The Google Cloud - The Google Guys

The Google Guys: Inside the Brilliant Minds of Google Founders Larry Page and Sergey Brin

Chapter 10 The Google Cloud

It’s best to do one thing really, really well.

—Number two on the list of “Ten things Google has found to be true” on the corporate Web site

To punish me for my contempt for authority, fate made me an authority myself.

—Albert Einstein

Larry and Sergey have always insisted that Google is a search company. Any assertion that it is becoming a media company or a PC company or a hardware company is dismissed as fantasy. The response is, according to Google spokesman David Krane, “We do search. But we reserve the right to define search any way we want.”

Really? Do PC applications, e-mail, cell phone operating systems, Web browsers, Wiki information sites, social networks, and photo editing sites really fall into the category of “search”? Pundits and competitors have criticized Google for stepping far beyond the boundaries of its core mission in an attempt to take over the entire computing industry. In fact, Google used to say that it focused on search but let its engineers and computer scientists dabble in whatever they wanted with 20 percent of their time. Now the company says that 70 percent of its research is focused on its core competency of search, 20 percent is dedicated to related products that include search elements, and 10 percent is anything else.

But there is logic behind most of the company’s seemingly scattershot approach to diversification. Put at the top of the list “Because they can.”

Larry and Sergey understand the Internet, and they know how to use it. In the year 2000, when pundits and publications were declaring the death of the Internet, Larry and Sergey understood its enormous potential. The Internet has made computers and communications astoundingly more efficient and less costly. When Larry and Sergey built Google’s computing infrastructure to handle rapid searches, they realized it could do so much more.

That infrastructure is too big to be contained in any one building—or any one city, for that matter. On the Internet, Google links millions of small computers into hordes that can speedily make their calculations for people in their region and speed the results back without having to travel halfway around the planet. Sometimes a single task is split among several computers that do the job in tandem, operating as a single computer. That’s the main reason Larry and Sergey decided to buy up miles of fiber optic cables at fire sale prices. This vision of massively parallel computing was articulated in the late 1990s by venture capitalists and other dot-com companies, but it took the vision of Larry and Sergey to make it happen. They didn’t just build a huge network to move information along quickly, they also created a new kind of supercomputer, one whose brain is scattered throughout the world, connected by fiber optic nerves that make it function as a single entity. It’s a network supercomputer.

Google’s supernetwork is by far the largest and most powerful network of computers in the world. It’s located in “server farms” scattered throughout the world—from Atlanta to China to Zurich—and it’s growing every day. One of the largest server farms is located in the town of The Dalles, Oregon, near sources of cheap hydroelectric power and a surplus of fiber optic cables. Called “Project 02,” the farm is the size of two football fields and has cooling towers four stories high.

The number of computers Google has installed is a closely guarded secret. When the company was just a couple of years old, Sergey revealed that the number was “over ten thousand,” and that’s still the number Google uses when asked. But the research firm Gartner Group has made an estimate based on the amount of money Google spends on computer equipment each year, and puts the total at more than one million.

These servers include load balancers, to move work from busy or broken computers to others that can handle the work; proxy servers, which store data temporarily and filter things such as spam; Web servers, which find search results; data gathering servers, which index Web pages; ad servers, to pick out relevant ads; and even spelling servers, which suggest alternate spellings of search terms.

Just as Larry and Sergey built their own network to run their search engine at Stanford, they designed a unique network when Google was young. And, just as they did at Stanford, they built it from the cheapest PCs they could get—commodity PCs running on Intel chips, running the open operating system Linux. They don’t buy the latest, most powerful generation of PCs for the network. They calculate which recent generation of PCs gives the most computing power per dollar.

Being PCs, the machines regularly fail. Because of the design of the system, that (usually) doesn’t bring down any part of the system. The work is automatically transferred to other PCs until the errant machines are pulled out of the system and replaced with new ones. Sometimes, however, too many may fail or a part of the network goes down, shutting down access to e-mail or documents or other services, causing havoc, and eliciting widespread complaints from the people who rely on them. This is the biggest potential problem in Google’s vision of network computing. Google crashes are infrequent compared to those of most systems, but many more people are affected by them.

