OpenAI Closes $3B Round at $300B Valuation to Fund Infrastructure
OpenAI has closed a $3 billion funding round led by SoftBank Vision Fund, valuing the company at $300 billion. The capital is designated for data center expansion and accelerating the o-series reasoning model roadmap.
Original sourceOpenAI has secured $3 billion in new funding led by SoftBank Vision Fund, pushing its valuation to $300 billion — up from the $157 billion it commanded during its October 2024 raise. The round signals continued institutional appetite for frontier AI infrastructure despite ongoing uncertainty about long-term unit economics at this scale.
The funds are earmarked for two specific priorities: expanded data center capacity and faster development of OpenAI's o-series reasoning models. The o-series line, which includes o1, o3, and their variants, requires significantly more compute at inference time than standard transformer completions, making infrastructure investment a direct product requirement rather than speculative buildout.
The SoftBank-led structure is notable given the firm's ongoing $500 billion Stargate joint venture with OpenAI, which was announced in early 2025 to build out domestic AI infrastructure. This additional $3 billion raise suggests the capital requirements for frontier model development continue to outpace initial projections, even within a framework already designed to address that problem.
At $300 billion, OpenAI is now valued above many established Fortune 100 companies, a figure that will invite scrutiny of its revenue trajectory and path to profitability. OpenAI has previously reported annualized revenue approaching $5 billion, though it continues to operate at a significant loss as compute costs, talent acquisition, and infrastructure spending consume capital faster than the subscription and API business generates it.
Panel Takes
The Founder
Business & Market
“At $300 billion on roughly $5 billion ARR, you're pricing in a 60x revenue multiple on a company that's burning faster than it's earning — that's not a valuation, it's a futures contract on AGI. The SoftBank relationship is a distribution and capital dependency wrapped in a partnership bow, and Masayoshi Son's track record of late-stage mega-rounds should make any operator nervous. The moat question hasn't changed: if the o-series models are what's worth funding, the question is whether OpenAI owns the compounding data loop or whether it's just the most expensive GPU renter on the planet.”
The Skeptic
Reality Check
“This round doesn't validate the business model — it delays the reckoning with it. OpenAI is raising $3 billion inside a Stargate framework that already exists to solve the capital problem, which means either Stargate isn't working as projected or the compute appetite for o-series is materially larger than disclosed. What kills this in 12 months isn't competition — it's that inference costs for reasoning models don't fall fast enough to make the API business profitable before the next raise is needed, and at some point SoftBank's patience has a price.”
The Futurist
Big Picture
“The thesis embedded in this raise is specific and falsifiable: reasoning-heavy inference will become the dominant compute workload within 24 months, and whoever owns the inference infrastructure at scale will capture disproportionate margin. The dependency that has to hold is that o-series-style chain-of-thought reasoning doesn't get commoditized by open-weight models before OpenAI can build switching costs through enterprise workflow integration. The second-order effect nobody is talking about is what a $300 billion OpenAI does to the talent market — at that valuation, the gravitational pull on researchers makes every competing lab's retention problem structurally harder.”
The PM
Product Strategy
“The earmarking of funds toward the o-series roadmap is the most product-relevant signal here — it tells you OpenAI has made a deliberate bet that the job-to-be-done worth owning is 'solve hard problems,' not 'generate fast output,' and they're willing to pay the inference cost premium to own that position. The risk is that the product surface for reasoning models is still underdeveloped: most users interact with o3 the same way they interact with GPT-4o, which means the distinct capability isn't translating into distinct usage. More data centers don't fix a product clarity problem.”