My views on AI and Automotive opportunities in semiconductor industries –
AI/ML need more computational power as they are handling more data. There are multiple use cases where semiconductor intercepts with AI.
- Data center for training model
- Edge devices which will use this algo
Real time application of AI/ML algorithm requires run time optimization, and to cater the same hardware is modified. These were done earlier as hardware accelerator implemented on FPGA not chips. Model training requires high computational power.
For automotive, you can think of semiconductor devices cost in car will increase to around 30 percent. Earlier 5 percent as it contributed only to infotainment and basic sensors. As automotive industries don’t have R&D capabilities to design computer that will support driving(self driving in future), they rely on semiconductor companies for the same.
NVIDIA –
On AI data center modelling related chips, NVIDIA stands out with its exceptional computational power ( because of its GPU), but it also has significant power consumption as well. On AI edge devices there is no clear winner. Also the market is growing, and you will start seeing dedicated architecture coming for the AI related use cases.
On automotive end, Qualcomm is ahead of nvidia in the race. NVIDIA has a pipeline of 14 billion while Qualcomm is having a pipeline of 30 billion.
My take – even if nvidia is doing great, there is rationale we should have on how much market size company will cater and what can be future earnings. At this level risk to reward is not that good.
Subscribe To Our Free Newsletter |