This supernetwork is an enormous distributed supercomputer. Anybody who uses an application from Google is tapping into this incredible store of computing power. This is the main reason Google’s competitors have such a hard time matching the company’s capabilities. And it allows Google to enter any business that Larry, Sergey, or their ambitious team of computer scientists and engineers finds interesting.

The Idea Machine

Google is using this power to change the rules of business, from news delivery to PC computing to books to watching videos. No wonder many business executives are—or if they aren’t, should be—afraid of this new giant on the block. If Larry and Sergey get their way, it will be extraordinarily difficult for traditional businesses to compete as they have since the start of the twentieth century. Any business that deals in the collection or dissemination of information is in danger of having its infrastructure collapse beneath its feet like Wile E. Coyote standing on an overhanging bluff. Larry and Sergey move like roadrunners, charging ahead with their visionary plans, saying nothing about where they’re headed, or why.

There’s a good reason for that. They often don’t know where they’re going until they get there.

Google is an idea machine. Some of the ideas come from the top. Larry, for example, thought of the idea for “street views” in Google Maps—that is, actual photographs of the streets and buildings to help people identify their destinations—before Google Maps was even created. But this is not Apple, and Larry and Sergey are not Steve Jobs. The company is not centrally controlled, and ideas can come from anywhere.

The problem with most corporations is that it’s hard to get people to think differently. People are more likely to think incrementally rather than boldly. Many tech companies deal with this problem by creating formalized “skunk works,” specialized teams put together to work on an innovative new idea. At Google, everyone is a skunk. Google’s innovation machine is designed to create hundreds or thousands of research projects, many of them oneperson teams. That’s the idea behind the concept of 20 percent time, which allows scientists and engineers at the company to spend up to one day a week working on their own ideas.

At a Technology, Entertainment, Design (TED) conference for big thinkers in these fields a few years ago, Larry explained how it came about. “Both Sergey and I went to Montessori schools. For some reason this has been incorporated into Google. We’ve embodied this as the 20 percent time. For 20 percent of your time you can do what you think is important to do . . . . After all, Mendel discovered the laws of genetics as a hobby. As companies get bigger they find it harder to have small innovative projects. We had that problem too for a while. And we said, ‘Oh, we really need a new concept.’ ”1

Not every engineer has new ideas to pursue, and not everyone takes that time. Some may join up with other people’s projects. Some may use the time to take classes. Some may formally set aside one day per week for a project, while others may occasionally dabble. All that’s required is letting their managers know what they want to do. “We’re flexible,” says senior vice-president Alan Eustace. “You have to have ways to set aside the time when you need to. Otherwise the upstream stuff isn’t going to happen.”

The process goes beyond giving researchers the ability to take time off from their regular jobs. Ideas are shared, discussed, analyzed, and criticized. To make that happen, Google has created a database of information about every project every engineer is working on. All the engineers in the company have access to the database and can find out what anyone else is doing. Any engineer can examine the design documents of any project, add comments about the projects, and e-mail members of teams to offer comments and opinions. They can even rate other people’s comments. In engineering, everyone is a critic.

Larry and Sergey created that system to prevent ideas from dying before they ever get past the concept stage. “Where most companies fail on innovation is that they don’t take into account the whole pipeline of innovation,” says Eustace. “They tend to have a suggestion box mentality. Next week maybe somebody will empty the box. You can’t really get feedback on your ideas, and they die on the vine. We have a lot of different ways to get feedback.”

But they also have to withstand the scrutiny of Larry and Sergey if they think it has commercial potential. The founders decide where the company will go next. And they are ruthless judges of product designs. The product designers and engineers schedule meetings with Sergey and Larry to report on progress, and the pair does not soft-pedal their critiques. CEO Eric Schmidt describes them as “brutal.”

Tough Critics

Although Schmidt’s role is to run the business and focus on growing the company, he defers to Larry and Sergey’s judgment about which technology and products will get them there. But it’s not always a harmonious meeting of the minds. “There were times when Larry and Sergey would do things I was very unhappy about because they were so precocious,” he says. Their arguments were often heated, and Schmidt decided that it was a bad idea to conduct them publicly. “Eventually we agreed that we would have an argument privately,” he says.

