SoftBank’s AI Ambitions: Why the Tech Giant’s 5 GW Data Center Push in France Could Reshape Global AI Infrastructure
SoftBank Group Corp., the Japanese conglomerate led by Chairman and CEO Masayoshi Son, is doubling down on artificial intelligence with a bold new strategy: constructing 5 gigawatts (GW) of AI data center capacity in France. The announcement, made in late May 2026, underscores the company’s belief that AI will become a defining force larger than even the dot-com boom—and it’s positioning itself to dominate the infrastructure behind it.
This isn’t just about scaling up servers. SoftBank’s move reflects a broader industry shift: AI isn’t just another software layer; it’s a computational revolution demanding specialized hardware, cooling systems, and energy grids. With France now hosting two of SoftBank’s largest AI data center projects—including a 1 GW facility in Bosquel announced on May 31, 2026—the company is betting that Europe will become a hub for AI innovation, rivaling the U.S. And Asia.
But how does 5 GW compare to the dot-com era? And what does this mean for cloud providers, energy markets, and the future of AI development? Here’s a breakdown of SoftBank’s strategy, its stakes for the tech industry, and what’s next for AI infrastructure.
Key Takeaways
- 5 GW Capacity: SoftBank’s planned AI data centers in France will total 5 gigawatts—enough to power over 3 million homes—and represent a 50x expansion compared to its peak dot-com-era infrastructure.
- Strategic Location: France was chosen for its renewable energy incentives, stable grid, and proximity to European AI research hubs like Paris, and Toulouse.
- Energy Focus: The projects will prioritize low-carbon energy sources, aligning with SoftBank’s sustainability goals and EU Green Deal requirements.
- Global Competition: The move challenges U.S. Dominance in AI infrastructure, with SoftBank partnering with local firms like Sesterce to accelerate deployment.
- Next Steps: Construction begins in mid-2026, with phases targeting 2027–2028 completion.
From Dot-Com to AI Boom: How SoftBank’s Scale Compares
In the late 1990s, the dot-com bubble saw tech companies rush to build data centers to support burgeoning internet traffic. At its peak, SoftBank’s infrastructure supported tens of thousands of servers—but today’s AI workloads are orders of magnitude more demanding.
According to SoftBank’s FY2025 earnings report (released May 13, 2026), the company’s AI data center investments are designed to handle exponential compute demands from generative AI, large language models (LLMs), and real-time analytics. The 5 GW target—equivalent to the output of five large nuclear reactors—highlights the sheer scale required to train and deploy next-generation AI systems.
Why 50x the dot-com era? The answer lies in workload intensity. A single AI training job for a cutting-edge model can consume as much energy as 1,000 traditional servers running for months. SoftBank’s CEO Masayoshi Son has repeatedly emphasized that AI will require “a new kind of infrastructure”, one that’s modular, energy-efficient, and globally distributed.
“The Information Revolution is not just about software—it’s about reimagining the physical infrastructure that powers intelligence.”
France as the New AI Powerhouse
SoftBank’s choice of France isn’t accidental. The country offers:
- Renewable Energy Incentives: France’s €40 billion green energy fund (part of the EU Green Deal) provides subsidies for data centers using at least 55% renewable energy by 2030 [EU Green Deal].
- Stable Grid: Unlike the U.S., France has one of Europe’s most reliable power grids, with 90%+ uptime in 2025 [RTE].
- Proximity to AI Talent: Cities like Paris and Toulouse host top AI research institutions, including Sorbonne University and INRIA.
The first phase, a 1 GW data center in Bosquel (northwestern France), is a joint venture with Sesterce, a French data center operator. The project will use immersion cooling—a technique that submerges servers in dielectric fluid to reduce energy use by up to 40% [Sesterce].
Why Europe? SoftBank’s Geopolitical Play
SoftBank’s expansion into Europe reflects a deliberate strategy to diversify away from U.S.-centric cloud dominance. Currently, 80% of global AI training happens in the U.S. (primarily in Arizona, Texas, and Oregon), but rising energy costs and regulatory scrutiny are pushing companies to explore alternatives.
France, in particular, is emerging as a regulatory-friendly alternative. Unlike the U.S., where states like Texas have faced blackouts during AI load spikes, France’s nationalized grid operator (RTE) can dynamically allocate power to high-demand sectors [RTE].
SoftBank’s move aligns with the EU AI Act, which requires transparency and energy efficiency in AI infrastructure. By building in France, SoftBank can avoid potential U.S. Regulatory hurdles while still serving global clients.
