Applied Materials sets a 30% cycle-time target for its new $5B EPIC Center

The user in this story is a chipmaker's process engineer, and the pitch moves part of that person's tuning work into Applied's building. Prabu Raja, president of Applied's Semiconductor Products Group, said: "Because our customers' own engineers help optimize the technology before it ever reaches their factory, what leaves EPIC arrives at the fab ready to ramp to high performance and high yield sooner." ^[13]

That is what Applied says customers will do. What users have actually done is not yet on record, because the site has just opened ^[1]. The report does not include usage figures, such as how many partner engineers will work on site, or say how results are shared among partners that compete with one another. What exists today is a building, five named founding partners ^[7] and a target.

The target comes from Dickerson, as SiliconANGLE relays his New York Post interview: "The goal here is to cut that whole cycle time at least 30%, improving the time to market." ^[9] Applied to a ten-year program, 30% is three years ^[15]. The source gives a decade as the upper bound for traditional chip R&D ^[8], so three years is the most the target could save on one program, and shorter programs save less.

Dickerson also set the scope of the problem. "There are five to 10 key technologies that determine performance and yield," he said ^[10]. Intel's Lip-Bu Tan described the motive from the customer side: "As architectures become more complex and R&D costs continue to rise, we need to move faster and work more closely with our equipment partners." ^[11]

We think a team weighing a seat at a shared site like this should sort its own programs on two axes. The first is how much schedule is lost at handoffs between you and a materials or equipment supplier. Raja's pitch addresses that gap: he said EPIC is "the only place where materials, process technology and equipment will be innovated in parallel" ^[14]. The second axis is how much of the work you can do in the same building as competitors, since the site is built for cross-industry R&D under one roof ^[6].

High handoff loss and high tolerance is the case for sending one project and timing it against the program's previous cycle. High handoff loss and low tolerance points to a bilateral arrangement with the vendor. Low handoff loss leaves little for the site to save either way. The tradeoff sits in the first quadrant: it is also where your process work is closest to competitors. The comparison to run is Dickerson's, at least 30% off the earlier cycle ^[9].