SoftBank's $85B France Build Signals Power Shift in Cloud Infrastructure Supply

SoftBank Commits $85 Billion to 5 GW French Data Center Capacity

SoftBank announced plans to invest up to €75 billion ($85 billion) to develop 5 gigawatts of data center capacity in France, starting with three sites in Dunkirk, Bosquel, and Bouchain targeting 3.1 GW by 2031. The scale matters: 5 GW equals entire national data center markets, making SoftBank a strategic supplier of wholesale capacity to cloud providers and large enterprises, not just another colocation player.

For enterprise buyers, this creates a new alternative to hyperscaler-owned regions from AWS, Microsoft Azure, and Google Cloud in Europe. The France-based build-out supports EU data sovereignty requirements, making it easier to keep AI training and inference data local to EU jurisdiction. At wholesale scale, SoftBank can introduce new pricing dynamics versus buying directly from hyperscalers, particularly for high-density AI clusters.

The timeline presents risk. Delivery extends to 2031, meaning buyers must treat this as medium-term capacity planning, not immediate relief. EU energy regulations and AI rules will evolve over that period, requiring long-term contracts to account for geopolitical and regulatory exposure.

Australian Contract Absorbs 40% of National Capacity Before It Exists

CDC Data Centres secured the largest data center contract in Australian history—555 MW with an unnamed US investment customer. The deal equals approximately 40% of Australia's total operating data center capacity in 2025, with operations starting in FY28–FY29.

The competitive implication is stark: large platforms are locking in capacity years in advance, reducing flexibility for smaller enterprise tenants and tightening local supply for high-density racks. If one or a few hyperscalers absorb much of this capacity, enterprises relying on colocation in Australia face scarcity risk and should consider multi-region or hybrid designs now.

The 555 MW contract signals that AI infrastructure has reached national-scale investment levels. Enterprise buyers should expect upward pressure on power, land, and premium rack pricing in the Asia-Pacific region as platforms compete for long-term capacity commitments.

TikTok's $29 Billion Thailand Investment Diversifies Southeast Asia Beyond Singapore

Thailand's Board of Investment approved TikTok's $29 billion investment plans to strengthen the country as a regional hub for data centers, cloud services, and AI infrastructure. This puts Thailand in direct competition with Singapore, Malaysia, Indonesia, and Hong Kong as regional infrastructure hubs.

For enterprise buyers running workloads in Southeast Asia, this creates options beyond Singapore, which has faced land and power constraints. Thailand's emerging role provides alternative failover and disaster recovery locations. The regulatory tradeoff: Thai regulatory environment and BOI incentives look attractive now but require scrutiny on data localization and AI rules that may evolve.

The size of TikTok's investment reinforces that OTT and AI platforms are now among the biggest buyers of data center capacity, competing directly with enterprises and cloud providers for power and land. This changes the leverage dynamic in contract negotiations.

What This Means for Capacity Planning and Pricing Power

TeleGeography tracks an immediate pipeline of approximately 650 data center sites globally—more than double historical volume. That expansion is driven by AI and cloud workloads, but it is concentrating in the hands of a small group of companies. Microsoft, OpenAI, Nvidia, and Oracle are investing heavily in one another to secure AI compute pipelines, creating vertical integration in the supply chain.

Enterprise buyers face three direct implications. First, hyperscalers and platforms are pre-committing capacity at national scale, which reduces availability and pricing power for enterprises buying smaller increments. Second, regional diversification becomes critical as single markets face capacity constraints driven by a few large players. Third, long-term contracts now carry more geopolitical and regulatory risk as data sovereignty rules and AI regulations evolve across the EU, APAC, and other regions through 2031.

The capacity expansion is real, but control is consolidating. Enterprises should model scenarios where preferred regions face scarcity or pricing pressure, and build contingency into multi-year infrastructure roadmaps. The days of assuming elastic cloud capacity at stable prices are ending for high-density AI and compute workloads.