Not a Chip off the Old Block

Not a Chip off the Old Block

Texas Instruments is constructing an innovative new million-square-foot microchip fabrication plant in Texas, thanks to a little help from the Rocky Mountain Institute. By Christina Page

Paul Westbrook would love to show you his house. It's a passive-solar, award-winning structure in Fairview, Texas that uses a third of the electricity and less than a fourth of the water of conventional homes in the region. It's got superwindows, an aerobic septic treatment system, active solar water heating, a geothermal heat-pump heating system, and a pair of 1,600-gallon rainwater tanks. Its design details, favorable economics, and carefully measured performance are all posted online.

Achieving resource efficiency and sharing information about it are two of Westbrook's talents. When he isn't showing people energy-efficient design, systems, and devices in his home, he's working hard to bring those innovations to his company -- Texas Instruments, where he's the sustainable design manager.

When Westbrook encountered skepticism about the cost-effectiveness of switching from energy-hogging cathode-ray-tube computer screens to flat-screen displays, he wrote an extensive spreadsheet showing the company-wide benefits from energy and other savings. But he had loftier goals in mind. When Rocky Mountain Institute CEO Amory Lovins addressed a Sustainable Dallas event in 2001, Westbrook grabbed him to give a seminar at TI on efficient "fabs" -- microchip fabrication plants. In spring 2003, Amory returned to TI; the crowd was much larger and interest was growing. To raise awareness about the opportunities for sustainability and whole-system design, Westbrook brought three TI vice presidents to his house and showed them his utility bills (just as RMI has done at its superefficient headquarters). This got their attention -- and got them talking about implications for the company. "It helped demonstrate that applying good design could allow a much more efficient system with minimal capital investment," Westbrook pointed out.

A year and a half later, on 18 November 2004 in Richardson, Texas, TI broke ground on a state-of-the-art, high-efficiency, million-square-foot chip fab (including 220,000 square feet of clean room), designed in part with ideas generated during a three-day charrette1 with Rocky Mountain Institute.

RMI has worked for chipmakers before -- mainly consulting for another world leader, STMicroelectronics. During 1998–2000, RMI's experts surveyed eight ST fabs, finding large potential savings with fast paybacks. Sure enough, STMicroelectronics (like IBM and DuPont) has been cutting its energy use per unit of production by 6% every year, paying back in 2–3 years. But at TI, RMI had the prized opportunity to help design a new chip fab from scratch -- bound to save more and cost less than retrofitting old ones. Normally, in the boom-and-bust chip business, redesigning the next fab is either premature or too late; there's never a good time to do it. But Westbrook astutely timed his intervention, and engaged RMI just in time to change the design mentality, both in his firm and in its equipment providers and consulting engineers.

When Texas Instruments officials first began discussing their new facility, sustainability wasn't the foremost thing on their minds. Wafer fabs are complex, extremely capital-intensive (often several billion dollars), and highly energy-intensive. Numerous layers of submicroscopic chips, with features smaller than a flu virus, are etched, sputtered, and baked onto silicon wafers by exotic high-tech “tools” inside climate-controlled "clean rooms." Chip manufacturing is extremely sensitive to disruption and contamination. Reliability is crucial -- production stoppages can cost more than $1 million per day. Workers must be kept comfortable inside special smocks to keep dust and lint out of the ultra-clean air (the tiniest speck could ruin a chip whose features are less than a hundred-thousandth of an inch across). The tools' need for precisely controlled temperature and humidity can be even more critical.

