(FORTUNE Magazine) – From the look of its stock-price chart, you'd think Millennium Pharmaceuticals was a hot Internet company. Its share price almost quadrupled between September and February--nearly matching Yahoo's surge over the period, before the latest rage for Internet stocks swept Wall Street. Millennium has no products or even drugs in clinical trials, yet its market value has surpassed that of all but a dozen or so biotechnology players, including many with medicines in advanced stages of testing.

To biotech mavens, all this may seem familiar--remember Cetus, the meteoric gene splicer that burned out in the early 1990s? But Millennium is different. For one thing, pharmaceuticals companies have awarded it research pacts potentially valued at more than $1 billion, including a contract last fall for up to $465 million from Germany's Bayer AG--biotech's largest research deal. Another difference: Thanks to the contracts, Millennium has been profitable in three of its six years of existence, not counting a one-time charge for an acquisition. Its CEO is different too. A former shoe salesman with an engineering degree, he's known for showing up in drag at company parties with his wife and daughter in tow.

Millennium's specialty is genomics, the science of identifying genes and how they work in humans and other animals as well as in microbes and plants. Eventually the company aims to become a major drugmaker, but drugs are not what sustains it now. Instead, Millennium has taken proprietary laboratory technology, such as software that helps analyze genes' functions, and integrated it with off-the-shelf devices, such as machines that decode a gene's sequence of molecular building blocks, to create a novel "technology platform" for drug companies doing genomics work.

Millennium executives grinningly refer to their partnerships with such companies as "Millennium inside," a nod to the runaway cachet of Intel's microprocessors. The smiles don't mean they're joking. Millennium is light-years from Intel in sales and profits, but its strategy resembles Intel's. With the first version of its technology platform, Millennium earned a reputation as a definer of the cutting edge. That has generated capital for updating the platform at a pace few, if any, rivals can match.

It's about time bioengineers cloned a better business model. Now two decades old, the biotech business is paying off for patients--according to a recent tally, more than 100 biotech-based drugs are on the market, and nearly 450 are in clinical trials. But for investors, its main offering has been sucker bets.

In retrospect, biotech's original game plan seems fatally flawed. Unlike large pharmaceuticals companies, its players are generally too small to spread the huge risks of drug development over many projects. Yet spreading risks is crucial--fewer than one in ten promising drug candidates makes it to market. When a biotech company's flagship drug fails in clinical trials, its ability to raise capital is undermined, often leading to a fire sale.

As such failures multiplied in the early 1990s, investors began shying away from biotech, especially its riskiest, early-stage players. Starved for capital, many were forced to license their promising products on the cheap to "big pharma." The drug giants love their little biotech sisters so much they've almost hugged them to death: Only a half-dozen of more than 1,300 biotech companies in the U.S. have gotten major drugs into the market without selling majority stakes to big pharma. And just one independent biotech company has attained a valuation that marks it as a major drug company: Amgen, whose market cap recently topped $30 billion.

The new biotech companies focused on genomics have advantages the earlier startups lacked. For one thing, the pharmaceuticals industry now faces a tough challenge: not disappointing analysts' predictions of profit growth. Wall Street is expecting the industry's earnings to rise 13% annually on average over the next five years, according to a recent report by the Boston Consulting Group. The report contends that's way too optimistic--the industry's annual growth is likely to be just 8%.

Wall Street's overoptimism partly reflects enthusiasm about blockbusters like Viagra. Yet such megasuccesses are rare: Only about three of ten drugs that reach the market ever recoup the average development costs of new medicines, according to a startling 1994 study by Duke University economists who examined drugs introduced in the early 1980s. Patents on some 30 major drugs will expire in the next three years, putting unprecedented pressure on industry profits--and on industry players to meet expectations with new blockbusters.

So drug companies desperately need genomics. Its power to delineate the molecular roots of disease speeds the search for "targets," illness-related molecules that drug compounds can stick to in order to yield benefits. Valium, for instance, binds to molecules called benzodiazepine receptors that jut out of certain brain cells. When Valium sticks to these, the cells slow their rate of sending electrical signals to other neurons, calming us down. Targets also exist in infectious microbes--antibiotics work by binding to essential molecules in the germs and gumming up their metabolic machinery.

With a target in hand, researchers can bring to bear recently developed automated systems to screen thousands of compounds quickly for any that stick to it in the test tube. These "hits" provide a vastly expanded pool of drug candidates. The result is head-snapping acceleration in drug research. Consider: Medicines developed over the past century are based on fewer than 500 targets. In its contract with Bayer, Millennium agreed to feed an astounding 225 new drug targets into the German company's research machine within five years. The contract's breadth is remarkable too: The targets Millennium promised to supply will relate to cardiovascular diseases, cancer, osteoporosis, pain, liver and blood disorders, and viral infections. It remains to be seen whether the goal can be met, but its audacity shows radical changes are afoot.

