Now "an idea for something great in CPU design" is unlikely to result in a marketable chip these days, but chip startups exist, and some are very small. Adapteva for instance is a 5-people company.
I was pondering the viability of almost exactly this. Could one invest in a hardware pipeline that efficiently combines e.g. 3D printing with FPGAs to create the consumer-electronics equivalent of Lulu.com's just-in-time small-batch book printing + drop-shipping.
Or, in other words: you can do pretty much anything with the sensors in a modern smartphone. All most manufacturers probably need is smartphone boards (with maybe some options of extra sensors) in custom cases with fancy buttons and displays, maybe custom remotes, and a cute little 4-colour double-walled box. That describes everything from Nest to Roku to some drone controllers.
JIT small-batch assembly seems roughly analogous to other low-volume electronics, which are obscenely expensive. For example, specialty bench equipment like spectrometers can cost as much as a house.. There are many factors at play there, but you can bet one of them is very low volumes.
This is why, at least so far, the kind of play you describe is accomplished with large batches of a generalizable platform that can be sold into many different devices.
Happy to make some free parts for anyone's ideas on here!
Don't forget all the hardware companies you know today were once startups. Yes there's a current fashion to call companies that are basically small businesses "startups", but the term encompasses a far more profound class of enterprise.
And Apple itself was founded by an engineer (Woz) who first got his start at a big company, HP, which is celebrated as a hardware startup that grew.
I think the article and your comment do get to an important point, which is that Apple is big enough to have solid (and at the moment very successful) processes and infrastructure so that you only get experience in that matrix, and don't have to spend any time learning stuff outside your own area. You can, of course, and plenty of companies have come from Apple alums.
I also think its right to compare now with the 1980s when F500 companies didn't know how to compete in the new tech hardware space. Companies like Samsung rush into new product spaces very early in the adoption cycle.
Unfortunately the winds of VC have changed a lot since then. Originally, VCs were willing and eager to invest millions in a promising innovation. Now they expect you to demonstrate traction before any large investments—which is infinitely harder with hardware startups.
Wrong. Gates and Allen used Harvard's University mainframe to do their first gig, then paid for other mainframe time:
All the students wich are entitled to access to the mainframe worked typically on a single mainframe which was also used for the university accounting etc.
Mainframes were, using the more modern words, the big servers with the capital S, effectively.
Hobbyist keyboard alone (without the TV) had the price of nearly 1000 US dollars equivalent today in 1975
http://en.wikipedia.org/wiki/TV_Typewriter
The cheapest self containing computer that year had a minimal cost of 30,000 of today's US dollars.
http://en.wikipedia.org/wiki/IBM_5100
The "operating system" of it were Basic and APL, and it had the storage of 200 KB.
Now compare all this to the "server" which started this discussion. The claim was that when Gates started he didn't need "servers."
But were they selling Altair_BASIC running on said mainframe? No they were not, hence they did not have servers. They used a mainframe as a dev box.
You can't redefine my terminology just to suit your own viewpoint.
Hardware startups are still being funded. Software startups are still being funded. Life science startups are still being funded. And it's a lot of the same old line VCs who are doing these deals. I'm thinking $15M-$20M first round, with companies need significant funding but get quite large.
Then there are a ton of small companies, many of which are really small businesses (just look at the list of companies at https://news.ycombinator.com/item?id=9666013), that are getting small amounts of funding ("spray and pray" funds). Are their funders really "VCs" in the classic sense when the amount they invest is =< $100K and they can't participate in future rounds? Although they get a most of the press for "startups" and "VCs", and they are the volume in absolute sense, most of these seem designed to run for a little while and then be aquihired away. So no need to build most of the infrastructure for a sustainable business. In that role, the "VCs" are really more like agents getting a commission on the aquihire.
It's the latter that aren't particularly picky. The traditional VCs seem mostly to still look for the same things.
To Amazon's credit, I can imagine Bezos reading your comment, frowning, and muttering, "Why the hell not?"
Could such an extreme outsourcing service (not necessarily only fabricating chips) paired with the accessibility of AWS be viable? Does such a company exist?
There is some truth to that but it's not the custom chip manufacturing that drives it. It's not even what I guess I would call 'excellence in material design' that drives it. Really it's a combination of the choice to not cut corners on hardware combined with the scale and market position to support the cost.
As an example a lot of has been made in the past about Apple's charges, both laptop and iDevices, they are small, sleek and powerful for their size. There isn't anything special going on there from an electronics perspective, they didn't invent new types of power conversion, they just didn't cut every possible corner to reduce cost and they had the budget to allow custom parts when off the shelf didn't fit. I don't mean custom IC's here I mean custom metal fab, maybe custom caps that are 'normal' other than being shaped a bit differently, etc. Apple has the volume and price point to let them do this when the vast majority of companies don't.
Apple does a lot of work in hardware and hardware manufacturing, both areas which are pretty capital-intensive and might not lead themselves as easily to the startup world.
Say, you are an Apple engineer working on the Ax chips for the next iPhone and have an idea for something great in CPU design, you cannot exactly rent a scalable 14nm chip fab from AWS to try to build it on your own and sell it market.