OpenAI Considered Buying Cerebras Systems in 2017: A Missed Opportunity?
- 23/11/2024 10:11 AM
- Kevin
Back in 2017, OpenAI explored the idea of acquiring Cerebras Systems, a company that builds advanced AI hardware. At the time, Cerebras was a young startup, and OpenAI’s founders saw the potential to gain a competitive edge in the AI chip market. However, the deal didn’t happen, and OpenAI’s path took a different direction.
This revelation comes from legal filings in Elon Musk’s lawsuit against OpenAI. It sheds light on how close the company came to breaking into the semiconductor industry early on, which could have changed its trajectory.
What Was the Plan?
OpenAI co-founder Ilya Sutskever proposed the idea to Musk and CEO Sam Altman in emails. He even suggested that Tesla, Musk’s electric car company, might help fund the purchase. By September 2017, they were seriously discussing terms and doing due diligence.
However, the acquisition never moved forward. While the exact reasons are unclear, it seems OpenAI decided to step back from hardware ambitions at that time.
Why Cerebras Was a Big Deal
Cerebras, based in California, is known for its Wafer-Scale Engine (WSE), a powerful chip designed for AI tasks. These chips are considered faster and more efficient than Nvidia’s GPUs, which dominate the AI hardware market. If OpenAI had acquired Cerebras, it could have gained:
- Less Dependency on Nvidia: OpenAI relies heavily on Nvidia’s GPUs to train its AI models. Owning Cerebras could have reduced costs and reliance on external suppliers.
- Better Tools for AI: With Cerebras’ expertise, OpenAI might have developed custom hardware optimized for its needs, making model training faster and more efficient.
- Simpler Scaling: Proprietary chips could have streamlined the process of building and fine-tuning large AI models.
For Cerebras, joining OpenAI could have meant avoiding a complex IPO process and gaining funding to scale faster.
Where Cerebras Stands Today
Since the acquisition talks fell through, Cerebras has raised $715 million in funding and is preparing for a public listing. However, the company faces challenges:
- Revenue Dependence: A single client, Abu Dhabi-based G42, contributed 87% of Cerebras’ revenue in early 2024. This reliance raises financial stability concerns.
- Geopolitical Pressure: U.S. lawmakers are scrutinizing G42’s ties to China, which could complicate Cerebras’ growth plans.
- Leadership Issues: CEO Andrew Feldman has faced criticism for past controversies, including a guilty plea related to accounting violations at another company.
Despite these hurdles, Cerebras remains a key player in AI hardware, competing with giants like Nvidia.
OpenAI’s Current Hardware Strategy
Although OpenAI didn’t acquire Cerebras, it hasn’t abandoned hardware ambitions. The company has shifted to building its own AI chips, following in the footsteps of Google and Amazon. Its approach includes:
- Hiring Experts: OpenAI is recruiting chip designers and engineers to develop in-house solutions.
- Partnering with Industry Leaders: Collaborations with Broadcom and TSMC aim to produce proprietary AI chips by 2026.
- Reducing Costs: Custom chips are expected to make training and deploying AI models more affordable, addressing one of OpenAI’s biggest challenges.
What We Can Learn from This
OpenAI’s interest in acquiring Cerebras highlights how the company has always looked beyond software. It understands that hardware plays a critical role in advancing AI. If the deal had happened, OpenAI might have avoided some of the cost and scaling issues it faces today.
Still, OpenAI’s efforts to build its own chips show that the company is tackling these challenges head-on. By 2026, OpenAI hopes to launch its first proprietary chip, which could reshape how it trains and deploys its advanced AI systems.
Conclusion
The story of OpenAI and Cerebras is a fascinating “what if” moment in the history of AI. While the acquisition didn’t happen, it reflects OpenAI’s bold ambitions and its willingness to explore new frontiers.
Today, OpenAI is not just focused on creating advanced AI models—it’s also working to control the hardware infrastructure that powers them. As it builds its own chips, OpenAI aims to lead in both AI innovation and the tools that make it possible.
