What you need to know
- Microsoft is reportedly working on its own chip to train and enhance its AI.
- The chip has reportedly been secretly tested since 2019 and could be available with Microsoft and OpenAI by next year.
- By making its own chips, Microsoft can be less dependent on NVIDIA, which currently has a strong foothold in the AI server space.
According to a recent report in The Information, Microsoft has plans to create its own AI chip. (opens in new tab)In fact, the report says the tech giant has been working on a chip in secret since 2019. You can save money and move away from relying heavily on NVIDIA.
Some Microsoft and OpenAI employees already have chips that can run tests on large language models such as GPT-4. If Microsoft can train his LLM using its own chips rather than his NVIDIA GPU to power its AI, it could lead to significant savings.
Microsoft invests billions in OpenAI to power Bing Chat with GPT-4. AI is also integrated into several Microsoft services. If Microsoft succeeds in creating a chip that can train and enhance AI, Microsoft can directly reduce costs by using its own hardware and also benefit from improvements in the artificial intelligence that its services rely on. increase.
NVIDIA currently dominates the AI server space. It is estimated that ChatGPT may need 30,000 of his NVIDIA GPUs to meet demand. NVIDIA A1000 GPUs are priced between $10,000 and $15,000, so NVIDIA could make him as much as $300 million just from the GPUs that power ChatGPT. This is just an estimate, but it gives you an idea of the range we’re looking at.
Amazon, Google and Meta are already making their own AI chips. Microsoft’s follow-up lawsuit could give the company more control over its hardware, potentially saving millions and even billions of dollars.
