Already, the question is not if AI will reshape global industries, but how fast its infrastructure scales up to meet that need. A new $40 billion funding round values OpenAI at $300 billion, nearly double from October, making it one of the most valuable private firms, a league including SpaceX and ByteDance. The enormous investment of capital into making software more capable and robust is tied inextricably to one of the most ambitious hardware projects ever: the Stargate data center network.
It had committed $10 billion in April and another $30 billion by year‑end, pending OpenAI’s complete transition into a so-called public benefit corporation. This makes formal a dual mandate returning investor capital while pursuing a chartered public mission and is aimed at courting long‑term capital without abandoning commitments to AI safety, health research, and education. In the new structure, nonprofit OpenAI Foundation retains a 26% stake worth about $130 billion, while Microsoft holds 27% valued at $135 billion, its technology rights secured through 2032.
Of the funding, approximately $18 billion is going to Stargate, its $500 billion hyperscale data center program which it has co-designed with SoftBank and Oracle. So much of Stargate’s engineering vision is inspired by scaling laws for multimodal large language models: the more compute and the more data, the better the performance. To fuel model development, OpenAI is building several 1–2 GW campuses across the U.S., the first of which is already live in Abilene, Texas. The first phase on the site delivers more than 200 MW of capacity, powered up in September 2025, but this will expand beyond 1 GW. Oracle will lease the site for 15 years, and is negotiating deals of up to $40 billion for Nvidia GB200 GPUs for deployment.
The scale of Stargate’s buildout rivals that of whole regional data center markets. OpenAI has commitments for 4.5 GW in additional U.S. capacity through Oracle partnerships, plus projects in Michigan, Wisconsin, New Mexico, and other states. Planned international sites will be in the UAE, Norway, and the UK, while initial exploratory talks are happening in South Korea, Argentina, and Canada. On the engineering side, some of the challenges faced include multi-gigawatt power provisioning, multi-campus training on fiber bandwidth, and rack design for up to 7,000-pound GPUs. Networking constraints also limit single-site clusters past 2 GW, driving multi-campus architectures with synchronous and asynchronous training methods.
Another one of the key technical challenges is that of power stability. AI training clusters can create enormous load swings on utility grids. On-chip power management, UPS buffering, and capacitor-based isolation have been proposed in plans by OpenAI, Nvidia, and Microsoft to ease the problem. Others, such as xAI, have used large Tesla Megapacks to absorb the fluctuations. Vantage, developer of the Stargate Wisconsin facility, promised net-new renewable generation and battery storage, while the Michigan site will boast its own dedicated battery infrastructure.
This funding round lands in a venture capital environment defined by extreme concentration. Global VC reached $97 billion in Q3 2025, with 46.4% $45.1 billion flowing into AI. Mega-rounds of $500 million or over captured more than 30% of total capital, and just 18 firms took one-third of all funding. It was a late-stage world, too, with deals totaling $58 billion; early-stage rounds saw only modest growth. OpenAI’s $40 billion raise is nearly triple the largest private tech round on record before it, dwarfing Ant Group’s $14 billion in 2018.
The infrastructure race also reflects GPU supply constraints and hardware-software co-design. Complementing the $100 billion in planned investment from Nvidia to deliver 10 GW of AI data centers in OpenAI will be AMD’s 6 GW supply agreement and Broadcom’s 10 GW of custom accelerators. OpenAI’s compute planning team needs to balance across heterogeneous hardware: Blackwell GPUs for memory-intensive multimodal workloads, TPUs for specific inference tasks, and custom silicon for specialized reasoning models.
Given CEO Sam Altman’s increased focus on research and product, operations during this capital-intensive expansion will fall to COO Brad Lightcap. Usage of ChatGPT swelled 100 million in a month to 500 million weekly active users, while OpenAI projects revenue to triple to $12.7 billion by year‑end. It is this unprecedented funding, structural realignment, and hyperscale infrastructure execution that positions OpenAI at the center of a generational technology build‑out-one where engineering capacity, not just algorithmic innovation, will determine the pace of AI’s next leap.
