OpenAI Closes $5B Round at $400B Valuation to Fund Data Centers
OpenAI has raised $5 billion in a Series F round led by SoftBank and Thrive Capital, pushing its valuation to $400 billion. The capital will primarily accelerate data center expansion as compute demand continues to outpace supply.
Original sourceOpenAI has closed a $5 billion Series F funding round, with SoftBank and Thrive Capital leading the investment. The round values the company at $400 billion, making it one of the most valuable private companies in history and cementing its position as the dominant commercial player in the large language model space.
The capital is earmarked primarily for infrastructure — specifically data center construction and expansion — as OpenAI continues to face constraints on compute capacity. The company has been aggressively scaling its model training and inference workloads, and the bottleneck has increasingly been physical infrastructure rather than model architecture or talent.
This round follows a pattern of accelerating capital concentration in AI infrastructure. OpenAI's valuation has grown roughly 8x in under two years, a trajectory that reflects both genuine product traction — ChatGPT remains the dominant consumer AI product — and significant investor appetite for exposure to the AI buildout regardless of near-term profitability. The company is reportedly still operating at a significant loss, with inference costs and staffing expenses outpacing revenue growth.
The SoftBank involvement is notable given the firm's history of large, thesis-driven bets and its existing exposure to AI infrastructure through its Vision Fund portfolio. For OpenAI, the partnership likely extends beyond capital to include distribution relationships and potential enterprise sales channels across SoftBank's global network.
Panel Takes
The Founder
Business & Market
“A $400B valuation on a company still burning cash at scale is a bet that the infrastructure moat compounds faster than the losses do — that's not crazy, but it's also not obvious. The SoftBank lead is interesting because SoftBank doesn't just write checks, they bring distribution, and OpenAI's real ceiling is enterprise penetration, not consumer. The number I'd want to see before calling this validated: gross margin by product line, specifically whether API revenue is trending toward positive contribution margin or still subsidized by the fundraise.”
The Skeptic
Reality Check
“$400 billion for a company that is, by most accounts, losing money on every inference call is a valuation that only makes sense if you believe OpenAI will own the margin layer of AI permanently — and Google, Anthropic, Meta, and Mistral are all actively working to prevent that. The data center buildout story is real, but it's also a hedge against the risk that model commoditization forces OpenAI to compete on infrastructure rather than model quality. I'd predict the kill condition here is Microsoft, not a startup: if Azure continues building first-party models and squeezes OpenAI's API margins, a $400B valuation looks like a very expensive way to become a cloud vendor.”
The Futurist
Big Picture
“The thesis baked into this round is falsifiable: compute scarcity persists long enough that owning data center capacity becomes a durable structural advantage, not just a temporary bottleneck. That's plausible if model scaling continues to yield capability gains — the moment scaling plateaus and inference efficiency compounds, the capital advantage flips into a liability. The second-order effect nobody is talking about is geopolitical: a $400B OpenAI with SoftBank as a major backer changes the conversation about where AI infrastructure gets built and who controls it, well beyond what any single product decision does.”
The PM
Product Strategy
“The stated use case for this capital — data center expansion — is actually a product decision disguised as an infrastructure one. If OpenAI is capacity-constrained on inference, that means users are hitting rate limits, latency is degrading, and enterprise contracts are getting qualified with SLA caveats — all of which erode the core job-to-be-done of 'give me a capable AI that's reliably available.' Solving compute before it becomes a visible user problem is the right call, but the risk is that $5B in infrastructure spend locks the product roadmap to a scaling paradigm that may not be where the capability gains are coming from in two years.”