026 (first DC) |
| Aligned — PHX-13 (100-acre / 40.5 ha) | Glendale, Arizona, USA | Campus build; APS 230 kV line | Broke ground | 2025–2026 (phased) |
| Joule — Utah AI Campus | Millard County, Utah, USA | 4 GW on-site power; 4,000 acres (1,619 ha) | Partners named; power program set | 2026 (initial launch) |
| Google — Oklahoma AI Build | Stillwater & Pryor, Oklahoma, USA | $9B for new campus + expansion | Announced | 2025–2027 (phased) |
| DayOne — Lahti Campus | Lahti, Finland | €1.2B; up to 128 MW (50 MW first) | Committed; demolition Q3’25 | 2027 (operations) |
| Novva — Tahoe Reno | Reno, Nevada, USA | 60 MW facility | Launched | 2025 |
| CyrusOne — LON6 UK Campus | London Area (Iver Heath), UK | 90 MW campus | New build announced | TBD |
| CyrusOne — DFW7 Fort Worth | Fort Worth, Texas, USA | 70 MW building | Broke ground | 2026 (target) |
| Colt DCS — Paris 2 | Villebon-sur-Yvette, France | Part of €2.3B France plan; 40 MW site | Broke ground | TBD (under construction) |
| AWS — New Taiwan (ap-east-2) Region | Taiwan | $5B multiyear program | Region launched | June 2025 |
| Stargate — UAE 1-GW Cluster | Abu Dhabi, UAE | ~1 GW AI cluster | Announced | 2026 (planned) |
| TECfusions — Chile AI Campus | Chile (Greater Santiago) | AI-ready campus (scale TBD) | Plan announced | TBD |
| HPE + KDDI — Osaka AI DC | Sakai/Osaka, Japan | New AI data center | Announced | Early 2026 |
| IIJ — Shiroi DC Expansion | Chiba Prefecture, Japan | AI liquid-cooled expansion | Expansion announced | TBD |
🌐 Analysis
The redirection of global CAPEX toward hyperscale AI campuses is reverberating far beyond the data center sector. Network equipment vendors are already experiencing demand shocks as operators design for terabit-scale east-west traffic and ultra-low latency interconnects between GPU clusters. Traditional spine-leaf architectures are being pushed to their limits, creating openings for new switching silicon, CPO/LPO optics, and 1.6T pluggables.
Vendors such as Cisco, Arista, Juniper, Nokia, and Broadcom are competing aggressively to supply the fabrics that tie together thousands of GPU nodes per hall. Optical component suppliers like Coherent, Lumentum, and Infinera are racing to ramp 1.6T capacity and 6-inch InP wafers to keep pace. At the same time, operators are demanding radical efficiency in power, cooling, and interconnect density—forcing vendors to re-engineer products for liquid cooling compatibility and higher rack power envelopes.
In effect, the CAPEX wave that is building hyperscale AI campuses is also rewriting the roadmap for network and optical equipment. Suppliers that can align with the scale, thermal profile, and bandwidth intensity of these builds stand to capture the next decade of growth.
🌐 We’re tracking the latest developments in AI infrastructure and hyperscale data center CAPEX. Follow our ongoing coverage at: https://convergedigest.com/category/data-centers
