In a year and a half, the pain begins. Or two years, depending upon whose estimates of the chip cycle you believe. Thursday night, three top chip analysts came together to divine the future of the industry during the Churchill Club's fifth-annual Semiconductor Forecast in Palo Alto, Calif. The analysts picked their favorite stocks and the ones they would short, and explained big, new forces behind semiconductor industry growth. Dan Niles brushed off his departure this week from Robertson Stephens (he's now heading up the chip and PC areas at Lehman Brothers) and dived into the predictions. By Niles' estimates, we're one and a half years into the chip cycle, with two or three to go. "Hence my job change," he joked. Don't begin the celebrations yet, however: Niles and his colleagues agree that it's going to be a cruel summer. "The next four to five months are going to be very choppy," he says. Morgan Stanley's Mark Edelstone agreed that until August, chip investors should sit tight. Yet, he cut short Niles' estimates for a 24-to-36-month chip heyday. "I'm not sure it lasts as long as Dan said," Edelstone said. "It started at the end of '98 and will go until the end of 2001, or early 2002." Edelstone thinks we're enjoying the most prosperous, though fleeting, moments of the cycle, with more than 90% of chip companies beating quarterly estimates. But he still believes "we'll get close to $30 billion in revenues a month before the cycle is over, and we're only at $15 billion now." Which leads us to self-styled doomsayer Drew Peck of SG Cowen. Peck isn't so sure the rest of the cycle will benefit investors because "it's tough to find an investor who doesn't know that. It's difficult to surprise anyone." In other words, even with the Philadelphia Semiconductor Index's nearly 35% decline since early March, Peck says, "I wouldn't rush out and buy. That 30% growth rate is already factored in." Fine, ruin the joyous mood. Go ahead! Peck adds that he thinks chips sales are really growing at 25% rates; the 30% annual revenue-growth figures come from the esteemed Semiconductor Industry Association. He attributes the 5% difference to inventory building. And that inventory build, fair reader, is what these three chip diviners think will bring an end to the skyward trajectory of the chip cycle. Clutch those Kleenexes tightly. Niles explains that the current chip cycle likely will be felled by an experience similar to the one at the end of '95, when chip companies forecast banner shipments for PCs, bolstered by Windows 95 demand, but the demand never came. Chip companies were stuck with too many chips, had to lower prices and, wham-o, welcomed back the dirty downturn. Niles thinks this time chip companies will probably get sucked into the continuing excitement over all the new consumer and mobile devices that need flash memory and digital signal processors (DSPs). "Say it's Q1 2002, and the data-capable phones didn't sell. People didn't really want to surf on a screen this size. That's how these things happen," says Niles, explaining one of the main risks for the chip sector. So what can you do to save yourself? You've heard about component shortages and PC market sluggishness (the first true, the latter false, according to the panelists). You thought at least the continuing chip-market climb would give you some optimism. Pessimist Peck has a tip just for you: Divide stocks into companies fueled by the economy, and those fueled by technology, and shy away from those tied too closely to the economy during periods of economic uncertainty. He considers the PC industry an economy-driven business. Technologies he likes are mixed-signal components (i.e. DSPs), wave-division multiplexing (WDM) for packing more information onto fiber-optic cables, high-speed switching, and Fibre Channel, a high-speed storage networking technology. Not Intel's ( INTC) biggest fan, Peck says the chip granddaddy is too bound by economics, as is Analog Devices ( ADI), despite a 400% gain last year. His pick: SDL ( SDLI), which makes pump lasers for wave-division multiplexing. "The economy's impact on SDL is very low. Expectations are still not high enough for SDL; they can surprise on the upside." SG Cowen did underwriting for SDL in 1995 and 1996. Niles, on the other hand, likes Intel, along with Atmel ( ATML), LSI Logic ( LSI), National Semiconductor ( NSM) and Microchip Technology ( MCHP), because their revenue growth is accelerating. He also taps companies that are either "inexpensive" or are at an attractive point in their product cycles: Cypress Semiconductor ( CY), Lattice Semiconductor ( LSCC), Nvidia ( NVDA) and Xilinx ( XLNX). Niles' intentions are innocent; Lehman doesn't cover any of these companies, nor has it done banking business with them. Now, for the analyst picks. Each year, panel moderator and ex-chipper Tom Lai of Institutional Venture Partners asks the group to pick a short and long favorite. Over the last year, Peck won the longs by pinning his hopes on Cypress Semiconductor. This year, Peck went outside the cozy semiconductor world and chose Crossroads ( CRDS), because he thinks that, although it's gotten "crushed" lately, there is a market for Crossroads' routers used in storage-area networks. (SG Cowen did underwriting on Crossroads' IPO.) He went double or nothing on his monstrously bad short pick of last year, Gateway ( GTW). "It's an accident waiting to happen," he says with childlike glee. Edelstone swooped up National Semiconductor as his long pick, calling it "cheap," compared with its competitors, and concluding that it was settling down after many disruptive strategic moves. Edelstone has a strong buy on National Semiconductor, and his firm has no banking ties to the chipmaker. Edelstone wouldn't pony up a short, prompting a little ribbing from emcee Lai. "That tells you the investment advice you get out of Morgan!" Finally, Niles graced unlikely Micron ( MU) with his long affections, and squashed NETsilicon ( NSIL) with a damning short. After a little prodding, Niles confesses, "One of its big customers is about to become a big competitor." Take heart, NetSilicon: Niles lost for the first time this year because he picked AT&T ( T) to go up. Analysts are only human, it's their predictions that are divine.