The argument that led to the private meetings came a few years ago, when the company had to decide which browser Google should focus on when designing products. Microsoft’s Internet Explorer has the overwhelming market share, but technology-savvy Interneters such as Larry and Sergey favor the more elegant and open Firefox browser from Mozilla.org, a nonprofit group that allows the geeky public to design new features. As an open system, it’s easier for outsiders to create add-on programs that users can choose to add or not. Some Web sites will only work properly and fully with Internet Explorer.

A meeting was scheduled to include Larry, Sergey, Schmidt, and Google engineers to make the decision. Sergey arrived late to the meeting, and it started late. When he arrived, he hated the approach the team was favoring. “Sergey throws up all over the decision in front of everybody,” recalls Schmidt. “I said, ‘Stop! Everyone leave the room.’ I sat down and told them, don’t do that. If we’re going to have an argument, let’s have it in private.”

In that meeting, Schmidt told Sergey, “I’m willing to do what you want, although I don’t agree with it.” But he told Larry and Sergey they had to come to an agreement on the issue, and gave them until 6:00 P.M. the following day to do so. Then he left the room to let them fight it out. “This was a test to see what would happen,” says Schmidt.

What happened was not what he expected. The next day, the deadline passed with no answer. So Schmidt went to Larry and Sergey to get one. Their response: “Well, we gave the team a new set of assignments, and we’re meeting again tomorrow. We came up with a completely new approach.”

The following day, after feedback from the engineers, Larry and Sergey agreed on the new approach. They had come up with a way to reengineer the development process so that the programs would work with any browser. The issue was now moot. “All the team was happy and I learned something,” says Schmidt. “I didn’t need to decide the answer. I just needed to make sure all the ambiguities were exposed and that the right people were in the room to get it set up. Sometimes there’s a better answer that’s not in the current solution set. Had we stopped and just done what Sergey wanted or just done what I’d approved we would not have gotten to the better algorithm.”

This realization was a real mind-opener to Schmidt. “It helped me understand my role,” he says. “Now I try not to make important decisions unless they’re really important ones. Rather, I’ve tried to make the culture work. The way it feels to me is they are smarter, quicker, more opinionated and earlier than I am. They’re on an issue earlier than I am. My job is to make sure that reality and what we’re doing match together for a positive outcome.”

Schmidt still operates on that principle today. At the time I met with him in November 2008, he had just come from another debate, although a much milder one this time, with Larry. (Sergey was out of town.) Larry had long believed that there was incredible opportunity for creating applications for mobile phones. Schmidt was ready to bag that market for a while. So he met with Larry and told him that despite all the applications Google and others had built for cell phones, “there’s not anything incredible.”

So Larry pointed out one such application and told Schmidt about all the opportunities it represented for the future. Schmidt was impressed with his thinking and scheduled a meeting for the following day to discuss the potential. “The meeting happened because Larry and I were thinking similarly, but he was thinking with better specificity. I was just generally complaining.”

That’s a big part of the reason that the number of services Google offers grows constantly. Although management insists that it’s still primarily a search company, outsiders, including Wall Street analysts, see what appears to be a scattershot diversification. There are, however, specific targets in mind. Larry and Sergey are focused on all the possibilities of the Internet, and many others besides. Many of their efforts are designed to fill in spaces they feel are not being filled properly, and they’re very fond of showing others what they believe is the right approach. Says Stanford president and Google board member John Hennessy, “Larry and Sergey have very broad and eclectic interests.”

Those interests are especially strong when it comes to getting people online so they can use Google’s services, and coming up with new ways to gather information for people to search through, from e-mail to news to people’s own documents.

The Media Biz

Increasingly, Google is looking like a media company. In 2006 it paid $1.6 billion to buy YouTube, the world’s leading supplier of online video. People post homemade videos, pieces of television programs, TV commercials, and anything else that strikes their fancy. But that’s the key difference between Google and a traditional media company. Google does not create its own video content. It merely provides the means for others to create content and distribute it through Google’s services. It’s another effort to build the world’s greatest online library.

Nevertheless, YouTube has gotten Google into a cauldron of trouble. Soon after Google bought YouTube in 2006, Viacom filed a $1 billion lawsuit against YouTube and Google for allowing users to post copyright-protected videos on the site.

