There's a reason why entrepreneurs don't often stick around long once their startups are bought by a big technology player. The ability to make quick decisions, take risks and other cherished freedoms are often out the window once a small company is absorbed.

The notion of the incubated company-within-a-company was developed to try and reinstate these entrepreneurial mores. The world's biggest networking gearmaker, Cisco Systems Inc., for example, operates a handful of adjunct networking businesses it calls "emerging technologies," including Internet protocol-based security camera systems and digital signage, with the eventual aim of growing them into billion-dollar businesses. Some, such as Cisco's video-conferencing business, have graduated into big moneymakers.

When they're successful, these businesses often get spun off to the public, or sold, or absorbed in a buy-in. But in some rare cases, an entrepreneurial project within a company is such a big winner that it provides the parent with a whole new business focus.

Such is the case with Cypress Semiconductor Corp., a Silicon Valley chipmaker founded in 1982 that traditionally focused on memory, interface, data connection and timing chips. The company has a long history of incubating new businesses, and few have woven the incubator model into their corporate fabric like Cypress. One of its most successful is SunPower Corp., a maker of highly efficient solar-cell technology that was spun out and went public in 2005. The company, staring down bankruptcy when it joined Cypress early last decade, now generates $1.8 billion in annual revenue.

But perhaps more important to Cypress is the success of Cypress MicroSystems, an "internal startup" founded in 1999. The initial aim of the unit was to steal business from rival Microchip Technology Inc., a leader in microcontrollers, which are essentially tiny computers on a single chip that provide digital control to power tools, toys, car engine control systems and a host of other devices.

To do this quickly, the unit developed programmable systems on chips, which are highly customizable components that incorporate interchangeable pieces to suit customers' needs.

The bid to wrest business from a rival wasn't particularly successful at first, as Cypress MicroSystems had to offer its products for prices that were too low. But after absorbing $54 million in cash before reaching profitability, the project, now dubbed "programmable systems on a chip" or PSoC, ended up producing what is now Cypress' flagship product, adding some $150 million to the company's revenue line, significantly more attractive gross margins and opening up a much larger potential market.

The key was the flexibility to create custom designs with the "Lego-like" building blocks, as Cypress chief financial officer Brad Buss puts it. "It could probably be a half-billion company on its own," he says. "Five years ago we were 70% memory, but now we're 70% programmable. Cypress Micro changed the game for us, and it's the heart of the new company."

But, just as it is with venture-backed startups, Cypress has had its share of failures with its internal experiments. Buss delicately refers to these as "turds." They include Silicon Magnetic Systems, which attempted to develop so-called magnetic random access memory, or MRAM, chips. The project failed to live up to the hype that MRAM would supplant other traditional memory technologies, and was shut down after absorbing $50 million.

Another was Ross Technology Inc., which was formed to compete against Intel Corp. As the pull-no-punches CEO of Cypress T.J. Rodgers wrote once, the startup's "culture of corporate dishonesty and poor quality did them in."

To date, Cypress has funded ten internal startups and acquired two (Silicon Light Machines, an optical chipmaker purchased in 2000 for $166 million and considered a flop, and SunPower).

Right now, the company's handful of emerging-technology startups include Cypress Envirosystems, which develops specialized thermostat systems for energy-efficient buildings; AgigA Tech, a memory chipmaker acquired as part of Cypress' 2008 purchase of Simtek Corp. for $46 million; a unit focused on selling end products in China; and its optical navigation system business, which encompasses a sensor used in a computer mouse that evolved from Silicon Light Machines technology.

In the fourth quarter, these businesses contributed about $4 million to Cypress' $227 million in total revenue for the period. This may seem to be a small drop in the bucket, but the potential for another company-changing success is always there, Buss says.

"I market this as a call option for the future," he says. "There's always some failure in there, but you only need a couple of winners."

See the complete archive of Silicon Valley Special