A naukri.com initiative
Medium
4w
187
Image Credit: Medium
AI Needs Chips
- For AI to be transformative, it requires a lot of enablers such as semiconductors to scale or go to the cloud or the data center.
- Nvidia is AI chip leader and the 800-pound gorilla in the space that deserves most of the credit.
- Intel has lost the plot and cannot keep up with AMD and Nvidia’s high-performance chips.
- TSMC dominates semiconductor manufacturing and has nearly 60% market share in 2023, rising to 90% for most sophisticated chips.
- Nvidia’s success is rooted in their CUDA software platform that keeps improving to deliver better products for customers.
- DPU allows data center computers to operate more efficiently at scale and allows the CPU and GPU to focus on higher order compute issues.
- Cerebras builds a wafer-scale solution that is basically a super computer on a single chip, aimed to bring performance while not shrinking the chip.
- Groq's LPU is built to challenge Nvidia in LLMs computation with a 10X improvement specifically for LLM acceleration and 1/10th energy used compared to other solutions.
- SambaNova has raised $1 Billion for their architectural approach that improves LLM performance in a smaller footprint with lower energy usage.
- The AI chip market is forecast to have a CAGR of 20% per year for the foreseeable future, becoming a target for new competitors.
Read Full Article
11 Likes
For uninterrupted reading, download the app