That suit is still crawling through the courts. In October 2008, Viacom issued a statement that said, essentially, if Google can settle with book publishers, it can do the same with Hollywood. “Copyright laws provide creators with the incentive to create the works consumers crave,” said the statement. “It is unfortunate that the publishers had to spend years, and millions of dollars, for Google to honor that principle. We hope that Google avoids the wasted effort and comes more quickly to respect movies and television programming.”

These days, such videos regularly disappear as soon as they’re posted, as Google complies with demands from the programs’ producers. But just as quickly, they reappear, posted by someone else. It’s an ongoing battle, in which Larry and Sergey firmly believe they are protected by the Digital Millennium Copyright Act of 1996, which says that Internet service providers (ISPs) are not liable for illegal postings by others using their services, as long as the operators remove infringing content and do not induce users to upload those postings.

So why get involved in video distribution in the first place? There are two reasons. For one, it’s a huge business that promises to give Google more places to park ads. The other is that it really does affect Google’s ability to improve search.

Google does not have access to all the world’s information—yet. Just as they want to digitize books so people can search through them, Larry and Sergey want to ensure that there is a large pool of video content to search through, too. That is not an easy task. The future of search is much more complex, more three-dimensional. It involves the ability to search through images, sound, music, time, and space. Google is experimenting with ways to parse any type of information so that it can be recognized and found. The easiest approach, searching through the labels applied to pictures and sounds, is extremely limited. Google needs to be able to capture words, recognize their meaning, and put them into the search index.

This is also the reason Google bought Picasa, a picturesharing site that also allows people to edit and manipulate their photos. Its research is still secret, but many other researchers are experimenting with techniques to automatically search for and identify photographs. At Stanford, researchers have been trying to identify the content of photographs by recognizing whether they include faces, or the blue skies and green grass of landscapes. Today’s cameras also capture the dates and times that images were taken, and one day it may even be possible to use global positioning satellites to determine where a picture was taken. Through Picasa, Google will have access to that data.

It’s also very important that Google can observe how people use different types of information. Everything someone posts online says something about him: his interests, opinions, likes and dislikes. By integrating people’s data from many different sources, Google’s computers collect what’s known as metadata, a powerful tool for improving its products and presence on the Internet.

The PC Biz

Metadata can come from almost anywhere. While they’re at it, how about adding everyone’s personal content to the Internet? This means e-mails, spreadsheets, documents, and other products of individual creativity. The goal is primarily to make each individual’s information available to her, to keep it from unwanted prying eyes and ears, but also to share it if she wishes. Every piece of data says something about the individual using it, allowing Google to further customize searches to that individual’s needs. And if people are using Google products to create that information, it’s easier for the company to analyze it, categorize it, and add it to the reams of metadata it has access to. When people use Google’s applications programs, such as Google Docs, the company has free access to the information (although it promises to use that info only for the creator’s benefit).

This category also pits Google against one of the most powerful technology companies of all time. Microsoft sees Google as an infiltrator into its core business, and is trying to respond in kind. The two companies have started an aggressive dance around each other’s business. So far, Google is steadily gaining ground on Microsoft, whose efforts to have an impact on search and online advertising can best be described as luke-cold.

It’s easy to say that Google is entering Microsoft’s business in order to counterattack Redmond’s forays into search and advertising. There is some truth to that, although Google executives all deny it, sometimes credibly. Says Dave Girouard, who heads Google’s applications business, “It’s not in the DNA of Google. Not that we’re not competitive, but I don’t think people at Google would ever see this as going after Microsoft. But it’s a pleasant side effect.”

But there is another temptation Larry and Sergey cannot resist. The computer industry is once again cracking at the seams, showing its inefficiencies and becoming too expensive—just as it did at the end of the mainframe era. Companies such as Google, Salesforce, and Zoho, a division of software company AdventNet Inc., are using the Internet to make computing more efficient, more powerful, and cheaper. The programs and data no longer reside in the PC, but online—in Google’s case, as part of its supernetwork. People tap into these applications through the Internet. The Internet is becoming the computer.

This trend is known as “software as a service” (SaaS) or “cloud computing.” Software is not sold and installed in personal computers; it’s run online in the amorphous Internet cloud, and rented or given away to users. The data can also be stored in the cloud.

