OpenAI and Oracle Expand $500B Stargate Project to 5 GW Capacity as Musk's xAI Aims for 50M H100 Units in 5 Years

OpenAI and Oracle have expanded their $500 billion Stargate project, a major AI infrastructure initiative, while Elon Musk’s xAI revealed an ambitious five-year plan to deploy 50 million H100-scale AI chips. The collaboration between OpenAI and Oracle now includes a total of 5 gigawatts (GW) of data center capacity, sufficient to power over 2 million AI chips. Oracle’s latest contribution includes a $40 billion investment in Nvidia’s GB200 chips for a new data center in Abilene, Texas, furthering OpenAI’s goal to scale compute capacity across the U.S. [1]. OpenAI CEO Sam Altman has emphasized the project’s significance, stating it is “a gigantic infrastructure project” and confirming that over 1 million GPUs will be operational by year-end [1].

Musk’s xAI, meanwhile, aims to achieve 50 million H100-scale units within five years, a milestone that would represent a 500-fold increase over the compute power of the world’s most advanced AI supercomputer from a year ago, according to estimates by a X user [4]. The xAI project’s Colossus 2 supercomputer, set to launch soon, will utilize 550,000 GB200 chips—equivalent to 5.5 million H100s—highlighting the competitive edge Musk is targeting [4]. This aggressive timeline positions xAI as a direct rival to OpenAI’s Stargate, as both firms race to dominate the AI landscape through infrastructure and computational scale [1].

Despite the momentum, the Stargate project faces notable hurdles. A recent Wall Street Journal report highlighted delays and internal disagreements between key partners, including OpenAI and SoftBank. Initially promising to deploy $100 billion immediately, the project has scaled back its near-term goals to constructing a single data center by year-end [4]. Analysts suggest that the absence of concrete infrastructure deals six months after the project’s launch raises questions about its execution timeline, even as Oracle and OpenAI emphasize their long-term ambitions [4].

The strategic moves by OpenAI and xAI reflect broader industry trends, where tech giants are prioritizing infrastructure scalability to secure market leadership. Oracle’s 5 GW capacity—comprising 4.5 GW of new U.S. data center development—aims to support both OpenAI’s and xAI’s ambitions, underscoring the interdependence of major players in the AI sector [1]. However, challenges such as energy sourcing and chip supply chain bottlenecks remain critical constraints for large-scale projects [4].

While OpenAI and Oracle focus on expanding compute capacity to meet surging demand, xAI’s strategy leans on rapid unit deployment to establish dominance. The Texas data center, powered by Nvidia’s GB200 chips, is expected to be pivotal for OpenAI’s Stargate goals, but the project’s success will depend on resolving logistical and financial complexities [2]. Analysts argue that the competition between these initiatives could accelerate innovation in energy-efficient computing, yet execution risks persist amid the high stakes [4].

Source:

[1] [OpenAI Expands Stargate, Musk Reveals xAI's 50M H100](https://cointelegraph.com/news/openai-oracle-stargate-expansion-musk-xai-h100-plan)

[2] [Abilene Tx Data Center](https://www.tiktok.com/discover/abilene-tx-data-center)

[4] [AI Daily News July 22 2025](https://www.reddit.com/r/learnmachinelearning/comments/1m6n44l/ai_daily_news_july_22_2025_openais_500b_project/)