Why is so expensive to fabricate custom IC

[ARISTOS8898]

I had read that the fabricate a custom IC is very expensive (>10000$) but why?

For example an 555 timer cost <1$.
If I design and 555 timer with an inverter at the output in some program (don't include that the license is also expensive) and then I want to fabricate that why will cost so much? What the 555 timer don't have but my IC have which makes the price >10000 times more expensive?

 

[(*steve*)]

Why can you buy a mass produced car for some tens of thousands of dollars, yet manufacturers invest a billion (if not more) for a completely new model?

"The price tag to develop a new vehicle starts around $1 billion. According to John Wolkonowicz, Senior Auto Analyst for North America at IHS Global, "It can be as much as $6 billion if it's an all-new car on all-new platform with an all-new engine and an all-new transmission and nothing carrying over from the old model."

The reason for cars is pretty much the same reason as it is for semiconductors.

 

[hevans1944]

The reason for cars is pretty much the same reason as it is for semiconductors.

For the layman, it's called "economy of scale" as pioneered by Henry Ford in the early part of the 20th Century. It always costs more to make the first of anything. After that, if you can invoke mass production methods of manufacture, the cost per unit manufactured plummets as the fixed costs are distributed across thousands or millions or billions of products.

Cell phones are an excellent example. Because the world wide market is so large, several cell phones of different designs and capabilities can be manufactured by more than one entity and compete in the open market successfully, How big a market your "enhanced" 555 would have depends on how many of these "new" products you will sell. The larger that number, the less the initial cost per unit will be. The initial "startup" cost of about $10,000 when spread over a million identical products (which is "low volume" for a product distributed world wide) becomes insignificant. For what it's worth, the start-up cost for a new semiconductor product is probably lower than just about anything manufactured on a large scale today.

I have heard it said that manufacturing semiconductors is a lot like printing a newspaper or magazine in terms of the cost to do so. Substitute a few tons of sand (silicon) for paper and few pounds of exotic chemicals for ink. If you pump out a few billion copies the cost of materials vanishes into the other fixed costs of marketing and distribution.

 

[BobK]

In addition to the high cost of setting up for production, IC manufacturing has another attribute. ICs are fabricated on 12 inch wafers. That is 113 square inches. A high end microprocessor chip is about a 1/2 inch square, or 1/4 of a of square inch. So, the minimum number you can make is about 400 of them.

For a chip the size of a 555 is is much worse. It is probably not even 1/8 inch square. So that would make the minimum quantity about 7000 of them.

That would be one wafer. It would be ridiculously expensive to make only 1 wafer, as you would have to do all the setup, make one, then go on to the next item.

Then you have to cut them all up, test them, and put them in a package connecting the leads.

And the price of a new fab is billions of dollars. That also has to be paid for over it's lifetime.

For low quantity ICs, we have programmable logic arrays. These have hundreds to many thousands of basic logic circuits on a chips where you can subsequently program the connections between them. They also have

Bob

 

[hevans1944]

You would think, with U.S. Government support, i.e., a research and development contract, it would be easy to develop a new semiconductor device.

In my former job at UES, Inc. we did just that at the request of the Defense Threat Reduction Agency (the military subsidiary of Homeland Security). It took several years of research and development, minimally and incrementally funded by the U.S. Government, but we succeeded in making a practical Photo-Conductive Semiconductor Switch (PCSS):

This was accomplished by adding some novel process steps to a PCSS invented earlier by engineers and scientists at Sandia National Laboratories (SNL), Albuquerque, NM during the 1990s. Those brilliant folks did their PCSS development work mainly by using discretionary "in-house" funding for almost a decade. They made considerable progress and published their results in the open literature, but eventually they were cut off and told to "get back to work" doing the classified nuclear weapons business for which SNL was created. With the time and money they were allotted, they couldn't find a practical solution to a serious problem, so their technology lay dormant for more than two decades.

It's a rather long, and maybe even a possibly interesting and entertaining story, that I may write up and post as an attachment to a new thread here in the Forums if anyone appears to be interested. I worked one summer as a contractor, employed by the University of Dayton Research Institute (UDRI) for the Air Force Weapons Laboratory in Albuquerque, NM, helping to establish their Optical Components Evaluation Laboratory (OCEL). SNL is right next door, but I didn't have a clue (much less the Q clearances) to know what, exactly, they did. We were kept busy in the Weapons Lab trying to develop airborne laser weapons as part of Ronald Regan's Star Wars anti-ballistic missile defense program. The last time I saw the Airborne Laser Laboratory (ALL) aircraft, it was parked outside at the U.S. Air Force Museum at Wright-Patterson AFB, Dayton, OH. Someone had removed all the optical "goodies" by the time it got there. It is my understanding that the Air Force later rigged up another airplane with a "bigger and better" laser, but so far the program has gone nowhere and produced zilch.

Our research, OTOH, was successful. Prototype devices were fabricated, delivered and tested, and a patent in my supervisor's name and my name (as co-inventor) was applied for. Our prototype won an R&D 100 award and national recognition in 2015, but the USPTO has now twice denied our patent application, first citing prior work by SNL, who used optics to create parallel illumination lines from a laser light-source to trigger the PCSS. This was true but totally unlike anything we did and claimed. Later, the USPTO cited a totally unrelated ion implant procedure they claimed also took precedence over our claims.

My former employer does not have deep pockets of cash to spend on attorneys to pursue the patent. It relies on government contracts to support its business model, much like Northrup Grumman or Lockheed Martin, but on a much smaller scale. Nor does the government care if we were issued a patent or not. The government paid for the R&D and the prototype PCSS devices. It therefore has an interest in the technology we developed at taxpayer expense, whether patented or not. It would be nice if there were commercial applications lined up waiting for this, but we could find no one after more than a year of searching. Maybe I was looking in the wrong places, but in the end I lost my job and was "forced" into retirement because we had a "solution" but apparently there was no one with a "problem" that needed it. Kinda reminds me of the situation during the early years of laser development.

The story hasn't ended. Patent or no patent, UES, Inc. knows how to successfully make these PCSS devices. Recently the company was awarded a small contract that requires a few more of them to be made. It's not something you can order online from DigiKey or Mouser yet, but if anyone really needs a high-voltage (kilovolts, scale-able to megavolts), high-current (kiloamperes), really fast (picosecond turn-on), high optical gain (micro-joules to trigger) pulsed switch... or a few thousand of 'em... I am sure something can be worked out.

So that's just an example of one possible deterrent on your path to sprucing up the venerable 555 integrated circuit. Starting with something that almost works the way you want, like the 555 or your late twentieth century PCSS, and improving it is not enough. You need lots of money, meaning investors, who will hire lawyers and subcontract out masking and manufacturing to independent semiconductor fabrication plants (there are several, worldwide). Then you need more money for marketing, since there are competitive alternatives to what you proposed. It's not like buying an empty corn field, building a ball park there, and believing they will come, waving dollar bills in your face to purchase your "new and improved" 555 timer. And investors like to have "patent protection" for their investment money.