It’s a reversal of the trend that Bill Gates exploited to make Microsoft famous. Before the invention of the microprocessor, only corporations could afford computers. The work was done on large mainframes that employees tapped into through private networks. But mainframes and their networks were easily overloaded if too many people tried to use them at once. After Intel invented the microprocessor in the late 1970s, putting most of the brains of a computer (albeit a very small one) onto a desktop PC, people no longer needed to rely on infinitely more expensive mainframes.

The Internet offers a huge benefit to even simple computing. Today, PCs are relatively expensive. A decently equipped new computer starts at about $1,000. They become obsolete after a few years. Adding too many applications causes internal communications problems as different programs try to access the same channels. Memory chips get full and slow down. People end up trashing them for a new generation of computers with better processors, more memory, and even bigger and more complex software. They spend another $1,000.

But imagine a PC designed for cloud computing. The software runs on Google’s supernetwork, so the PC does not need the latest microprocessors. The data is stored in the cloud, so the PC does not need the most expensive memory chips. The software is automatically updated, debugged, and improved remotely by Google. Under this scenario, PCs become much cheaper; the software, free (as long as you’re willing to live with a few ads); and obsolescence is slowed immensely. Fully realizing this vision will take some time, because it requires constant access to high-speed Internet connections that do not crash.

But for years Internet bloggers and analysts have repeated rumors that Google is creating its own PC operating system. It’s generally seen as an unwise attempt to undercut Microsoft. But it would also increase Google’s presence in our lives, and its access to our data, enormously. It makes sense for Google to do it.

Several bloggers have speculated that Google could adapt its Android smartphone operating system specifically for PCs and team up with PC makers to create the hardware. Some companies are already pushing the idea. A Motorola spinoff called Freescale recently announced it is developing a chipset to run Google’s Android smartphone operating system on “netbooks.” These pared-down $200 notebook computers, designed for Web surfing, are rapidly becoming the next major trend in computing. And a company called Quickoffice has announced it is creating software that will allow people to view Microsoft Word, Excel, and PowerPoint applications on their Android phones.

Google has even teamed up with Microsoft to take on Apple’s iPhone. It’s using Microsoft’s ActiveSync technology for a new utility called Google Sync, which allows people to synchronize their data from Google Calendar and Google Contact over the air with their iPhone. Apple charges $100 a year for that service.

On a lesser scale, rumors have discussed a supposed secret project within Google code-named Platypus, also known as the GDrive. Leaked documents indicate that Google is trying the system internally. GDrive is more than a memory device. Users can upload the entire contents of their computer to Google’s cloud, including all the data and software applications they use. Whenever that person is online, everything is automatically synchronized. Any data added either to the PC or the cloud is updated on the other device as well. The big advantage is this: you can drop your computer in the bathtub and completely destroy it, a disaster for most people. But with this system, you merely buy a new computer, connect to your personal GDrive with your password, and your old computer is reborn in a new shell.

Google denies all rumors, right up until the time it announces a product. For years there were rumors that Google was getting into the phone business with its own smartphone. Technically, the rumors were not quite accurate. Google created an operating system to control a cell phone, and in late 2008, cell phone manufacturer HTC announced it was building the hardware, while T-Mobile announced that the phone would work on its cellular network.

Google has already created a lesser version of the GDrive, a product called Google Gears. This program does not copy an entire PC, but it does copy all the documents users create, either on their PCs or online using Google Docs. When new documents are created or updated, the system is supposed to automatically update those documents either on the PC or online. It’s still buggy, though, and does not update reliably.

Not all the products these days are created in-house. Increasingly, Google is buying promising small companies with good ideas, a presence in a business Google covets, or technology that can enhance its own products. Google has acquired more than fifty companies to fill those needs, including blogging services, photo sharing and editing programs, e-mail, Maps, Google Earth, Web browsers, mobile phone applications, and analytical software to help run a business.

A recent article in Forbes magazine put it well. Google’s goal is to leverage the vast resources of the Internet to put together “little bits on an unimaginable scale.” With that approach, Google is unquestionably displacing Microsoft as the center of the computer industry. Microsoft’s model is PC-centric, while Google’s is Internet-centric.

