The Elon Musk-backed artificial intelligence startup, xAI, has decided to end negotiations with Oracle for a potential $10 billion deal that involved renting GPU servers in the company’s cloud to build its AI training infrastructure.
Change of Strategy: Building Data Centers
According to The Information, xAI has chosen to purchase chips to build its own data center instead of partnering with Oracle. This decision reflects a significant strategic shift, as the startup now plans to train its next flagship system in-house using Nvidia H100 GPUs. Elon Musk stated on the X social network that “our fundamental competitiveness depends on being faster than any other AI company.”
Details of the Grok 2 Model
Musk also revealed that xAI’s Grok 2 model was trained on 24,000 Nvidia H100 chips provided by Oracle and is “likely ready for launch next month” after additional tuning and bug fixes. This development underscores xAI’s ability to rapidly progress in the competitive field of artificial intelligence.
Oracle’s Offer
During a September 2023 earnings call, Oracle’s chairman, Larry Ellison, mentioned that his company had an agreement with xAI to provide cloud infrastructure for training their AI models, though he did not disclose the value or duration of the deal. The Information reported that Musk demanded Oracle to build a supercomputer faster than the company deemed possible, raising concerns about the technical feasibility of the project.
Logistical and Technical Concerns
In addition to performance demands, xAI also expressed concerns about Oracle’s preferred location, which reportedly lacked sufficient electrical supply to support the necessary infrastructure. These logistical and technical issues contributed to xAI’s decision to walk away from the negotiations and opt for an in-house solution.
In summary, Elon wants control over his computing
xAI’s decision to end talks with Oracle and build its own data center infrastructure marks a significant milestone in the startup’s strategy. This move highlights xAI’s determination to maintain its competitiveness in the AI field, ensuring full control over its AI training resources. As the artificial intelligence industry continues to evolve, the ability to quickly adapt to technological needs will be crucial for companies to succeed in this rapidly growing sector.