SoftBank’s big plans for a $500 billion AI infrastructure project, called Stargate, are falling behind schedule. During the company’s recent earnings call, CFO Yoshimitsu Goto admitted things are taking longer than expected. He said the project is moving “slower than usual,” which is a setback for such a massive initiative.

This delay comes about seven months after then-President Trump announced a similar data center push. The project faces familiar hurdles, like picking the right sites and getting everyone on the same page. Building these huge AI infrastructure setups isn’t easy, and the process involves many stakeholders, from engineers to regulators.

What’s causing the delays?

Goto explained that finding the right location is a key challenge. “There are many options,” he said, “and it takes time to select a good site.” Plus, coordinating with different groups—construction teams, local authorities, utility providers—is complicated. Reaching consensus and solving technical issues naturally slows things down.

Despite the slower pace, Goto remains confident. He said SoftBank wants to carefully build the first model of Stargate, even if it takes longer than planned. The company still aims to reach its original goal of investing 346 billion yen (about $3.2 billion) over four years. Some major sites have already been chosen in the U.S., and preparations are underway on multiple fronts.

Requests for comments from Stargate partners like Nvidia, OpenAI, and Oracle have gone unanswered so far.

Lessons for other CIOs planning AI infrastructure

These delays highlight important lessons for IT leaders. Experts say that what SoftBank faces is common. Sanchit Vir Gogia from Greyhound Research points out that delays in partner onboarding and service setup are typical hurdles. Oishi Mazumder from Everest Group adds that these delays show AI infrastructure isn’t just about having powerful servers but also about land, energy, and stakeholder cooperation.

Mazumder emphasizes that building AI infrastructure is a complex, long-term effort involving many parts of an organization and outside players. It’s not just a simple upgrade but a transformation that requires careful ecosystem-wide planning. Utilities, regulators, hardware suppliers, and construction crews all have their own timelines, which can clash or cause delays.

The costs for supporting AI data centers are huge. Goldman Sachs estimates that about $720 billion will need to be spent on grid upgrades through 2030. Companies need to find a balance—deploy quickly but carefully. McKinsey suggests breaking projects into stages instead of trying to do everything at once. Mazumder warns that even phased plans can stall if early coordination isn’t done properly. He advises organizations to expect multi-year timelines and treat AI infrastructure as a major capital project.

Adapting planning strategies for AI growth

Learning from Stargate’s hiccups, experts recommend a more practical approach. Instead of waiting for massive, all-at-once projects, CIOs should focus on modular, hybrid strategies. Deploying workloads across multiple clouds and in smaller phases can keep progress moving, even if some sites face delays.

Gogia warns that relying too heavily on one flagship facility can be risky. It’s better to plan for external factors and coordinate closely with all involved providers. This way, companies can adjust and resequence their deployment plans without losing momentum.

Many enterprises already using Arm-based chips show that there are alternatives to waiting for huge projects to finish. Smaller, adaptable infrastructure upgrades can help organizations stay ahead while waiting for larger initiatives like Stargate to fully mature.

In summary, the delays in SoftBank’s Stargate project are a reminder that building massive AI infrastructure isn’t just a technical challenge—it’s a complex orchestration involving land, energy, stakeholders, and careful planning. Companies looking to scale AI should adopt flexible, phased approaches and build strong coordination into their strategies.

Inspired by

• https://www.networkworld.com/article/4036679/stargates-slow-start-reveals-the-real-bottlenecks-in-scaling-ai-infrastructure.html

Sources

• gmpg.org
• seekingalpha.com
• computerworld.com
• cio.com
• greyhoundresearch.com