Advanced Micro Devices is following in NVIDIA's footsteps by teaming up with a competing chip company.

On Thursday, Advanced Micro Devices announced a partnership with Cerebras, a rival in the AI server chip space, enabling AMD server racks to support Cerebras' wafer-scale chips. This allows both types of chips to work together in parallel. The setup is a distributed inference architecture: a single AI task is split across different chips to boost overall computing speed and efficiency.

The most compute-intensive part of AI processing, known as the prefill stage, will be handled by Advanced Micro Devices' graphics processing units (GPUs). The remaining decode stage will be managed by Cerebras' chips. NVIDIA has previously launched similar collaborative products with several other chip rivals.

Cerebras Chief Executive Officer Andrew Feldman stated on stage at AMD's annual conference for chip buyers and lessors that the joint product from Advanced Micro Devices and Cerebras will officially launch later this year. Following the news of this partnership, Cerebras' share price rose by 7%.

Advanced Micro Devices' new server rack, named Helios, is equipped with 72 AMD GPUs and is positioned to compete with NVIDIA's upcoming Vera Rubin rack. Demand for AI servers is currently strong, and NVIDIA's top-tier customers are also key clients for Advanced Micro Devices, including companies like OpenAI, Anthropic, Microsoft, and Meta.