Google's brilliant new plan to provide super-speedy Internet access.

Last week, Google announced that it plans to launch an "experimental" broadband network in several cities across the country. In other words, Google will become an ISP—it will provide Internet access to a small number of customers (50,000 to 500,000, depending on the cities it chooses), serving as an alternative to the broadband service offered by phone and cable companies. The news itself wasn't much of a surprise. For years, Google has been buying up "dark fiber"—extremely fast fiber-optic lines that were built during the Internet boom of the late '90s and had since been lying fallow—and observers had expected that the search company would do something with it soon. Still, the scale of Google's plans is shocking. The company wants to build home Internet connections that run at 1 gigabit per second. That's 100 or 200 times faster than the connection you're using to read this story.

Why is Google doing this? Because its future depends on better broadband. Like all Web firms, everything Google does goes through other people's lines. What's worse, the companies that run those lines have shown little interest in innovation. Most things in Google's orbit advance at a breakneck pace—computers keep getting faster, hard drives keep getting bigger, software keeps getting better. But broadband is static; it's not getting much faster, cheaper, or more widespread. Broadband—especially broadband in America—is Google's limiting factor.

The company has been trying to fix American broadband for years. Google successfully lobbied the FCC to impose "open access" policies on next-generation mobile connections and has been pushing for the government to use "white spaces"—the radio spectrum in between old analog broadcast TV channels—for wireless Internet. But Google's efforts to revolutionize broadband haven't been especially successful. It had high hopes for municipal Wi-Fi systems, but many cities abandoned those plans after the firms responsible for a few dozen networks bungled the roll-out and eventually pulled the plug. (Google still provides free Wi-Fi access in its home city of Mountain View, Calif.). Google's plan to circumvent wireless carries by selling its Nexus One phone directly to consumers also doesn't look like it's going to hurt the telecom industry very much. According to one estimate, Google sold only 20,000 Nexus Ones in its first week (Verizon's Droid saw 250,000 first-week sales, while sales of the AT&T-exclusive iPhone 3GS topped 1.6 million).

Google's decision to operate an ISP is not something a normal business would do. The company admits that in the long run, it doesn't want to be in the business of providing Internet service. It wants to show customers the promise of high-speed Internet, but it understandably doesn't want to deal with the headaches of the ISP business like customer service and tech support. This is sort of like if I ran a (very profitable) company that made travel pillows and decided to launch my own airline to prove to existing carriers that they needed to do better (and, thus, would get more people to fly and use more pillows).

Circuitous as it is, though, the ISP initiative might actually work, especially if Google picks the right cities to serve. I'd urge it to go after a few high-profile tech hubs—places like San Francisco; Austin, Texas; Boston; or some areas around Washington, D.C., and New York—where there are loads of people who want (and can afford) super-fast Internet. If Google manages to steal phone and cable companies' best customers in these few areas, it could light a fire under the telecom industry to provide better service all over. That'll be especially true once people see all the possibilities of 1GB Internet lines. Super-fast broadband has the power to overthrow the business models of entire industries. As more and more people get access, we'll be able to quit our TV and phone plans, and we'll finally have the ability to store all our data in the Internet cloud. I can't wait.

Yet the way things are going now, I'm going to have to wait. Indeed, waiting is pretty much all we do on the Internet in America. According to Akamai, which publishes a quarterly "State of the Internet" report, the average connection speed in the United States is 3.8 megabits per second. That puts America at 18^th in the world—below Denmark, the Czech Republic, Latvia, Romania, Hong Kong, Japan, and South Korea, the world's leader in connection speeds (its average speed is 11 megabits per second). What's more, we're falling further behind. A new report by Harvard's Berkman Center for Internet & Society that measures American broadband across several dimensions—speed, penetration, price—puts this country in the "middle of the pack" on "first-generation" broadband systems. We lag way behind, however, on next-generation systems. According to Akamai, nearly 10 percent of broadband customers in South Korea enjoy speeds of greater than 25 MBPS. In the United States, only 1 percent of customers get those speeds.