(FORTUNE Magazine) – As dawn breaks in Las Vegas, an unlikely crew of venture capitalists spills onto the people mover at Bally's Casino, stogies in hand, bleary-eyed from a night at the blackjack tables. Thrown in among the rowdy regulars (they're visiting January's Consumer Electronics Show) are two of the hottest hands in VC-dom, Geoff Yang and John Walecka of Redpoint Ventures. As they stroll the Strip, the pair can't help joking about why they would blow good money in Vegas when back in Silicon Valley they are in the habit of collecting lottery-sized, 300-to-one payoffs, often in the billions of dollars. "I always say Vegas is a lot like our business," joshes Walecka. "Only our odds are a lot better."

Want a piece of that action? Dream on. By the time you or I can buy into one of Redpoint's offspring, it'll probably be too late to make big money--once these startups are public, they sell at absurd P/E ratios. The Redpoint partners stand right at the center of an ultralucrative space, Internet equipment. It is that mundane stuff, not e-tailing or B2B, that has been generating some of the biggest returns in the history of capitalism--triple digits, for more than a decade. And broadband infrastructure is just getting started: Most of the world has yet to be wired. The whole digital revolution depends on a constantly-improving network of routers, switches, filters, and fibers.

Redpoint's founders, Yang and Walecka, both 41, became infrastructure experts early, when the field was an overlooked niche. Their track records pre-Redpoint are incredible. In 1999 they formed a dream team by essentially merging the infrastructure departments of IVP, Yang's alma mater, and Brentwood Associates, Walecka's old firm. It's far too early to measure Redpoint's success, but look at the performance of the IVP and Brentwood funds, driven mostly by Yang's and Walecka's infrastructure deals. Between them, Yang and Walecka backed no fewer than 30 of the decade's top 50 infrastructure successes, including 3Com, Juniper Networks, Avici, MMC, Xylan, and Turnstone. And that's just sampling the hit parade.

Overall, IVP and Brentwood--thanks largely to Yang and Walecka--handed their investors average returns of over 100% annually for 12 years. Anyone who placed $10,000 with either firm in 1988 and then kept rolling the boodle into their new funds now has over $30 mil- lion. That dwarfs the $1.5 million you'd have pocketed backing Bill Gates at Microsoft. Those results make Yang and Walecka the most successful capitalists you've probably never heard of.

The Ruthian size of their hits is a sign of the times. In 1996, Yang invested $6 million in Foundry Networks, maker of high-speed switches tailored to the Net. Foundry went public in June 1999; Yang's investors have pocketed a gain of $800 million. Says Walecka: "In today's infrastructure world, it's heads you win big, tails you win moderately big."

Redpoint has plenty of loot to pursue its broadband vision. It just raised $1.25 billion from universities, foundations, and other institutions, an extraordinary figure for a new firm. But what if the infrastructure phenomenon is just another bubble, like the dot-com craze? Isn't it possible that the shooting stars Yang and Walecka nur- ture now will prove tomorrow's eToys or Webvans? Not likely. "An American Airlines or AmEx needs huge amounts of new bandwidth to keep improving their marketing and back office," says John Armstrong, an analyst with the Gartner Group. Jupiter Communications, the economic research firm, reckons that Internet equipment sales will jump from $153 billion in 1999 to $348 billion by 2003.

The dot-coms are growing fast, too, but they generate piddling profits or none at all. The infrastructure players hold a giant edge: Their specialty is translating sales growth into immense, consistent earnings. Companies like Juniper and Foundry aren't selling books or CDs. Their wares are the world's most innovative, sophisticated industrial goods. The uniqueness of their switches and routers ensures big, well-protected margins. Amazingly, their startup costs are also relatively low--generally $150 million or so, compared with $300 million-plus for a dot-com--because the value is in the product's design and because marketing costs are much lower. Internet equipment aces the value test: the ability to sell products for far more than it takes to finance, make, and market them.

