[miohtama]

The problem with hardware business is that it’s capital intensive: large, expensive product runs needed. Not many are happy to upfront the bill for a new, unproven, company. Whileas in software business, the upfront cost is fractions, and thus VCs love it more.

Thus, in hardware, funding comes from the existing players who already know the hardware business with its cyclic business and other associated risks.

[throwup238]

I don't think it's a problem with how much capital hardware requires. Biotech startups require an order of magnitude more capital with tons more risk with regulatory approval yet the biotech VC industry is huge - the amount invested annually frequently exceeds all of tech VC. Most of the startups that go public do so with zero revenue, let alone profit, and everyone is perfectly fine to bankroll them through clinical trials even though they're all or nothing.

I think there's two problems with hardware: first the marginal profit per unit doesn't scale so to make more profit you have to sell more widgets. The same is mostly true for biotech but the profit margins on a drug are usually >95%, with a much higher ceiling, and are heavily recurring, often for the life of a patient. Since biotech customers are mostly insurance companies, the value of a drug is easy to calculate based on its quality of life improvements and past deals.

Second, success is very all or nothing for hardware companies. Each hardware startup will have a limited number of possible acquirers who specialize in their field so they either become profitable and go public or fail. On the other hand, failed tech companies get acquihired by the tech giants and pharmaceutical companies acquire tons of companies before they even finish clinical trials. Any startup that makes it past phase 3 trials has a 99% chance of getting acquired by a pharmaceutical company so the economics of investing are very enticing, despite the massive capital outlays.

[dartos]

This 100%.

One (A SINGLE) photolithography machine is like $4B.

There’s no way for a new player to enter the CPU market

[bittwiddle]

This is why most "semiconductor" companies are actually fabless, focusing on design while working with a partner(like TSMC) to have their designs manufactured.

[mlyle]

I remember hanging out with Carl Amdahl after we were pitched our nth fabless semiconductor company, and him making the joke that if he ever started another company making IC designs, this time he would call it Fabulous Semiconductor.

[Scoundreller]

And even fabbed semiconductor companies de-fab themselves, like AMD.

https://en.m.wikipedia.org/wiki/GlobalFoundries

[jstanley]

You wouldn't normally be physically manufacturing CPUs yourself.

Most likely you'd be supplying an "IP core", which is basically just software. More niche than that, if you actually wanted to supply physical chips, you would contract out the manufacturing to TSMC for example.