Meta gave up on GPU and CPU to take a RISC-y route for AI

Dec 16, 2023 ·

Meta, a startup making AI-driven accelerators, has made a bold move in the field of AI training and inference acceleration: they have chosen to abandon traditional GPU and CPU architectures in favor of a new approach. Instead of relying on established silicon chips, Meta is leveraging technology from RISC-V, an open-source instruction set architecture.

This will be a major blow to Nvidia and AMD, who are used to dominating the market for AI-related accelerators. By not relying on their hardware, Meta is now able to offer customers access to cheaper AI processing capabilities. Additionally, by having an open-source architecture, customers can customize and tailor solutions to their particular needs. Meta can offer greater flexibility, scalability, and customization than Nvidia or AMD.

The implications of this move could be huge for the industry. Nvidia's position of being a one-stop shop for AI acceleration products could be weakened if Meta is successful and other companies follow suit. Additionally, GPU-based systems are often too slow and expensive for certain applications, so Meta’s RISC-V-based approach could provide faster, more efficient processing at lower cost.

Finally, with the emergence of Edge computing, there is an increasing need for specialized accelerators that are both powerful and power-efficient. This is another area where Meta's RISC-V architecture could have an advantage over traditional GPUs or CPUs.

Overall, Meta’s decision to use RISC-V for AI acceleration could prove to be a game changer for the industry. With the promise of cheaper, faster, and more flexible AI solutions, it looks like Nvidia and AMD may have some serious competition on their hands.