Yang and Walecka talked extensively with FORTUNE about their formula for sifting winners. Both are tech experts--Yang an information systems major from Princeton, Walecka a Stanford electrical engineer. But as VCs they blend different strengths. Yang, the son of Chinese immigrants, is a visionary who keeps an eye out for the Net's next big turn. Entrepreneurs value his soft-spoken diplomacy. "When times are tough in a startup, he's the long-lost friend you call at 3 A.M. for comfort," says Bobby Johnson, CEO of Foundry (market cap: $7 billion).

Walecka, tall and laconic, follows the money. He filters the ideas for their potential for immense profits. For Walecka the trick is demanding that entrepreneurs focus on industry-rocking products with a ready, ravenous market. He'll send candidates back to the labs until they've found it. "Now you've got the Big Idea," Walecka will say. "Don't screw it up!" Yang's and Walecka's talents dovetail with those of partner Tom Dyal, 34, a veteran of AT&T and Bay Networks. Dyal, a Brad Pitt-handsome soccer fanatic, specializes in operations--helping startups with gritty issues like budgeting and recruiting.

A VC firm like Redpoint used to be content developing niche products, then selling the startups to a Cisco or Lucent. But these guys are convinced that giant new industries will arise in the next few years, and that they can back the startups that will dominate them. They're looking for game-shattering home runs (forget the doubles, even the triples).

How do they comb winners from the 20,000 ideas Redpoint auditions each year? One way is to have the imagination to get a step or two ahead of the marketplace even as the rate of technological change accelerates. Take, for instance, StorageWay, a company founded less than a year ago to take advantage of the explosion in fiber-optic networks.

In late 1999, Peter Shambora, a former salesman of network storage equipment, observed that long-haul phone companies were building fiber-optic trunk lines across swaths of the country. The phone companies were also erecting "data centers" with direct connections to the trunk lines and renting out space to Internet companies that needed to ship large amounts of data.

Shambora saw a huge market for data storage aborning: Along with Web servers, the dot-coms would all need vast amounts of old-fashioned storage. Shambora could create an outsourcing business to provide storage to all the new companies crowding into the data centers. He approached Redpoint, and less than six months later, StorageWay emerged. Look for a big IPO next year.

Redpoint is especially keen on entrepreneurs who have started successful companies before--so much so that it maintains a program of entrepreneurs-in-residence, investing in the people and counting on the ideas to come later. It's an idea that Yang used with success at IVP. In 1997, for instance, Yang brought in Kingston Duffie, an entrepreneur he had backed before, and a marketer named Rick Tinsley. They received salaries of $150,000 and their only obligation was to scour the Internet landscape for fresh opportunities.

Duffie and Tinsley spotted what became a huge trend. In 1996 legislation had made it far easier for new companies to provide phone service in competition with existing local phone carriers. The newcomers mainly targeted the data market, using DSL technology to transmit high-speed data over the copper network. Yang was hoping Duffie and Tinsley would invent a spectacular switch or another type of new gear to speed the DSL revolution. Instead, the two proposed something positively mundane. When the copper wires created static or sprang other problems, the phone companies, tied down by union rules, sent out crews in trucks to fix them. They even did their preventive testing manually. But Duffie and Tinsley reckoned that new equipment in the central office could monitor lines and make most of those routine repairs and tests remotely.

Duffie and Tinsley proposed creating testing gear tailored to the emerging DSL providers. Yang yawned over the concept but bet on the team and helped them found a company called Turnstone. When the DSL business took off, newcomers like Covad and Rhythms gorged on Turnstone's gear. Turnstone--which went public late last year--is heading toward $170 million in revenues this year and $60 million in operating profit. Redpoint's $5 million investment is now worth $800 million.

Right now the Redpoint partners are looking at three frontiers where they expect fortunes to be made in infrastructure: the intersection of voice and data; seamless optical networks; and the junction of the Web and TV.

