First, I agree that the industry has started slowing down. This is because of physical limitation of the silicon technology. We as researchers have not yet found a material that can replace silicon or have not yet perfected Quantum computing (although promising it is far from becoming true).

However, I disagree with those who believe that the industry will die or collapse. The fundamental reason is, VLSI is the foundation of all industry! if VLSI collapses (Chip industry or silicon industry comes under same domain), the rest of the industry will eventually collapse due to a mere dependence of computers! So, the industry has to stay and it will. Innovations and creativity required to work on this industry would increase. Rapid changes as occurred in the 1990s -2000 may not happen unless it’s disruptive! (should give 100x improvement over current technology which is bleak).

Demand for engineers will increase in slow phase or it may fall down slightly or say constant. Layoffs will occur eventually ( you can already see layoffs in this industry). Specialized skill sets creative thinking, innovations and envisions the future will become crucial skill sets. The industry will sustain itself with a slow growth period until disruptive technology emerges like QC. Customized solutions will emerge in the market like accelerator designs.

There is a renewed interest where people have started concentrating on merging VLSI + Computer architecture and system designs. The performance improvement from Moore’s law scaling is providing diminishing returns in terms of cost and yield. Thus, several researchers have started expanding the work towards building software-hardware co-designs to meet performance requirements. Parallel algorithms are evolving slowly and new students are encouraged to write massively parallel algorithms (not all algorithms can be parallelized but many could be).

Personally – I believe the VLSI industry has dug deep only one sub-domain,i.e silicon technology! Rest of the domains (software architecture, parallelism, virtual memory/cores, security, ASIC, system design, etc) I believe they have merely scratched the surface! The industry is exploring data science methods to understand VLSI and generate new insights that may assist them to generate higher profits. On the other hand, the research community has already understood the performance limitations provided by the software. Hardware-software co-design is providing a good return in performance on each design from generation to generation (check Nvidia, Intel, IBM generation trends). Customized ASIC or FPGA are being developed to provide higher performance for segmented markets. (bitcoins, autonomous vehicles, for examples). Vertical efforts in sub-domains of VLSI has started to see some rapid changes (GPU/ML Accelerators). However, these vertical efforts require specialized skills and understanding across the industries. AI/ML is yet to reach VLSI (for example, power estimation and sleep state changes can be learned over time using ML models in PMU or area estimations can be quickly converged using inference algorithms or better memory access pattern for prefetching can be deployed and many more). A lot is going to change when AI/ML starts integrating with VLSI!

Thus, the future is bright and exciting !!!

From Layla-Ec.Com

Follow me!