SANTA CLARA, Calif. — A year ago, Brian Halla was king of the world. He'd just turned National Semiconductor Corp. around in the wake of Gil Amelio's lackluster reign. He pumped life into earnings and helped lift the stock price to more than $40 a share. He sold businesses that didn't make sense and whittled the company strategy down to something that employees and customers alike could understand: system-on-a-chip.

This year, it's a different story. National's stock price has collapsed to $8 and the company is reporting losses instead of earnings. A financial report two weeks ago offered a glimmer of hope, but Halla acknowledges that neither the company — nor the industry — is out of the woods yet. For a moment, it seemed, National was back to the bad old days.

But the optimistic Halla is more intense than ever, vowing to second-source popular linear parts that have eaten into National's historic analog legacy, beat his alma mater Intel Corp. to 0.18-micron process capability, take on copper interconnect and make the world a safer place for low-priced, highly integrated microprocessors.

"We've seen snippets of [encouraging] data points," Halla said in an interview. "Is it a PC Christmas build or is it going to last? People are going to keep buying cell phones . . . and even [in Korea] we'll be seeing renewed CDMA orders" for the first time in a year.

What's different this time around is that everybody is hurting, not just National. Billion-dollar or near-billion-dollar companies like Analog Devices, LSI Logic, Atmel and Microchip Technology all are trading below $15 a share, getting the bum's rush from investors who were spending two and three times as much to buy their stocks a year ago. The Asian contagion has hit them all. Halla, however, is charging through the fire.

Angry that analog companies have taken business from National over the years, Halla has National embarking on an aggressive second-sourcing strategy to take back market share.

The first shot in that volley is the LM1117, a second-source 800-mA dropout regulator to Linear Technology's LT1117. First-quarter revenues from the National part were $6 million.

"If it's a huge market that we're not participating in, we're looking at it," Halla said. "Everybody out there got their starts by ripping us off. Turnabout is fair play."

National also is stepping up non-second-source analog development. It churned out a dozen new designs in the first quarter and 18 in the second.

The analog thrust is one of four corporate dicta that National formulated in February. Realizing that a difficult year lay ahead, Halla gathered managers together and the group devised a quartet of strategies, one of which was to generate more new-product revenue in analog. The other three: Ramp Cyrix microprocessor products at National's fab in South Portland, Maine; pick the next clock speeds for the Cyrix parts; and administer CPR to the company's LAN business.

The last dictum stemmed from a painful socket loss at 3Com Corp. that not only cost the company a "couple hundred million dollars," said Halla, but revealed a soft underbelly in its decision-making structure.

The device, the MACPHYTER, was the "most closely watched product we had each week in our staff meetings." Halla and his managers were trying to get the product to fab as quickly as possible. "We argued over whether we should do it as pure analog or DSP. We put so much pressure on the engineering guys that they said 'screw the die size, let's just do it fast.' "

To deal with crosstalk from a neighboring digital part, the engineers decided to design in noise-isolation circuits, which bloated the die size. As a result, "We were late and die size was so large that our cost was too much to manufacture it," Halla said.

National didn't bid the part on the 3Com job. "It was poor execution," Halla said. "The responsibility for this rests squarely on my shoulders."

Its lesson learned, National has switched to peer-group engineering reviews, where a combination of digital, analog and DSP engineers can duke it out without having to worry about having the boss there, he added.

Meanwhile, on the process front, Halla pledged 18 months ago to bring National to the leading edge of process technology. He spent several hundred million dollars to install a quarter-micron, 8-inch line on the Maine coast.

Now, Halla says he'll beat Intel's implemention of 0.18-micron technology by three months. National plans to leverage those design rules to crunch down power, so that in one design — the Media-PC, an integrated PC-on-a-chip expected in the middle of next year — battery life will start out at nine hours and move quickly to one week, Halla said.

The aggressive process moves — a complete leap from 0.65-micron to quarter-micron, skipping 0.35 altogether — has not been without its pain.

"That giant sucking sound is our new fab waiting for product to run through it," Halla said. "Forty-one percent utilization just ain't going to cut it."

Copper will soon enter the National arsenal — perhaps within a few weeks. Halla said National will engage with a partner to develop a copper process for high-speed interconnect, a technique first deployed by IBM but now embraced by the likes of Texas Instruments Inc. and Motorola Inc.

National will avoid following the IBM chemical vapor-deposition method, which requires specialized equipment, and likely will adopt a plating method similar to the one TI and Motorola use, because commercial equipment is available for the technique.

Also moving ahead aggressively is the Cyrix processor road map. Halla is continuing his push to take on Intel and others by tightly integrating key functions onto single pieces of silicon for the portable information-device market.

"We created the sub-$1,000 PC market, and AMD has 50 percent of it," Halla said, with annoyance in his voice. "We can pull ourselves out of this goo even if the market stays in recession by gaining market share."

This is where those hundreds of millions of dollars in fab investment will pay off, said Halla, who insists that no one can make a long-term business in microprocessors without having control over their process technology.

"You can't wait for a six-to-eight-week foundry turn on a design spin," he said. "We can turn from tape to silicon in 13 to 14 days. We can't get that from foundry partners."

Halla also dismisses the concept that the billion-dollar foundry model dies with the emergence of the sub-$1,000 or sub-$500 PC market.

"Intel probably makes 90 percent margins on its high-end processors," he said. "We target 45, going to 50 percent, margins. And we're happy. We can sustain that on the low end. It's not a bloodbath at all."