Genomics has ushered in an era of outsized deals between biotech and drug companies, says Mark Edwards, who consults with Millennium and other biotech companies. The bellwether was a $125 million pact in 1993 between Human Genome Sciences of Rockville, Md., and Britain's SmithKline Beecham. "It was the founding of the feast in genomics," says Edwards, managing director of Recombinant Capital in San Francisco.

Even at the high price, SmithKline got a terrific bargain. It soon realized its genomics partner could identify far more disease-related genes than it could use, so the companies jointly licensed the surfeit to other drug concerns. SmithKline's share of the proceeds, plus gains from stock it acquired in the genomics company, "pretty much cover" its partnership costs, says William Haseltine, CEO of Human Genome Sciences. In effect, the deal let SmithKline fill its pipeline to overflowing for free.

Incyte Pharmaceuticals in Palo Alto next emerged as Wall Street's favorite genomics company with a clever ploy: Instead of developing drugs, it compiles information on genes and sells it to drug companies in the form of easy-to-use databases. By licensing the companies to use the data nonexclusively, Incyte showed that genomics could generate hefty returns right away--forget the long waits and appalling risks of drug development. Its database subscription fees topped $100 million last year.

A number of large drug companies soon jumped into genomics, with in-house programs and collaborations with Incyte and other players in the budding niche. Merck, for instance, launched a gene-searching program with Washington University in St. Louis. The genomics niche quickly became biotech's hottest area--more than a dozen startups have styled themselves genomics specialists in the past few years.

Millennium stood out early. Aiming to become the do-all Swiss Army knife of the field, it emerged as the niche's third major contender, with Incyte and Human Genome Sciences, for drug-company partnerships. CEO Mark Levin, a plain-spoken St. Louis native with a quirky sense of humor, says he conceived the idea for the company as a venture capitalist at the Mayfield Fund in Menlo Park, Calif. Trained as a chemical engineer, Levin, 48, worked briefly at his family's shoe store before launching a corporate career that included management stints at Miller Brewing and Genentech. In 1987 he joined Mayfield to specialize in biotech investing. He wasn't interested in the usual VC shtick. Rather than wait for business plans to come over the transom, he became an agent provocateur, creating companies from ideas he would get by studying research journals and talking to scientists.

Levin's genomics lessons began in 1989 at Rockefeller University in New York City, where researcher Jeffrey Friedman told him how the new genomics techniques would soon reveal genes underlying obesity. Antennae quivering, Levin spent the next three years interviewing leaders of the Human Genome Project, the burgeoning international effort to identify all 100,000 or so human genes. By 1993, after co-founding ten biotech companies, he'd synthesized a big picture of the field and where it was headed. With Eric Lander, a gene luminary at the Whitehead Institute at MIT, and three other scientific advisers, he formed Millennium to get there first. "We decided that finding genes wasn't enough," he says, "because we knew that over the next few years all human genes would be identified. So we asked, 'Then what?' "

An answer was suggested by his work in the early 1980s as a manager at Genentech. That company emerged as biotech's first superstar, says Levin, by building up its research team and its gene-splicing technology faster than rivals did. Millennium would aim to do the same in genomics. Instead of publicizing itself as a baby drug company--a breed of dubious repute on Wall Street--it would portray itself as a technology company that happened to be in the drug-research business. Levin liked the concept so much that he hung up his traveling shoes and stayed as CEO.

Making technology the message wasn't just a PR ploy. Millennium's brain trust realized early that a broad genomics effort would produce an abundance of intellectual property. Not only would it reveal thousands of disease-related genes in humans and germs, but also every single one might become a little patent factory. For starters, a gene and the specific protein it helps produce might be patented, providing molecular targets for developing traditional drugs. A separate patent might be awarded for a bioengineered drug generated by splicing the gene into a microbe. Another patent might cover novel medicines that directly affect the gene to ameliorate the disease it's involved with. Yet another might cover the gene's use in new diagnostics for spotting individual patients' predisposition to disease. Drug researchers' crummy batting average seemed about to change--imagine 20 DiMaggios stepping up to the plate each inning.

Other industry players shared this vision, but none capitalized on it more aggressively than Millennium. Launched with $8.5 million from a venture capital group that had backed Amgen and other biotech leaders, the company rapidly assembled a crew of hard-driving young scientists and equipped them with everything from "sequencers" for decoding genes' structures to a colony of mice, the mammal of choice for gene research.

