The cooperation between OpenAI and Broadcom has attracted widespread attention, and both parties plan to develop customized inference chips to optimize the efficiency of AI services and reduce cloud computing costs. This move not only demonstrates OpenAI's strategic layout in the hardware field, but also indicates a new direction for the future development of AI technology.
Recently, it has been reported that OpenAI is working with Broadcom to develop a customized inference chip. According to Reuters, the discussions between the two companies are very confidential, and Taiwan Semiconductor Manufacturing Corporation (TSMC) may serve as the foundry for the project. This news has sparked widespread speculation from the outside world about the future development direction of OpenAI.
Image source notes: The image is generated by AI, and the image authorized service provider Midjourney
So, why does OpenAI need its own inference chip? First of all, OpenAI's cloud computing costs are very huge. Although partners like Microsoft support some of the costs, controlling the hardware itself can undoubtedly significantly reduce operating costs. Many companies have found that choosing to build their own data centers is much more economical than renting cloud services.
In addition, developing dedicated chips that are adapted to their own services may also be a strategic goal of OpenAI. As we all know, AI applications consume a lot of energy, so by optimizing the synergy between hardware and software, OpenAI services will become more efficient.
OpenAI is also showing investors the idea of building large data centers that are specifically designed to run AI services that may also cost less to build or operate if equipped with custom chips. What's more, the consideration of decentralized supply chains cannot be ignored. Due to the limited global semiconductor production capacity, the risk of relying on external suppliers exists, and developing own chips can reduce dependence on third-party products.
While we can’t imagine OpenAI would be willing to enter the hassle industry of hardware sales, as it requires a lot of actual investment and increases the number of employees, OpenAI might be deployed at the edge of the network when inference tasks usually need to be as close to users as possible Related devices, like many content distribution networks and Netflix, this architecture is definitely a good idea.
Speaking of reasoning chips, it is actually not unfamiliar with them in the market. Inferentia of AWS, Google's Tensor Processing Unit (TPU), and Microsoft's Maia wafers can handle inference and training workloads.
Interestingly, news of OpenAI's partnership with Broadcom also drove the latter's stock price to rise slightly. Broadcom's latest quarterly earnings report shows that it is expected to sell $12 billion in AI wafers this fiscal year, a figure of $1 billion higher than previous expectations, but investors' reactions seem a bit disappointed. Therefore, working with this hottest name in the field of AI software will undoubtedly make Wall Street even more excited.
Key points:
OpenAI and Broadcom are negotiating to develop customized inference chips to reduce cloud computing costs.
The own chip can optimize the collaboration between hardware and software and improve the efficiency of AI service.
Broadcom expects to sell US$12 billion in AI silicon wafers this fiscal year, and the news of cooperation boosts stock prices.
The cooperation between OpenAI and Broadcom is not only a technological breakthrough, but also an important step in the future development of AI. By customizing chips, OpenAI is expected to occupy a more advantageous position in the field of AI.