To compete on cost, Texas Instruments was seriously considering building its new facility overseas. But of course fab's high cost is due not just to its specialized tools but also to the scale and complexity of the elaborate equipment that provides abundant chilled water, clean air, scrubbed exhaust, vacuum, and other "utilities." Using those services more frugally can make the fab cost less, work better, build faster, and win in the marketplace. -- -- In 2003, Chris Lotspeich, who'd led much of RMI's ST work, nicely summarized the challenges of chip fabs: --
  • Fabs have extensive heating, ventilating, and air conditioning (HVAC) systems with high-performance filters to maintain clean rooms' temperature and humidity precisely while filtering airborne particles. Fans, pumps, furnaces, and chillers deliver conditioned air and cooling water into the clean room via ducts and pipes. Depending on their size, fabs use anywhere from 3 to as much as 30 megawatts of power.2 HVAC systems consume 30%–50% of a fab's electricity; tools use another 40% or so. Energy accounts for [only a few]… percent of a chip's cost, yet electricity can be the largest single [non-labor] operating expense for a chipmaker, totaling millions of dollars annually at a single fab. Moreover, energy-saving measures can improve key operating parameters (yield, setup time, flexibility), and in new plants can save capital and construction time -- critical factors in competitiveness.
  • Despite great innovation, semiconductor manufacturing fosters a risk-averse corporate culture due to exacting process requirements, safety risks, the high cost of downtime, and brutal competition in a fast-moving market. Meeting production and time-to-market targets requires extraordinary control over thousands of variables. When something works, it is copied exactly. Firms also "copy exactly" previous fabs when building new ones. This saves some time and initial cost, yet retards improvements outside the clean room, including energy efficiency features -- thus raising operating costs. It's somewhat ironic that cutting-edge technologies are made in buildings designed decades ago, and thus those buildings now offer significant energy-and money-saving potential.
Additionally, a huge modern fab complex can easily go through 2–3 million gallons of water per day, a quarter of it for cooling.

Now came the hard part: TI's engineers and designers were told to cut the building and utilities cost by 30% over the previous project. "The cost challenge could have been a show-stopper," Westbrook said, "but turned out to be a benefit. We literally had to go back to the drawing board on many items. It gave us a chance to analyze old assumptions and challenge some conventional wisdom. RMI CEO Amory Lovins calls it 'good old Victorian engineering'" -- the art of wringing multiple benefits from single expenditures.

By driving revolutionary change and jettisoning incremental evolutionary design, the 30-percent-lower-capital-cost goal gave Westbrook and his unfunded "Fabscape" sustainability design team their opening to test the most innovative ideas. Starting in 2002, the team met every two weeks and generated a flurry of state-of-the-art concepts.

Their growing stack of white papers soon made a compelling case for a freewheeling-but-disciplined design process to distill out something useable. So in December 2003, a team of RMI consultants came to help TI bubble up and boil down hundreds of nifty notions into twelve "Big Honkin' Ideas" -- concepts that could fundamentally change how TI designed and built a fab and how TI worked with its industrial partners.

A wafer fab is full of exquisitely complex tools made by arcanely specialized suppliers. Energy efficiency is rarely a consideration when specifying tools. Because process and reliability requirements rule, the customer seldom asks for efficiency, and the toolmaker, who won't pay the utility bills, simply isn't used to providing it. It's not that they can't; rather, they've never been asked.

Yet the cumulative effect of all the power consumed by all the fab's tools and equipment led the charrette participants to trace how each watt of energy consumed by each tool ends up as heat that must be removed, making the cooling equipment bigger and power-hungry -- at a total present-valued cost around $7 per watt! So the biggest win wouldn't be simply making the cooling equipment more efficient, but making it smaller and simpler by buying efficient tools that would give off less heat in the first place. Equipment would be sized by measurement, not guesswork: as RMI designers say (borrowing from GM), "In God we trust; all others bring data."

Savings quickly started to breed and multiply. Nearly doubled-efficiency vacuum pumps, cut to idle speed when waiting for wafers, saved 300 tons of chiller capacity and 7% of the plant's total electricity. Vacuum-pump vendors, initially startled by requests for extra efficiency, soon saw the business logic. Optimizing temperature and pressure drops saved a fifth of internally cooled tools' cooling-water flow. Smarter exhaust systems saved 100,000 cubic feet per minute (cfm) of exhaust and its replacement (conditioned fresh air) -- each worth a present value of $62. Internally cooled tools with heat exchangers designed to lose less pressure and temperature cascaded into a 3,000-gallon-per-minute reduction in the size of the central process cooling water system, saving both capital and operating cost.