"From day zero," says Levin, "we insisted on getting passionate people who wanted to create a nothing's-impossible culture." That may be standard CEO cheerleading, but Millennium staffers echo it with feeling. "You get interviewed here about 12 times before they hire you," says Kenneth Conway, who joined the company in 1997 to head its diagnostics unit. "First they want to know whether you have enough drive and intelligence to do the job. Then they reinterview you six times to find out if you'll fit in personally."

Levin, however, takes pains to provide comic relief from the pressure to achieve--every year, for instance, he comes to the company's Halloween party dressed as a woman. "It was a lot easier to walk out of the bathroom with my costume on when there were 30 people at the company and I knew them all well," he says. "Now I walk out into this crowd of 800 wondering if most of them are thinking, 'Who is this idiot?' But it's fun."

One of Levin's early hires was Steven Holtzman, who had studied philosophy at Oxford as a Rhodes scholar before seeking his fortune in biotech. A lucid explainer whose idea of relaxation is to crack open Aristotle's Politics, Holtzman has taken the lead in crafting Millennium's mammoth deals--his gift for making fine distinctions is shown by the fact that some of the contracts are thicker than Aristotle's tome.

The deals' complexity stems from a key aspect of Millennium's strategy: slicing and dicing its intellectual property for parceling out to drug companies in maximally valuable pieces. For instance, the company has used its cancer discoveries to cut three different deals so far by divvying up rights both by tumor type and by product category: Eli Lilly bought rights to discoveries useful for developing prostate-cancer drugs, while Bayer bought similar rights related to other major cancers. In the most recent pact, Becton Dickinson agreed to pay up to $70 million for the rights to certain diagnostic applications related to skin, cervical, breast, ovarian, uterine, prostate, and colon tumors.

Drug companies, accustomed to biotech companies knocking on their doors with hat in hand, initially balked at paying premium prices for finely diced rights, says Holtzman. But the alternative, trying to build a broad genomics program in-house, posed competitive risks. In an influential 1996 paper, Jurgen Drews, retired research director of Hoffmann-La Roche, estimated that only 3,000 to 10,000 human genes, or at most 10%, will provide drug targets in the near term. It would be folly for a drug company to let rivals get ahead in the race to find and patent this limited set of key genes. Human Genome Sciences, for example, recently claimed it has already discovered more than 95% of all human genes and applied for patents on more than 3,000 of them.

John Maraganore, who manages a Millennium unit focused on bioengineered protein drugs, calls this rush to patent genes "the land grab." For the companies involved, he adds, "the important thing is to get California instead of Appalachia." In case you missed it, that's a plug for putting Millennium inside. With one of the world's best and biggest genomics programs, it can deploy more talent and technology to analyzing a novel gene's function than all but a handful of rivals. In principle, that should give Millennium's partners an edge, for the more that's known about a gene's role in the body, the easier it is to write a broad patent that covers its potential uses in developing drugs.

Roche awarded Millennium's first big contract, a $70 million pact for research on obesity and diabetes genes, only a few months after it opened its doors. After that, Millennium went on a roll--a string of ever bigger deals followed, each carefully limned to preserve the company's options for future contracts. "We fought like hell to retain as much as we could," says Levin.

All the while, Millennium refined its platform. From major medical centers it obtained access to tissue samples from sick people--researchers extract the DNA from such samples and scan it for genetic patterns correlated with diseases. To help detect such patterns, Millennium's researchers developed a system to build their own version of so-called biochips--devices pioneered by Affymetrix and others that employ miniature arrays of DNA probes to analyze hundreds of genes at once. From a pharmaceuticals partner the company imported a "library" of compounds to screen for those that bind to targets it finds via its genetic studies. The compounds thus singled out qualify as drug prototypes, which medicinal chemists can use as starting points to develop optimal versions that merit clinical tests. After raising $62 million in a 1996 initial public offering, Millennium added a major extension to its platform via an $89 million purchase of ChemGenics Pharmaceuticals, a Cambridge, Mass., company with mass-screening technology.

"Millennium has a prime-contractor mentality analogous to Boeing's," says Edwards of Recombinant Capital. Thus it can usually answer yes when prospective drug-company partners ask whether it has the latest genomics tool, such as mass-screening technology. Few, if any, competitors boast this "absence of fatal flaws" when negotiating deals with big pharma, adds Edwards.

In 1997, Millennium's opportunism prompted a surprising move: It farmed out its prized genomics platform to Monsanto for use in developing bioengineered crops. Earlier it had licensed only limited parts of its platform to research partners to prevent its technology and know-how from being copied. But in the agriculture area it faced a dilemma.

