Future of Hardware/ChipDesign
What is the future of hardware going to be like? Most folks are going into software now even EE guys. With Moore’s law reaching its EOL, is the future uncertain? What is goimg to be the emerging field? SOCs, ASICs, ? I know most of the big software companies are going in on designing their own SOCs now. eg Facebook, Google, Amazon, Microsoft.. Any insights from those in the industry will be appreciated.
Architecture will be more important than ever. I'm not convinced even the big players will have sufficient scale to make custom chips viable. It's a more exciting time to be in hardware than ever.
H/w is very global and competitive -- just look at intel's stock price over the last 18 years since the tech bubble 1.0. But there are growth areas. S/w is currently taking advantage of very low h/w and internet costs and there are several high growth areas. You can still enjoy h/w, but you'll need to pick the high growth areas that are least affected by China (e.g. China/Huawei killed telecom, even as it became more important and the driver for much of mobile -- except mobile processors). I wouldn't count on the big webscale companies to bail out the electronics industry --- just look at the size of their capex -- but they may contribute to some high growth areas like TPU, where there may be a moat against China competition (like Qualcomm was for a time).
RISC-V or gtfo
Azure
That's so Microsoft. Answer to any question irrespective of what is being asked is Azure
It's the best answer, we're using open hardware so anyone can participate in the engineering process. Go to opencompute.org to learn more.
I think AI brings some fresh air in the hw field. Lots of new companies are designing radically different chips: in memory computing, wafer scale chip, very low power, very high power. Not all of them will succeed but this explosion of creativity will move the field forward. I started hearing more hw leaders admitting the importance of sw which is healthy if it translates into action.
No future
Now that we can’t rely on just new nodes, custom architectures will become ever more important. Which means the design process needs to be faster at iterating. I’m personally excited about the likes of Catapult or Chisel for silicon design. It’d bring some much needed reuse and configuration ability to the aging and imperfect SystemVerilog
As someone who works in the area of extending Moore’s law, it’s very exciting. We will not be able to scale the same way forever, which means all kinds of interesting new innovations are happening in silicon, some revolutionary new concepts in the works.
Hardware, including silicon/semiconductor in general, isn’t going away. We live in a physical world. However, it is a commoditized industry. The margins are lower, snd you actually have to sell something tangible in order to make money (as opposed to ad revenue, media rights, etc.). The jobs do not pay as well as in SW. Old school leadership can also make it less appealing, as mentioned above.
How are they ruining the industry? Examples?