SoftBank Reportedly Weighs $25 Billion OpenAI Investment 

SoftBank, startups, Light, investment cuts

SoftBank is reportedly in discussions to invest up to $25 billion into OpenAI.

The deal, as reported by the Financial Times (FT) and other outlets Wednesday (Jan. 29) evening, would see the Japanese conglomerate overtake Microsoft as OpenAI’s largest investor.

The news comes days after OpenAI and SoftBank announced they would join Oracle on an AI data center project dubbed “Stargate,” with a price tag that could reach as high as $500 billion over the next four years.

Sources tell the FT that SoftBank is in talks to invest between $15 billion and $25 billion into OpenAI, in addition to the more than $15 billion it has committed to Stargate. In all, SoftBank could spend upwards of $40 billion on its relationship with OpenAI.

SoftBank has said it wants to invest $100 billion into AI infrastructure in the U.S. over the next four years. The company had previously invested $500 million into OpenAI.

 “The talks are ongoing and the amount that SoftBank could invest in primary equity into OpenAI is a moving target,” said one source familiar with the matter.

OpenAI also plans to invest about $15 billion in Stargate. Several sources said that SoftBank’s equity investment in OpenAI could cover the latter’s commitment to Stargate.

Speaking with PYMNTS about the project earlier this week, Deborah Perry Piscione, co-founder of  the AI/Web3 advisory firm Work3 Institute, said that AI data centers are unique in that they need specialized hardware and infrastructure to deal with the massive parallel processing AI workloads require.

“Traditional data centers focus on storage and basic compute, while AI facilities need dense configurations of GPUs and AI accelerators, like Nvidia’s H100s, designed specifically for the complex matrix calculations that power AI models,” she said.

Meanwhile, debate continues in the AI world about the need for massive amounts of investments in AI projects following the debut of a model by Chinese company DeepSeek, which claims to have trained its model for only $5.58 million on 2,048 Nvidia H800 chips.

DeepSeek proved that “small companies, individual developers, and even researchers are now able to harness the power of AI without breaking the bank,” Roy Benesh, CTO at eSIMple, told PYMNTS earlier this week.

Analysts at Bank of America argue that the $5.58 million is “misleading” as DeepSeek did not include factors like research, experiments, architectures, algorithms and data. Still, the analysts said the bigger picture is that the startup introduced innovations demonstrating that less costly training is possible.