feature article
Subscribe Now

Lost and Foundry

Chip making, like America’s Cup yachting, is a rich man’s game. To actually make the physical silicon chips—that is, to run a semiconductor fab—costs many billions of dollars. We’re talking NASA space program, government bailout, gold-plated washroom fixtures kind of money.

The cost to build a brand new chip-making plant is around $5 billion, and that’s just for starters. Then there’s the raw material, labor, taxes, R&D, waste-disposal, and many other expenses.

Oh, and the whole enterprise will be obsolete in less than five years, so you’ve got about 1300 business days to make back your $5 billion investment. That’s $3.8 million a day, every day, just to break even. It makes the cost of heroin addiction seem like a faint craving for salty snacks.

It’s no wonder that very few companies—or even countries—can afford to do it. And yet the chip business is thriving. We’re clearly not hurting for DRAMs, microprocessors, and A/D converters. They’re not even getting more expensive. How does that work?

The answer, Dear Reader, lies with third-party foundries. These are independent factories-for-hire that make batches of chips for anyone who can write a check. The big foundry players are TSMC (Taiwan Semiconductor Manufacturing Company), GlobalFoundries (formerly Chartered Semiconductor), UMC (United Microelectronics Corporation), and SMIC (Semiconductor Manufacturing International Corporation), plus about a dozen others scattered around the world. Your average foundry makes chips for 20–50 customers at once. They even serve competitors, making chips for Company A alongside chips for rival Company B. Foundries are the mercenaries of the chip-making world.

A Heartbreaking Work of Staggering Expense

Because of the rise of the foundries, very few “chip companies” actually make chips any more. IBM and Intel are the last big holdouts in the United States. Toshiba in Japan and Samsung in Korea make most of their own chips. But that’s about it. Big names like AMD, Freescale, and Infineon all stopped producing their own chips months ago. Texas Instruments and Renesas will outsource their leading-edge manufacturing soon. Biggish companies like Altera, nVidia, and Xilinx never made chips at all; they’ve always outsourced their silicon manufacturing.

The net result is that more and more chips are being made by fewer and fewer companies. And most of those companies are probably located halfway around the world from you. Is this a cause for concern or celebration?

On the one hand, outsourced silicon manufacturing makes good business sense and is a natural evolution of the semiconductor industry. There’s no real advantage to having a thousand different chip companies all cooking up their own chips in their own kitchens. That would be a wasteful duplication of effort. Besides, the question is moot; most chip companies can’t afford their own production facility, so it’s either outsource or die.

On the other hand, centralized manufacturing might reduce differentiation. If all of the world’s chips are being made in just a handful of foundries, aren’t we heading toward a silicon monoculture that lacks variety and diversity? The answer there is no. Not all foundries make their customers’ chips the same way—that’s how they compete. Some foundries emphasize low-power silicon processes, while some push high performance. Natural competitiveness keeps them from cloning one another. 

On the third hand, the rise of the foundry business has enabled all sorts of startup companies that could never have existed otherwise. With a $5 billion ante, the semiconductor poker game wouldn’t have very many players. But now, two guys and a dog can start a chip company with little more than a bright idea, some EDA software, and a few million in venture capital. That wouldn’t happen without independent foundries willing to make small batches of chips.

Finally, the global nature of foundries makes some customers nervous. Three of the four biggest foundries are located in Taiwan or China. What if an earthquake or other natural disaster disrupts their supply? Or military action stops shipments? Or a trade embargo inflicts sky-high tariffs? Similar things happened in the 18th and 19th centuries; who’s to say it won’t happen again tomorrow?

GlobalFoundries is the exception. True to its name, it has facilities in Singapore, Germany, and the United States, so it can serve xenophobic or politically sensitive customers around the world. Its next fab will be in Abu Dhabi, giving the company a big presence in the Middle East and making it a high-tech Mecca, so to speak.

So the “chip making” business has really become the “chip design and marketing” business. The design work is done by white-collar workers in nice, clean office buildings, while the manufacturing work is done by white-suited workers in very clean office buildings. It’s a nice symbiotic relationship.

Foundry production doesn’t limit innovation or differentiation; it actually encourages it. Would your chip design be different—in any significant way—if you manufactured it yourself using some unusual process? Unless you’re working on a scientific research project, chances are it wouldn’t. But would your chip design get manufactured at all if you weren’t using a third-party foundry? Unless you’re working at IBM, Intel, or Toshiba, chances are it wouldn’t. 

Automakers don’t make their own tires, refine their own oil, or develop their own electronics. Boeing buys its engines from Rolls-Royce and its airline seats from independent suppliers. That’s a normal supply chain. Economics 101 teaches us that specialization is the cornerstone of civilization and economic growth. There’s no reason chip companies should be any different.

So let us not mourn the decline of the vertically integrated semiconductor manufacturer. Let us instead praise the rise of the foundry and the innumerable little fabless startup companies they have enabled. Hail, horizontal integration! Hail, innovation! Hail, the submicron millennium!

Leave a Reply

featured blogs
Jan 19, 2021
If you know someone who has a birthday or anniversary or some other occasion coming up, you may consider presenting their present in a Prank-O gift box....
Jan 19, 2021
As promised, we'€™re back with some more of the big improvements that are part of the QIR2 update release of 17.4 (HotFix 013). This time, everything is specific to our Allegro ® Package Designer... [[ Click on the title to access the full blog on the Cadence Communit...
Jan 19, 2021
I'€™ve been reading year-end and upcoming year lists about the future trends affecting technology and electronics. Topics run the gamut from expanding technologies like 5G, AI, electric vehicles, and various realities (XR, VR, MR), to external pressures like increased gover...
Jan 14, 2021
Learn how electronic design automation (EDA) tools & silicon-proven IP enable today's most influential smart tech, including ADAS, 5G, IoT, and Cloud services. The post 5 Key Innovations that Are Making Everything Smarter appeared first on From Silicon To Software....

featured paper

Overcoming Signal Integrity Challenges of 112G Connections on PCB

Sponsored by Cadence Design Systems

One big challenge with 112G SerDes is handling signal integrity (SI) issues. By the time the signal winds its way from the transmitter on one chip to packages, across traces on PCBs, through connectors or cables, and arrives at the receiver, the signal is very distorted, making it a challenge to recover the clock and data-bits of the information being transferred. Learn how to handle SI issues and ensure that data is faithfully transmitted with a very low bit error rate (BER).

Click here to download the whitepaper

featured chalk talk

Time Sensitive Networking for Industrial Automation

Sponsored by Mouser Electronics and Intel

In control applications with strict deterministic requirements, such as those found in automotive and industrial domains, Time Sensitive Networking offers a way to send time-critical traffic over a standard Ethernet infrastructure. This enables the convergence of all traffic classes and multiple applications in one network. In this episode of Chalk Talk, Amelia Dalton chats with Josh Levine of Intel and Patrick Loschmidt of TTTech about standards, specifications, and capabilities of time-sensitive networking (TSN).

Click here for more information about Intel Cyclone® V FPGAs