Companies like Monsanto, says Holtzman, were "at a moment of history"--they realized genomics' power would transform their industry and were ready to pay dearly for it. (At the level of molecules and genes, plants aren't all that different from animals.) But forming traditional research alliances with them would have been a distraction for Millennium, which is focused on drugs. "We struggled with ourselves," says Holtzman, before deciding to pack up the family jewels and license them to a single partner for plant research. "We finally realized that to invest enough in our platform to stay ahead, we had to focus on the bucket of gold and not worry about how many little pieces might fall through the cracks."

Monsanto brought a big bucket: It agreed to pay Millennium up to $218 million over five years for technology transferred to a new Monsanto unit, Cereon Genomics, which Monsanto set up next door to Millennium's headquarters. At the time, the contract reportedly set a record for biotech research deals. Just as important, its structure averted a major hidden cost, notes consultant Edwards: Since Millennium doesn't own Cereon, it doesn't have to add staff or labs to earn its payments from Monsanto. That's a big plus--one of biotech's nasty little secrets is that players often find themselves stuck with infrastructure they can't afford after their much publicized partnerships with drug companies wind down.

While negotiating the Monsanto deal, Millennium worked out a way to squeeze even more value from its platform--and to set up its stunning deal with Bayer. The maneuver was inspired by a bottleneck: By using biochips and other tools, genomicists can quickly detect genes that show abnormal activation patterns in diseased cells and hence may play causal roles in illness. But the researchers still must rely on laborious methods to pin down what a newly discovered gene does--knowledge that drug companies customarily want before trying to develop medicines based on the gene. The task often involves experiments that take months, such as disabling the analogous gene in mice to see what happens to the animals. If such studies establish that a gene plays a central role in a disease, researchers deem it "validated" for producing a protein target to use for automated drug-screening studies.

Unfortunately, years of validation work may lead nowhere, for many genes produce poor drug targets. Some, for instance, generate proteins that are tucked away in cells where drug molecules can't get to them.

By studying the 400 or so targets of existing drugs, however, scientists have found that certain families of genes tend to be "druggable"--the proteins they make are readily affected by known medicines. That led to Millennium's new ploy: Skip the long validation studies for every new gene that's found. Instead, select the druggable ones and quickly test whether their activation patterns are correlated with a disease. Then put genes tentatively linked to an illness in this way immediately to work generating proteins to use as drug targets. Rigorous validation studies could be done later on a small subset of genes whose proteins are found to be highly susceptible to tweaking with prototype drugs.

The strategy is akin to a bank doing cursory, up-front credit checks on mortgage applicants so that its loan officers don't waste a lot of underwriting time on deadbeats. If it works as Millennium hopes, drug discovery will become more of an automated process than ever. But putting it into effect will require placing a large bet on a novel idea--Millennium needed a partner willing to take risks.

It found one in Bayer. The aspirin king was late to the life-science party, the growing focus on biotech among drug and chemical companies. In late 1997, Bayer went shopping for a genomics partner by holding a beauty pageant. It invited Millennium and other biotech companies to pitch ideas for research alliances at an all-day session in New York City.

It wasn't much of a contest, says Kenneth Bate, who advised Bayer in its negotiations with Millennium. "Millennium just got bigger and better faster" than its rivals, says Bate, a managing director of MPM Capital in Cambridge, Mass. Its new drug-discovery idea meshed perfectly with a Bayer forte: automated screening of target proteins to find drug candidates. After nine months of negotiations, the companies announced their record-setting deal last September. For up to $465 million over five years, Bayer would get a 14% stake in Millennium plus the 225 targets. In a characteristic move, Millennium arranged to get back rights to some 90% of the targets it produces for Bayer after the German company selects a subset for use in developing drugs.

Late last year, Millennium's stock price took off in an apparent delayed reaction to the Bayer deal. "For a long time the marketplace didn't understand how much we had retained in our early deals," comments Holtzman. When investors finally got it, Millennium rocketed to the head of its class in market cap--$1.18 billion recently, vs. $827 million for Human Genome Sciences and $489 million for Incyte, its closest rivals.

Is Millennium the next Amgen, the company whose antianemia drug, EPO, was bioengineering's first blockbuster?

Not necessarily. Human Genome Sciences already has three drugs in clinical trials--if any work as hoped, that company will take off fast. And don't forget that Millennium faces rising risks and costs as it tries to evolve from a contract researcher to a company that makes and clinically tests drugs--an ambition Levin has quietly pursued from the start. Millennium says it plans to have a drug in clinical trials within a year or so, a goal it may meet by buying rights to one that already has shown promise in animal or human tests. Still, no drug-company wannabe has mustered as much value with as little red ink. If Millennium had ".com" after its name, it might well be worth more than Amgen already.