Foundry model

Companies that both designed and produced the devices were originally responsible for manufacturing microelectronic devices. These manufacturers were involved in both the research and development of manufacturing processes and the research and development of microcircuit design.

The first pure play semiconductor company is the Taiwan Semiconductor Manufacturing Corporation, a spin-off of the government Industrial Technology Research Institute, which split its design and fabrication divisions in 1987,^[5] a model advocated for by Carver Mead in the U.S., but deemed too costly to pursue. The separation of design and fabrication became known as the foundry model, with fabless manufacturing outsourcing to semiconductor foundries.^[6]

Fabless semiconductor companies do not have any semiconductor fabrication capability, instead contracting with a merchant foundry for fabrication. The fabless company concentrates on the research and development of an IC-product; the foundry concentrates on manufacturing and testing the physical product. If the foundry does not have any semiconductor design capability, it is a pure-play semiconductor foundry.

An absolute separation into fabless and foundry companies is not necessary. Many companies continue to exist that perform both operations and benefit from the close coupling of their skills. Some companies manufacture some of their own designs and contract out to have others manufactured or designed, in cases where they see value or seek special skills. The foundry model is a business model that seeks to optimize productivity.

The very first merchant foundries were part of the MOSIS service. The MOSIS service gave limited production access to designers with limited means, such as students, university researchers, and engineers at small startups.^[7] The designer submitted designs, and these submissions were manufactured with the commercial company's extra capacity. Manufacturers could insert some wafers for a MOSIS design into a collection of their own wafers when a processing step was compatible with both operations. The commercial company (serving as foundry) was already running the process, so they were effectively being paid by MOSIS for something they were already doing. A factory with excess capacity during slow periods could also run MOSIS designs to avoid having expensive capital equipment stand idle.

Under-use of an expensive manufacturing plant could lead to the financial ruin of the owner, so selling surplus wafer capacity was a way to maximize the fab's use. Hence, economic factors created a climate where fab operators wanted to sell surplus wafer-manufacturing capacity and designers wanted to purchase manufacturing capacity rather than try to build it.

Although MOSIS opened the doors to some fabless customers, earning additional revenue for the foundry and providing inexpensive service to the customer, running a business around MOSIS production was difficult. The merchant foundries sold wafer capacity on a surplus basis, as a secondary business activity. Services to the customers were secondary to the commercial business, with little guarantee of support. The choice of merchant dictated the design, development flow, and available techniques to the fabless customer. Merchant foundries might require proprietary and non-portable preparation steps. Foundries concerned with protecting what they considered trade secrets of their methodologies might only be willing to release data to designers after an onerous nondisclosure procedure.

In 1987, the world's first dedicated merchant foundry opened its doors: Taiwan Semiconductor Manufacturing Company (TSMC).^[8] The distinction of 'dedicated' is in reference to the typical merchant foundry of the era, whose primary business activity was building and selling of its own IC-products. The dedicated foundry offers several key advantages to its customers: first, it does not sell finished IC-products into the supply channel; thus a dedicated foundry will never compete directly with its fabless customers (obviating a common concern of fabless companies). Second, the dedicated foundry can scale production capacity to a customer's needs, offering low-quantity shuttle services in addition to full-scale production lines. Finally, the dedicated foundry offers a "COT-flow" (customer owned tooling) based on industry-standard EDA systems, whereas many IDM merchants required its customers to use proprietary (non-portable) development tools. The COT advantage gave the customer complete control over the design process, from concept to final design.

• Pure-play semiconductor foundry is a company that does not offer a significant amount of IC products of its own design, but instead operates semiconductor fabrication plants focused on producing ICs for other companies.^[9]
• Integrated device manufacturer (IDM) semiconductor foundry is where companies such as Texas Instruments, IBM, and Samsung join in to provide foundry services as long as there is no conflict of interest between relevant parties.

As of 2009, the top 17 semiconductor foundries were:^[17]

(1) Now acquired by GlobalFoundries

As of 2008, the top 18 pure-play semiconductor foundries were:^[18]

(1) Merged with CR Logic in 2008, reclassified as an IDM foundry

As of 2007, the top 14 semiconductor foundries include:^[19]

For ranking in worldwide:^[20]

As of 2004, the top 10 pure-play semiconductor foundries were: ^{[citation needed]}

Like all industries, the semiconductor industry faces upcoming challenges and obstacles.

The cost to stay on the leading edge has steadily increased with each generation of chips. The financial strain is being felt by both large merchant foundries and their fabless customers. The cost of a new foundry exceeds $1 billion. These costs must be passed on to customers. Many merchant foundries have entered into joint ventures with their competitors in an effort to split research and design expenditures and fab-maintenance expenses.

Chip design companies sometimes avoid other companies' patents simply by purchasing the products from a licensed foundry with broad cross-license agreements with the patent owner.^[21]

Stolen design data is also a concern; data is rarely directly copied, because blatant copies are easily identified by distinctive features in the chip,^[22] placed there either for this purpose or as a byproduct of the design process. However, the data including any procedure, process system, method of operation or concept may be sold to a competitor, who may save months or years of tedious reverse engineering.

• Compound Semiconductor.net: "Foundry model could be key to InP industry future"