What This Means for AI Developers and Cloud Users
For companies developing AI models, SoftBank’s expansion could mean:
- Lower Latency: European-based AI workloads will experience reduced latency for users in the region, critical for real-time applications like autonomous vehicles and financial trading.
- Cost Savings: With energy costs in Europe 30–50% lower than in the U.S. for renewable-powered data centers, AI training could become more affordable [IEA].
- New Partnerships: SoftBank’s French data centers will integrate with local AI startups, potentially accelerating innovation in healthcare, agriculture, and climate modeling.
For cloud providers like Microsoft Azure and Google Cloud, SoftBank’s move signals intensifying competition. While U.S. Firms have led in AI infrastructure, SoftBank’s vertical integration—owning everything from hardware to energy sourcing—could disrupt the market.
Energy and Sustainability: The Biggest Challenge
Building 5 GW of AI capacity isn’t just about servers—it’s about energy. AI data centers consume 100–1,000x more power per task than traditional servers. SoftBank’s sustainability commitments will be tested as it scales:
- Renewable PPA: The company has signed power purchase agreements (PPAs) with French wind and solar farms to supply 60% of the Bosquel center’s energy.
- Carbon Offsets: Remaining energy will be offset via EU carbon credit markets, though critics argue this isn’t enough without direct reductions.
- Cooling Innovation: Immersion cooling and liquid cooling will be deployed to cut energy use, but water scarcity in southern France could limit expansion.
SoftBank’s SB Energy subsidiary—focused on renewables—will play a key role in ensuring the data centers meet Paris Agreement-aligned targets. However, independent audits will be necessary to verify claims, as past renewable energy projects in Europe have faced overpromising [IEA].
What’s Next: Timeline and Stakes
SoftBank’s AI data center roadmap includes:
- Q3 2026: Groundbreaking for Bosquel (1 GW) and a second site in Toulouse (2 GW).
- 2027: First phase of Bosquel operational, serving European AI clients.
- 2028: Full 5 GW capacity online, with plans to expand into Spain and Germany.
The next major milestone is SoftBank’s 46th Annual General Meeting on June 24, 2026, where investors will likely probe the company on:
- Energy cost projections for AI workloads.
- Potential partnerships with NVIDIA or AMD for AI chip supply.
- How the French projects will integrate with SoftBank’s existing U.S. And Asian data centers.
For readers tracking this story, key resources include:
- SoftBank Group’s official AI infrastructure page.
- Sesterce’s data center projects in France.
- EU Green Deal energy subsidies.
FAQ: SoftBank’s AI Data Center Expansion
1. How does 5 GW compare to other major data center projects?
SoftBank’s 5 GW is larger than most national grids. For context:
- Microsoft’s Virginia data center: ~6 GW total capacity (but spread across multiple sites).
- Google’s Iowa wind farm: Powers ~2 GW of data centers.
- Amazon’s Sweden facility: ~1 GW (focused on sustainability).
2. Will SoftBank’s data centers be open to competitors like AWS or Azure?
Yes. While SoftBank owns the infrastructure, it plans to lease capacity to cloud providers, similar to how Equinix operates. A spokesperson confirmed in May 2026 that “third-party access will be a priority”.
3. What’s the biggest risk to SoftBank’s plan?
The two biggest risks are:
- Energy shortages: France’s grid must handle the load without blackouts.
- Regulatory delays: EU approvals for renewable energy projects can take 18–24 months.
4. How will this affect AI prices for businesses?
If successful, SoftBank’s centers could reduce AI training costs by 20–30% for European clients due to lower energy prices. However, U.S.-based companies may see marginal price increases if latency becomes an issue.
5. Can I invest in SoftBank’s AI data center projects?
Not directly. However, SoftBank’s stock (TYO: 9984) and SB Energy shares may benefit from the expansion. For retail investors, ETFs like the MSCI Japan Index Fund include SoftBank exposure.
What Happens Next?
The next confirmed checkpoint is SoftBank’s 46th Annual General Meeting on June 24, 2026, where CEO Masayoshi Son will likely provide updates on:
- Construction timelines for the Bosquel and Toulouse sites.
- Partnerships with AI chip manufacturers (e.g., NVIDIA, AMD).
- Financial projections for AI infrastructure revenue.
For real-time updates, monitor:
This is a pivotal moment for AI infrastructure. With SoftBank leading the charge in Europe, the question isn’t if AI will dominate the next decade—but where it will be built. The answer may well be written in France.
What do you think? Will SoftBank’s AI data centers succeed where others have struggled? Share your thoughts in the comments below—or tag us on X @worldtodayjrnl.
Like this analysis? Share it with colleagues in tech and policy: