Home AI Infrastructure NewsHow AMD’s bold move shakes up the AI hardware landscape

How AMD’s bold move shakes up the AI hardware landscape

by Susana SchwartzSusana Schwartz
0 comments
amd

How AMD’s bold move shakes up the AI hardware landscape

amd

The OpenAI-AMD partnership could massively expand OpenAI’s compute capacity for GPT-5 while positioning AMD’s chips as a strong alternative to Nvidia’s

In sum – what to know:

Turning point – The deal marks a turning point for AMD, which is jockeying for position as a viable alternative to Nvidia in the market for high-performance AI chips.

Easing dependence – By partnering with AMD, OpenAI eases the industry’s dependence on a single vendor while also diversifying its own supply chain for more rapid infrastructure improvements.

AMD GPUs at OpenAI facilities – Collaborative AI infrastructure will drive installations of AMD Instinct GPUs at OpenAI facilities, which could include existing sites and new data centers under construction.

As we reported yesterday, AMD and OpenAI announced a multi-year partnership to deploy 6 gigawatts of the former’s Instinct GPUs, a pivotal announcement for competition in the AI infrastructure market, which until now was dominated by Nvidia’s CUDA-based ecosystem. 

In challenging Nvidia in the AI chip market, AMD is opening the door to hardware diversity, which could mean better scalability, performance, cost efficiency, and innovation for not only OpenAI, but for the enterprises and institutions that build on OpenAI’s models.

The deal elevates the standing of AMD’s open-source ROCm software stack and boosts its technical roadmap with OpenAI’s financial backing. OpenAI’s option to acquire a 10% equity stake in AMD means there will be a symbiosis in which OpenAI’s financial success will depend, in part, on AMD’s performance; conversely, AMD’s traction as an AI accelerator will depend, in large part, on OpenAI’s continued momentum with GPT-5 and more advanced models. In interviews yesterday, AMD chief Lisa Su indicated the company would possibly get tens-of-billions of dollars in new revenue over the next five years, which could help it creep into advanced AI systems, a market that has eluded AMD until now.

This deal is significant for OpenAI in that it diversifies the supply chain of cloud service providers and semiconductor companies that it can tap, including Nvidia, Broadcom, and now AMD. This diversification could accelerate infrastructure improvements, addressing the cacophony of complaints about GPT-5 lag times. By infusing OpenAI’s infrastructure with large quantities of AMD’s Instinct GPUs, OpenAI may achieve the scale of computing capacity and memory needed for its more advanced models and the larger number of users it aims to add by year’s end.

For perspective, OpenAI’s 800 million weekly active users currently generate more than 8 billion tokens per minute in the ChatGPT API. With  CEO Sam Altman’s aspirations to reach 1 billion users by end of year, it is predicted that dozens of “Stargate-class” data centers and an estimated 20 to 100 gigawatts (GW) of computer power would be needed. Scaling would require hundreds-of-thousands of specialized AI GPUs, and with this partnership, Altman may get some of the computational muscle he needs, especially in locations with new or planned OpenAI facilities, such as Texas, New Mexico, and Ohio.

AMD is paying a high price to participate in the AI chip market, but it secures a long-term revenue stream from OpenAI, and garners an important endorsement of its Instinct GPU. OpenAI also takes a risk, banking on the performance of those GPUs, but if successful, it shifts OpenAI (and others) away from a near-monopoly to a multivendor ecosystem. Still unclear is the long-term impact on the massive infrastructure buildouts necessary for the many AI-driven ecosystems that are emerging in the AI industry. For example, AWS, Microsoft Azure, Google Cloud and other hyperscalers are competing for the same scarcity of manufacturing resources, so how this multi-year, multi-gigawatt deployment will impact others seeking out GPUs, remains to be seen.

You may also like

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More