Have you ever wondered why dozens of companies are designing chips, as opposed to just a handful of giant companies with fabs? It's because the fabless semiconductor revolution has reshaped not only the semiconductor industry but the device-rich world we live in. A new book explains how all that happened and preserves an essential piece of electronics industry history.
The book is titled Fabless: The Transformation of the Semiconductor Industry. It is a project of Semiwiki.com, which has become a go-to news and blogging site for IC design and manufacturing. The authors are Semiwiki bloggers Daniel Nenni and Paul McLellan, and a number of key companies (including Cadence) contributed chapters to the book. The book is now available in Kindle (mobi) and iBooks (ePub) formats and is going to print in March.
You can tell a lot about a book from the opening paragraphs in the preface, so let's go there first:
"The purpose of this book is to illustrate the magnificence of the fabless semiconductor ecosystem, and to give credit where credit is due.
We trace the history of the semiconductor industry from both a technical and business perspective. We argue that the development of the fabless business model was a key enabler of the growth in semiconductors since the mid-1980s. Because business models, as much as the technology, are what keep us thrilled with new gadgets year after year, we focus on the evolution of the electronics business."
The book defines a "fabless" semiconductor company as "a company that designs their own chip but outsources the manufacturing to a third party, either a pure-play foundry or an IDM [integrated device manufacturer} that sells excess fab capacity." It notes that "this is the prevailing business model today."
It wasn't always so. In the early 1980s chips were designed in-house by large IDMs who also had fabs - Intel, IBM, AMD, and a few others. They mostly used internally-developed CAD tools. Then several things changed. First, fabs got really expensive (think $10 billion to start one today). Secondly, systems companies wanted more customization than off-the-shelf chips could provide. Third, an independent commercial EDA business emerged.
It's thus not surprising that the ASIC business model emerged in the 1980s. Here, systems companies did the front-end (logical) design and handed it over to an ASIC provider who would do the back-end (physical) design and manufacture (or pay somebody else to manufacture) the chip. This eventually evolved into a true "fabless" model, where a semiconductor company will do the complete design and a pure-play foundry will manufacture the chip.
Here's a brief overview of the chapters in the book:
Chapter 1: The Semiconductor Century. This chapter traces the invention of the transistor and the integrated circuit, the beginnings of Silicon Valley, and the continuing reality of Moore's Law. It looks at the emergence of business models from IDM to ASIC to fabless.
Chapter 2: The ASIC Business. This chapter presents a history of the ASIC business and shows how it dramatically changed the semiconductor industry. It includes "In Their Own Words" descriptions of VLSI Technology and eSilicon.
Chapter 3: The FPGA. Wondering what this chapter is doing here? It may not be widely known that FPGAs fueled the fabless business model, or that Xilinx and UMC pioneered the "virtual IDM" relationship. The chapter includes a history of Xilinx.
Chapter 4: Moving to the Fabless Model. This chapter goes into more detail about how and why the fabless model evolved. It includes an "In Their Own Words" history of Chips and Technologies, which the book describes as the first fabless semiconductor company (it was later acquired by Intel).
Chapter 5: The Rise of the Foundry. Pure-play foundries (manufacturing only, no design) emerged in the mid-1980s. This chapter shows why they came about and how they helped build the fabless semiconductor ecosystem.
Chapter 6: Electronic Design Automation. EDA has been a key enabler of the fabless revolution, and this chapter traces EDA history by identifying five "phases." These phases include:
- Phase one - Formation of Calma, Applicon, and ComputerVision in the 1970s to automate IC layout.
- Phase two - Publication of Introduction to VLSI Systems by Carver Mead and Lynn Conway; creation of ASICs; "big three" EDA vendors in the 1980s are Daisy, Mentor Graphics, and Valid.
- Phase three - System companies start doing their own back-end design. The book says: "The most important company of the third phase of the EDA industry was Cadence Design Systems."
- Phase four - Transition from schematics to synthesis-based design for digital ICs.
- Phase five - Era of "full service" EDA companies who attempt to span the entire design flow.
This chapter includes "In Their Own Words" histories of Mentor, Cadence, and Synopsys. (I worked on the Cadence history description).
Chapter 7: Intellectual Property. This chapter shows how the semiconductor intellectual property (IP) business arose, starting with standard cell libraries and reaching up to processors. It includes a history of ARM.
Chapter 8: What's Next for the Semiconductor Industry? A number of industry luminaries contributed short viewpoints to this concluding chapter, including CEOs of ARM, Xilinx, Synopsys, Cadence, Mentor, Atrenta, eSilicon, Jasper, and others. Semiwiki bloggers contributed viewpoints as well - and there's even one from yours truly.
So in conclusion, this book is a good read - dare I say a "must read" -- for anyone involved in, or who wants to understand, the semiconductor or EDA industries. I've been writing about EDA since 1985 and I still learned a lot from it. You will, too.
The book is priced at $15. For a digital download of the book, click here.