Startup Semiconductor companies

There is currently a general consensus among industry analysts, investors, and company executives that leading edge digital chip development projects costs $30-$100M to develop. This negative view of chip development costs has led to severely decimated levels of VC investments and has even led to certain companies scrapping whole products lines.[1] Today’s chips have a bigger impact on the world’s economy than ever and reduced investments in chip product development will eventually have a very negative impact on the economy. China for one has realized this and is pouring billions of dollars into their local semiconductor economy.[2] This article will go through some of the reported data and show that present assumptions regarding chip development costs are very much exaggerated and that it is possible to design state of the art chips at a fraction of the cost previously reported.

Why is Chip Development so Expensive?

Chip development has never been cheap and there is no question that it’s getting more expensive. The million dollar question is: “how much more expensive?” The consensus among industry pundits is that there is no way to do a leading edge chip with less than $30-$100M.[3-4] Let’s take a look at some of the most common arguments for the “exponential” increase in chip development costs.

“Mask costs are increasing exponentially!”

It used to be that you could design a product with production mask costs in the range of $100K. Today, production mask costs at leading nodes can cost $1-2M. This has the effect of increasing the minimum market size that is practical to address with a leading edge chip design. Still, given a decent size market or high chip selling prices, mask costs are certainly not show stoppers and don’t explain the $100M price tag.

“EDA tools are getting really expensive!”

While EDA tools are certainly not cheap, aggressive volume purchasing agreements, increased feature integration and increased efficiency in the tools have made today’s tools far more cost effective than those of ten years ago. Today many leading chips can be taped out using vanilla flows derived from reference scripts provided by the EDA vendor. Thus, per transistor and per project, the cost of EDA tools has actually gone down in the last ten years.

“Deep submicron design is really hard!”

Chip design is certainly getting more difficult as we move to finer process geometries. At 0.35um and above, life was simple. Gates were slow, wires were fast, and there was virtually no leakage to speak of. The only problem was that chips were slow, power hungry, and big! Since then, the design constraints and difficulties have been piling on with every process node. In order of introduction designers had to learn and deal with: wire delays, voltage drops, signal integrity, leakage, process engineering effects like stress, and most recently on chip variability and an explosion in the number and complexity of device design rules. The introduction of each one of these effects was painful, but within one process generation, the EDA industry always seemed to find automated methods of dealing with them, thus keeping engineering team growth to a minimum. Deep submicron chip design is definitely not for hobbyists, but for a small expert team equipped with state of the art tools, it is today possible to accomplish in weeks what used to take a generously sized team 6-9 months. [5]

“Chips are getting incredibly complex!”

Bingo! The simple reason for the explosive growth in chip design costs is that design complexities are growing exponentially thanks to Moore’s Law. At leading edge nodes it is now possible to integrate complete multi-billion transistor computer systems on a single chip. The more complicated the system, the more complicated the chip. Complicated SOCs are indeed very expensive to develop and there are few if any examples of complete products that have cost less than $10M from start to finish. An ASIC on the other hand tends to solve a very specific problem and be key part of a larger system. The limited generality of ASICs can lead to much higher performance levels and generally results in reduced complexity and development. For a great discussion of the relationship between project complexity, team sizes, and project costs, I highly recommend reading “The Mythical Man-Month”, by computer legend Fred Brooks. Unfortunately, there aren’t a lot of short cuts for saving costs when projects are complex and need to be completed in a short time. Note that complexity drives up development costs of any project, regardless of the underlying implementation fabric. As many designers are discovering, large FPGAs have all of the complexity problems of large SOC design, minus the long turnaround time and costs of manufacturing.

Source: www.adapteva.com

RELATED FACTS

Ahmet Tuncay has 18 years of corporate enterprise and startup
experience in business development, product development, product
management, technical marketing and sales support. Before Trapeze, he
worked for several technology companies, including Electronic...
Crocus Technology, founded in 2004, is a venture-capital-backed semiconductor startup company developing next generation magnetoresistive random-access memory (MRAM) technology. The company's products originated in a Grenoble-based Spintec laboratory. They company...
Lynne Greer Jolitz (born June 30, 1961) is a figure in free software and founded many startups in Silicon Valley with her husband William.
Lynne Jolitz is probably most famous for her work in pioneering open source operating systems with 386BSD with her husband. She...

Startup Semiconductor companies

RELATED FACTS

Share this Post

latest post

follow us

category