A holy grail in broadband is creating a central switch that carries both voice and data, in immense volumes, with excellent fidelity. The phone system is laden with 20-year-old circuit switches built for voice; they transport only piddling streams of data. On the other hand, the powerful packet switches designed for data frequently offer mediocre voice quality and don't provide services like voice mail or caller ID. But transporting voice is far cheaper over packet switches because it's a low-cost addition to what's mainly a data processor. Combining the two fosters big economies of scale. The challenge is finding a packet switch that also excels at transporting voice.

A breakthrough, Redpoint swears, is at hand. Its recent startup, Rapid5, is poised to introduce a voice-plus-data switch that founder Shirish Patel claims will reduce equipment costs for phone companies by a factor of five. It will also slash phone charges for customers, Patel says, because it can transmit high-quality voice signals via packet switching.

Yang and Walecka believe that this new world of switches will open up a much bigger market for features. They are backing a company called Syndeo that Walecka calls the budding "Microsoft of phone features." Founded by Ted Griggs, a thatch-topped 33-year-old with three successful startups on his resume, Syndeo is developing all kinds of new features for packet-switched phones. One would let customers place calls by clicking on a name in their Internet Rolodex. Another would let, say, a refrigerator service person automatically pull up warranty information on a caller's broken fridge. Another would allow your PDA to route calls to wherever your calendar said you were at a given time of day. On the dark side of telephony, the technology would permit polling outfits to conduct elaborate surveys by phone, with automated questions answered by touchtone. The future of Syndeo's business will depend on how easily Rapid5 and others can blend voice and data. If that happens fast, Syndeo could become the wunderkind Walecka predicts.

Redpoint finds more rich turf in optical networks. Today the trunklines built by Qwest, AT&T, and WorldCom face severe restrictions. The problem is that large volumes of optical signals can't travel across the country. Once they reach the 300- or 400-mile mark they misbehave, either fading or becoming distorted. The solution is to install "regeneration" equipment along the route that turns the optical signal into electricity, then transforms the electronic pulse back into light. That process cleans up the signal, until it wobbles again after another 300 miles.

Regeneration is extremely expensive. It also becomes far more difficult when the carriers pump huge volumes of voice and data through the cables. A Redpoint startup called Ultraband Fiber Optics has invented a filter that, using acoustic technology, automatically readjusts the wayward signals to perfect shape every 100 miles or so. Hence, each light wave can travel from New York to San Francisco without switching back and forth between photons and electrons. The technology is relatively inexpensive, and if reliable, it could spur the construction of coast-to-coast networks.

Geoff Yang's pet technology is the convergence of TV and the Net. Yang prides himself on bridging the service and hardware sides. He backed TiVo, the service that allows you to record, say, all the episodes of The Sopranos that air during your vacation so that you can watch them later. For good measure, TiVo will assess your tastes and record a bunch of other crime series for you. Yang loves it; he's a self-confessed TV addict who's glued to the tube two or three hours a day.

Now Yang is enthusiastic about Big Band, a venture that claims to solve one of the thorniest problems in broadband: routing both data and video--two signals that behave totally differently--over a single cable or wire. That should make it possible to provide all the data services you get on the Internet over your TV, as well as a galaxy of customized, interactive entertainment choices. With Big Band, viewers--in theory--will be able to order from a library of thousands of films. TV watchers will be able to click on an ad and flow right into a Website for hair transplants or cars. And programmers will be able to narrowcast commercials, putting different ads for different households into the same entertainment products.

Entrepreneurial experience, cutting-edge technology, industry-transforming ideas--these are things that have made the Redpoint partners among the hottest VCs in the land. But for Yang, at least, there's another thing that counts. Not long ago he strode into a ramshackle building to audition a couple of twenty-somethings, then realized he had backed a hugely successful startup, years before, in the same magical space. Exclaimed Yang: "That means the spirits are right!" No kidding.