Why Are Tech Giants Microsoft And OpenAI Looking To Develop AI Chips Internally?

AI chips have dominated recent AI events – but there’s a new trend emerging!

According to a recent report, Microsoft is planning to release its first AI chip next month. As a result of strong demand and limited supply, Nvidia's graphics processing unit (GPU) chips have become a bottleneck for Microsoft's operations. #WaitWhat

Let's get down to the nitty-gritty of this silicon fight by "chipping" away at the specifics!
The Athena AI chip was developed by Microsoft specifically for use in enterprise data centers. Microsoft and other cloud providers employ NVIDIA's flagship H100 GPU to run large language models (LLMs) and other AI applications, so this new GPU is likely to give the chip manufacturer a run for their money.

Microsoft's Ignite conference, which will take place from November 14-17, is widely anticipated to be where the chip is first shown off to the public.

Athena's potential release coincides with a period of increased interest in artificial intelligence processors. In particular, LLMs have high demands on computational resources, leading to a severe shortage of AI processing chips and consequently increased costs.

OpenAI, the Microsoft-backed AI research and development company, is apparently also looking at creating its own AI processors. Moreover, the fact that both Google and Amazon are also working on their own AI chips indicates a bright but unpredictable future for the AI chip industry.

 

Why Is OpenAI Considering AI Chip Manufacturing?

According to Reuters' sources, OpenAI is actively investigating prospective acquisition targets in order to further its AI chip manufacturing goals.

Due to the significant lack of processors needed for AI technology, OpenAI has been considering alternatives. If OpenAI follows through with its plan to manufacture its own chips, it would join the ranks of other tech giants like Google and Amazon that have already made the transition to fabricating their own chips in-house. OpenAI's potential chip fabrication might lessen its reliance on other suppliers and help it keep up with the growing demand for specialised AI processors.

OpenAI CEO Sam Altman expressed alarm about the chip scarcity earlier this year, which has caused delays in the company's programmes. In a deleted blog post, Altman discussed potential AI chip production with CEO of Humanloop, a LLM training company, Raza Habib, saying, “A common theme that came up throughout the discussion was that currently, OpenAI is extremely GPU-limited and this is delaying a lot of their short-term plans. The biggest customer complaint was about the reliability and speed of the API.

Sam acknowledged their concern and explained that most of the issue was a result of GPU shortages. Their finetuning API is also currently bottlenecked by GPU availability. They don’t yet use efficient finetuning methods such as Adapters or LoRa, so finetuning is very compute-intensive to run and manage.

While better support for finetuning will come in the future, dedicated capacity offering is currently limited by GPU availability.


With Microsoft and OpenAI considering rolling out their first AI chips, are we witnessing a game-changing trend in the AI world? How will the move affect AI chip manufacturers such as Nvidia?

Let us know your thoughts in the comments section below!