OpenAI’s Failed Bid for Cerebras: Insights from Elon Musk’s Lawsuit
OpenAI once considered buying Cerebras, an AI chipmaker planning to go public. This discussion began in 2017, shortly after Cerebras was founded and OpenAI transitioned to a nonprofit.
In an email from Ilya Sutskever, co-founder and former chief scientist at OpenAI, the idea of acquiring Cerebras through Tesla was proposed. At that time, Elon Musk was financially invested in OpenAI. Sutskever mentioned concerns about Tesla prioritizing shareholder returns, which may not align with OpenAI’s mission.
The acquisition did not happen, and OpenAI delayed its own chip development for several years. Cerebras, located in Sunnyvale, California, designs custom hardware for AI tasks. It claims its chips outperform Nvidia’s offerings.
Cerebras aims to increase its $4 billion valuation through an upcoming IPO after raising $715 million. However, it faces challenges including dependency on a single client for most of its revenue and scrutiny over ties to China. Additionally, Cerebras CEO Andrew Feldman has a controversial past regarding accounting practices.
What were the main factors that led OpenAI to reconsider the acquisition of Cerebras?
Interview with Ilya Sutskever on OpenAI’s Past Considerations of Cerebras Acquisition
NewsDirectory3: Thank you for joining us, Ilya. Let’s start from the beginning. Can you share more about the genesis of the discussions between OpenAI and Cerebras?
Ilya Sutskever: Thank you for having me. The conversations around potentially acquiring Cerebras began in 2017, shortly after Cerebras was founded. At that time, OpenAI had recently transitioned to a nonprofit model. We recognized an opportunity to leverage Cerebras’ innovative chip technology, which we believed could be a game-changer for AI development.
NewsDirectory3: The proposal involved Tesla, given Elon Musk’s financial ties to both companies. What were the implications of this connection?
Ilya Sutskever: Yes, Tesla was indeed proposed as a potential pathway for the acquisition. However, there were concerns about whether Tesla’s corporate priorities, especially regarding shareholder returns, would align with OpenAI’s mission of ensuring that AI benefits all of humanity. The nonprofit status of OpenAI and Tesla’s profit-driven model created a fundamental misalignment.
NewsDirectory3: It’s interesting to note that OpenAI ultimately delayed its own chip development for several years. What led to this shift in focus?
Ilya Sutskever: Several factors contributed to the decision to delay. After evaluating our strategic priorities, we concluded that building an in-house team of chip designers, in collaboration with well-established semiconductor firms like Broadcom and TSMC, was the most prudent path. This way, we could ensure that our goals aligned closely with both our technological ambitions and ethical commitments.
NewsDirectory3: Cerebras has announced plans for an IPO and aims to raise its $4 billion valuation. What are your thoughts on their financial path, especially considering their dependence on a single client?
Ilya Sutskever: The challenges Cerebras faces are significant, particularly their revenue reliance on a single client. While their technology shows great potential—outperforming Nvidia’s chips—having a limited customer base creates a precarious situation. An IPO could bring in much-needed capital, but they will have to address these risks to instill investor confidence.
NewsDirectory3: How would an acquisition have benefited both OpenAI and Cerebras had it gone through?
Ilya Sutskever: Had we succeeded in that acquisition, I believe it could have streamlined our chip development process. Cerebras would have avoided the complexities and public scrutiny of going public, and we would have gained access to valuable hardware resources and expertise that would have greatly accelerated our capabilities in AI technology.
NewsDirectory3: In light of everything, what is OpenAI’s current direction concerning chip development?
Ilya Sutskever: OpenAI continues to focus on building a dedicated team of chip designers, collaborating with established semiconductor manufacturers. Our first AI chip is expected to be ready by 2026, and we are committed to reducing our reliance on Nvidia, ensuring we can scale our AI technologies efficiently and ethically in the future.
NewsDirectory3: Thank you, Ilya, for providing such insight into OpenAI’s strategy and the historical context regarding Cerebras.
Ilya Sutskever: Thank you for having me. It’s crucial for us to reflect on these decisions and their implications as we continue to navigate the evolving landscape of AI technology.
If OpenAI had acquired Cerebras, it might have benefited both firms. Cerebras could have avoided a complicated IPO, while OpenAI would have gained critical resources for chip development. OpenAI has tried to lessen its reliance on Nvidia, which dominates the AI chip market.
Although OpenAI once planned to create chip manufacturing factories and considered acquisitions, it later shifted focus. OpenAI now aims to build a team of chip designers and work with semiconductor companies like Broadcom and TSMC. Its first AI chip could be ready by 2026.