In a dramatic unveiling just a few weeks ago, Masayoshi Son, the founder of SoftBank Group, announced an ambitious project in tandem with the new U.S. government—the Star Gateway project. This initiative, created in collaboration with tech giant OpenAI, is touted to possibly exceed a staggering $500 billion in costs. With such astronomical figures being floated, Son is currently in the throes of determining a viable financing strategy to support this monumental venture.

According to insider sources, Son is exploring the realm of project financing—a method typically linked with substantial oil and gas enterprises—as a means to generate the much-needed capital. This financial strategy is employed for the construction of massive infrastructure projects, similar to the Alaska oil pipeline, and its advantages include lower upfront investment requirements from anchor investors and extended financing based on anticipated cash flows from the project.

SoftBank's current considerations involve integrating aspects of this financing structure into the Star Gateway project’s financial framework. This incorporates multiple data centers and energy generation initiatives that the project will entail. Two sources familiar with these discussions indicated that a proposed solution might see SoftBank, OpenAI, along with their partners Oracle and Abu Dhabi’s MGX, contributing roughly 10% of the total project costs, with the remaining funds being sought from the bond markets.

Last month, SoftBank and OpenAI boldly proclaimed plans to construct vast data centers with a capacity measured in tens of gigawatts. This scale could signify the largest deployment of computational power in history. Not only has groundwork commenced on the first data center site in Abilene, Texas, but there is an active search for additional locations across several states for this groundbreaking project.

Son has indicated a readiness to allocate an immediate $100 billion toward this endeavor. However, the sources revealed that the financing sources for the majority of the Star Gateway's domains are still undetermined, prompting discussions aimed at bringing additional stakeholders into the fold.

The rise of low-cost and open-source AI solutions from Chinese startup DeepSeek presents a competitive challenge to future technology suppliers. This intensifying rivalry implies diminished profit prospects for such ventures. Moreover, the ongoing acquisition efforts spearheaded by Elon Musk add another layer of uncertainty to the financing framework for Star Gateway.

If SoftBank and OpenAI were to hold equal shares in Star Gateway, with Oracle and MGX taking smaller stakes, a combined 10% equity could equate to approximately $50 billion. Thus, SoftBank and OpenAI would each need to contribute between $15 billion and just under $25 billion.

However, in practice, Star Gateway will be composed of a series of projects, each potentially costing billions. Some sources noted that, beyond the equity pool, most of the funding will likely be allocated on a per-project basis. The complexity of the project's financial structure has been heavily emphasized in recent discussions. Previous media reports indicated that SoftBank intends to cap the total equity exposure in Star Gateway at around 10%, utilizing a per-project financing approach.

The financial team responsible for Star Gateway at SoftBank is likely to consider various funding mechanisms, such as preferred shares, mezzanine debt, and senior secured loans, to amass the requisite capital. One insider stated that the mix of preferred shares and debt might depend on the involvement of additional investors for each project, which may focus on niches like semiconductors or server technologies.

According to another individual, within the framework of project financing, a conceivable allocation scenario might involve common equity accounting for 10%, preferred equity and mezzanine debt at 20%, and senior debt making up 70%. Alternatively, there could be a model where preferred equity and mezzanine debt represent 40%, while senior debt stands at 50%. These structures would fall under highly leveraged frameworks commonly seen in infrastructure projects such as pipelines, power plants, bridges, or telecom networks. Nonetheless, SoftBank could leverage Japan’s ultra-low interest rates and popularity with retail bond investors to alleviate some of the financial burdens.

At this juncture, no definitive decisions have been reached, and discussions are continuously evolving. Insiders revealed that SoftBank is in talks with an increasing number of partners, suggesting that anchor investors could ultimately opt for alternative financing forms depending on market conditions. The feasibility of this financing structure remains uncertain, as project financing terms are often contingent on cash flow forecasts—something that, for AI services, is largely unpredictable.

A spokesperson from SoftBank declined to provide any comments regarding the ongoing developments. In a recent blog post, OpenAI's Chief Financial Officer Sarah Friar assured that their two companies would follow a "structured, phased investment approach." She explained that while OpenAI, SoftBank, Oracle, and MGX are the initial equity investors, the plan is to attract more partners. "We are actively pursuing a diverse group of institutional investors to build a robust and resilient capital base," she stated. Recently, during a meeting with Japanese Prime Minister Fumio Kishida and Samsung Electronics Chairman Lee Jae-Yong, Son stood alongside OpenAI’s CEO Sam Altman, rallying support for the Star Gateway initiative.

As SoftBank hones its discussions, colossal firms are actively seeking innovative funding methods to cover skyrocketing costs associated with AI hardware. Companies like CoreWeave, supported by NVIDIA, alongside backing from Blackstone Group and Magnetar, are utilizing chips as collateral to generate funds for billion-dollar data center projects. This financial model hinges on the assumption that cutting-edge NVIDIA chips, currently valued at tens of thousands of dollars, will retain their value over time.

Bloomberg analysts Marvin Lo, Sharon Chen, and Chris Muckensturm noted, "SoftBank's investment in OpenAI is primarily focused on proprietary large language models, with total investments potentially ranging from $15 billion to $25 billion. In a market increasingly leaning towards low-cost and open-source models, this may pose risks." The growing prominence of DeepSeek indicates that companies are beginning to prioritize the customization, control, and cost-effectiveness of these models. While Meta focuses on open-source, tech giants like Google and Alibaba have also joined the fray, releasing more affordable open-source alternatives alongside their proprietary LLMs. This emerging trend may democratize AI technology, making it accessible for the masses and potentially pressuring OpenAI's return on investment—rendering the current lofty valuations questionable.

With uncertainty swirling around how SoftBank will fund the Star Gateway initiative, considerable concern looms over the investment community regarding Son’s propensity for high-stakes gambles. Analysts estimate that SoftBank will report a net loss of approximately ¥155 billion (around $1 billion) in its third-quarter earnings report later this week. Kirk Boodry, an analyst at Astris Advisory, further noted that after stocks like Coupang, Inc. surrendered some of their quarterly gains, the Vision Fund's public portfolio experienced a loss of about $700 million in the fourth quarter of last year.

The performance of SoftBank's profits is intricately linked to a multitude of unlisted startups on its balance sheet. One such entity is ByteDance, which is under immense pressure to sell its widely popular video application TikTok in the U.S. or face a potential shutdown. Another noteworthy private asset is the Indonesian agritech firm eFishery, which just two years ago was valued at $1.2 billion but now faces allegations of fraudulent accounting practices and potential liquidation.