Hmmm what is a simple "hello world" project in chip design?
In computer science courses, that's as simple as a println().
In machine learning courses, that's training on mnist dataset to do character recognition.
In electrical engineering, that's buying a raspberry pi to blink led.
In chip design ... Chatgpt says to design a 1-bit full adder using verilog?
...
I understand why the article thinks the market is looking for graduate education. To design a simple chip requires an *initial investment* (as with all hardware startups really). This is different from software where one can simply launch a web app with a container hosted on your preferred cloud provider...
... That said, with the rise of LLMs lowering the barrier of entry of software even lower (e.g. vibe coding), may we see more rise of hardware startups/innovations?
FPGA dev boards are cheap nowadays, and you can start coding in a hardware definition language with a simulator. The ChatGPT answer of doing a 1-bit full adder as "hello world" makes sense.
You are obviously not going to etch silicon at home, but the design part is rather accessible as far as hardware goes.
You can absolutely etch silicon at home. Processes like wet etching (KOH, HF), reactive ion etching (RIE), laser ablation, and even electron beam lithography using repurposed CRTs are all viable at the DIY scale.
They're not used in high-volume manufacturing (you’re not replacing ASML), but they’re solid for prototyping, research, and niche builds.
Just don’t underestimate the safety aspect—some of these chemicals (like HF) are genuinely nasty, and DIY high voltage setups can bite hard.
You're not hitting nanometer nodes, but for MEMS, sensors, and basic ICs, it’s totally within reach if you know what you’re doing.
It's actually pretty scary how far this guy got: 10nm. Makes you wonder if something as powerful as an intel N100, famously a current 10nm chip + DDR4 memory, could be made like this. That would support quite a bit in terms of AI already, even if transformers are probably a bit much too ask.
> rise of LLMs lowering the barrier of entry of software even lower
Getting to your first wafer costs something like $250k and upwards of fab costs, depending on what process you're using. Hence much of chip design effort is already spent on verification, it's probably over 50% by now. This is the exact opposite of vibes because mistakes are expensive.
Businesswise it's quite tough B2B sales because you're selling into other people's product development pipelines. They need to trust you because you can sink their project, way over and above the cost of the actual parts.
Edit: I cannot emphasise enough how much more conservative the culture is in chip design and EE more broadly. It belongs to a world not just before "vibe coding" but before "web 2.0". It's full of weird closed source very expensive tooling, and is built on a graveyard of expensive mistakes. You've got to get the product 100% right on the first go.
Well, maybe the second go, production silicon is usually "B" rev. But that's it. Economics dictate you then need to be able to sell that run for a few years before replacing it with an upgraded product line.
The rule of thumb I use for chip design is that verification takes at least 2/3s of development. Sometimes more. 50% would be nice but I think is optimistic
Verification is indeed the majority of the time spent. Unlike programming, Verilog and VHDL and higher level things like Chisel aren’t executed serially by the hardware they describe like a von Neumann machine. Hello World for a chip isn’t designing the circuit, or simulating the circuit, or synthesizing the circuit to some set of physical primitives. No, it’s proving that the circuit will behave correctly under a bunch of different conditions. The less commoditized the product, the more important it is to know the real PDK, the real standard cell performance, what to really trust from the foundry, etc. Most of the algorithms to assist in this process are proprietary and locked behind NDAs. The open source tools are decades behind the commercial ones in both speed and correctness, despite heavy investment from companies like Google.
And so my point: the place where people best know how to make chips competitively in a cutthroat industry is NOT in schools, but in private companies that have signed all the NDAs. The information is literally locked away, unable to diffuse into the open where universities efficiently operate. Professors cannot teach what they don’t know or cannot legally share.
Chip design is a journeyman industry. Building fault-tolerant, fast, power-efficient, correct, debuggable, and manufacturable designs is table stakes. Because if not, there are already a ton of chip varieties available. Don’t reinvent the wheel because the intersection of logic, supply chain logistics, circuit design, large scale multi objective optimization, chemistry, physics, materials science, and mathematical verification is unforgiving.
I think you would just buy a cheap FPGA board and use that wouldn't you? No need to do a full chip until you know what you are doing. That would be like building a server farm just to do your software hello world
In computer science courses, that's as simple as a println().
In machine learning courses, that's training on mnist dataset to do character recognition.
In electrical engineering, that's buying a raspberry pi to blink led.
In chip design ... Chatgpt says to design a 1-bit full adder using verilog?
...
I understand why the article thinks the market is looking for graduate education. To design a simple chip requires an *initial investment* (as with all hardware startups really). This is different from software where one can simply launch a web app with a container hosted on your preferred cloud provider...
... That said, with the rise of LLMs lowering the barrier of entry of software even lower (e.g. vibe coding), may we see more rise of hardware startups/innovations?