2 min read
OpenAI has publicly confirmed that it currently has “no active plans to use Google’s in-house chip” to power its AI systems.
This clarification comes shortly after reports by Reuters and other outlets speculated that the AI research company might be adopting Google’s tensor processing units (TPUs) to support its expanding computing needs.
While a spokesperson for OpenAI acknowledged that the company is in the early testing phase with some of Google’s TPUs, they emphasized that the lab does not intend to deploy them at scale at this time.
As of now, OpenAI remains committed to Nvidia’s GPUs and AMD’s AI chips, which currently form the backbone of its computing infrastructure. The company is also working on developing its own custom chip, with plans to reach the “tape-out” stage, the finalization of the chip’s design for manufacturing, later this year.
Interestingly, OpenAI has signed up for Google Cloud services to help meet its growing demand for computing power. This collaboration between two of the biggest competitors in the AI space has raised some industry eyebrows. However, the bulk of OpenAI’s compute power will still come from GPU servers provided by CoreWeave, a rapidly growing neocloud company.
Google Expands Access to TPUs
In recent years, Google has made its once-exclusive TPUs available to external customers. This move has attracted major tech players such as Apple, as well as AI startups like Anthropic and Safe Superintelligence, both of which were founded by former OpenAI team members and are considered direct competitors to ChatGPT.
Despite this increasing availability and adoption, OpenAI appears to be taking a cautious approach, opting to continue using its current hardware providers while testing alternatives without full-scale commitment.
Though OpenAI is exploring new chip options, it remains focused on its current hardware setup and the development of its in-house chip solution. For now, OpenAI is testing Google’s TPUs but has not adopted them, reinforcing its strategy of carefully planning infrastructure to meet soaring AI demands.