As post-workshop design progressed, it became increasingly apparent that smarter tools and their smaller, more efficient supporting systems would cascade energy and water savings. The results included a split chiller plant that cools water to two different temperatures for different purposes (further innovation might even eliminate one of the two sections in the next fab); highly efficient fan filter units for air recirculation; prechilling incoming hot air with outgoing cool air; big pipes and small pumps to cut friction and capital cost; natural daylighting and highly efficient lighting fixtures in the office area; solar water heating; a reflective roof; and extensive water recycling and reuse (reclamation will save nearly a million gallons of city water per day). Recovering heat previously thrown away, and using high-pressure water spray rather than steam for humidification, reduced six boilers to just one plus a backup -- both of which will be off most of the year -- cutting emissions of nitrogen oxides by 60%.

Although the facility will have to be up and running before anyone can know for sure, Westbrook predicts that the new facility will cut energy use by 20% and water use by 35% compared with TI's previous wafer fab. The savings come about half each from better tools and their direct support equipment and from smaller, more frugal utilities and building systems.

"Whole-systems tool design" wasn't the only breakthrough idea that emerged from the workshop. For example, TI decided to test lighter-weight smocks for clean room workers. Particle tests revealed that eliminating facemasks shouldn't harm product quality, and could make workers more comfortable with less chilling.

Some of the design features explored at the workshop were standard components of green design for non-industrial buildings, and offered tremendous financial benefit. Each waterless urinal, for example, will save 40,000 gallons of water a year, plus the capital cost of flush valves and water pipes not installed -- helpful to a water-intensive industry in an arid climate. Energy modeling software such as eQUEST3 let the designers test immediately how their ideas would change performance: for example, rotating the administration building 30° could save about $30,000 annually in space cooling.

The participants' diverse enthusiasms quickly focused on winning a high LEED (Leadership in Energy and Environmental Design) rating -- a systematic way, evolved with RMI's help, to score points for elements of good design. As Westbrook noted, "The competitive nature of people is a strong force and can be harnessed for good. We like to save energy and reduce emissions -- we love it when we score a point for doing so."

The LEED focus seems to be working well for TI. The company will invest $2–3 million in LEED-related items -- mostly efficiency gains that would have been incorporated anyway. That investment will return an estimated $750,000 in operating cost just in the first year, and at full buildout, should save more than $3 million every year.

It'll be exciting to see what comes from those three days in Texas. So far, RMI's retrofit efforts with ST Microelectronics, wrote Lotspeich, have "identified potential HVAC energy savings of 30%–50%, plus other efficiency opportunities. Collectively these retrofits had payback periods of less than two years." But designing a new fab offers far greater scope for doing it right the first time: low-friction pipes and ducts, controls that run motors at the speed instantaneously required, even free cooling by exploiting cool or dry outside air. Such a system at ST's fab near Milan "costs 80% less to operate than conventional cooling, saving $500,000 annually with a payback of one to three years, depending on the weather."

In the end, such bottom-line benefits led TI to adopt most of the Fabscape team's dozen Big Honkin' Ideas (though some await further testing and analysis). All the energy and water savings changed the net capital cost by roughly zero -- at most 1% extra, but quite possibly a decrease. Total capital cost per square foot, as required, came in at 30% below normal, blowing away industry norms and keeping the new fab in the United States.

On 15 March 2005, sponsored by leading chipmaker Applied Materials, Amory Lovins will describe TI's breakthrough to the China Semicon exposition in Shanghai. His goal: to foster still further design improvements in China, which has smart engineers and abundant pollution, scarce power and water, and an urge to leapfrog the West. Such competition is good for the world, and further opportunities remain to be exploited. Could the next fab be designed even better, to save 50% of its energy? Seventy percent? Eighty percent? Let's find out. As such radical savings emerge from the next generations of tool and system design, they may work even better and cost even less. With dedicated innovators like Paul Westbrook and his remarkable team, we're off to the races -- helping one of the world's fastest-growing and most advanced industries to reduce all forms of waste to zero.

Christina Page is a researcher/consultant with Rocky Mountain Institute.

This article has been reprinted courtesy of RMI. It was first printed in the Spring 2005 edition of the RMI Solutions newsletter.