There is, of course, room both for PCs operating independently and for cloud computing, but Microsoft places the emphasis on the former, with the Internet as a peripheral system, while to Google, the cloud is the computer. There are merits to both approaches. Microsoft’s is more expensive for consumers, while Google’s has a higher risk of network crashes shutting down access to one’s programs and data.

More Microsoft Mischief

Google executives have long said that they are not trying to compete directly with Microsoft. “We try not to focus on what they [Microsoft] are doing,” says Larry. But it seems to work out that way. In February 2007, Google moved from being primarily a consumer products company to entering the enterprise software field, creating products for corporations, a field now dominated by Microsoft. Salesforce.com has pointed the way with a cloud computing system that displaces corporate software from Oracle, normally run on company mainframes or mini-computers. Google’s group is headed by David Girouard, president of the enterprise business and Google Applications. A Dartmouth grad in computer engineering with an MBA from the University of Michigan, Girouard has been a consultant with business consulting firms Booz Allen and Accenture, an executive at Apple, and an entrepreneur. He joined Google in 2004. He says that putting Google’s software into corporations and getting them to run programs online is another way to collect all the world’s information, although this information is in the restricted access section of the Internet library, available only to the corporations that own it. “A lot of the world’s information is in e-mail, in filing cabinets, in the workplace. It’s an acute part of what limits people’s access to information. Web content per se was never how we defined ourselves.”

Microsoft became an enterprise computer company by sneaking in the back door. Corporate IT departments were reluctant to switch from a centrally controlled mainframe system to a distributed system of PCs that individuals owned, used, and upgraded themselves. But PCs became cheap enough for managers to buy with a corporate credit card, and they spread to the extent that IT departments had to accept and integrate them into the corporate systems. Girouard is infiltrating corporations with the same tactic. “What Internet technology is used in every business?” he asks. “Google.com. Web search. The job of my organization is to transform cloud computing into something that corporate IT does care about and makes use of. It’s a rogue approach to changing corporate computing.”

Microsoft’s desktop PC programs still dominate in corporations. But Girouard says Google is making substantial progress, and that there are more than ten million business users of its applications programs—both free ad-supported versions and paid versions more appropriate to corporations. Half a million businesses have signed up for the paid versions, with more than three thousand more signing up every day. Most are small businesses, but a few are big companies, including Genentech, Procter & Gamble, and General Electric. Virtually all the companies, however, are still using Microsoft applications as well. “We view this as a marathon, not a sprint,” says Girouard.

Microsoft’s view is a remnant of founder Bill Gates’s vision, stuck in the old model he helped create. Like the mainframers—including IBM—that Microsoft helped displace from the center of the computing world, Microsoft finds it hard to completely displace its older, more expensive model. IBM had trouble weaning itself from its expensive mainframes, and Microsoft has trouble extricating itself from its more costly PC software model. Selling software is Microsoft’s money maker. In order to compete with Google, it will have to learn to make money from leasing out software, adopting free or cheap open systems such as Linux, and generating revenue from online advertising, where it has consistently failed. “The trajectory in computer science is clearly on our side,” says Girouard.

Bill Gates, in fact, had trouble accepting the power of the Internet. Before 1995, Microsoft was trying to create its own “information superhighway,” a proprietary networking system that it alone would own and control. In 1996, I conducted an extensive interview with then-CEO Gates, who was still angrily dismissing the future power of the Internet. “Oh, sure, the Internet is everything,” he sneered. The PC visionary had trouble looking beyond his own bubble.

At Google, any idea is fair game. “Most people would like to divide the world in certain ways—enterprise products, consumer products, media,” says Girouard. “Larry and Sergey don’t see things that way. They see people and the way people use technology. They ask, ‘What problems are people having that brave, smart computer scientists can help with?’ They never assess whether we have the DNA to do it right. They just look for hard problems to solve that would have a material effect on millions of people.”

Not everything works, and Google has killed a lot of projects. In December 2008, as recession put pressure on even Google’s bottom line, Larry decided to kill a lot of projects that seemed to be going nowhere or were not successfully competing with existing products from other companies. The company ceased work on Dodgeball, a mobile social networking service, and Jaiku, a Twitter-like microblogging service. And it killed Mashup Editor, a product in private testing that allowed outsiders to create new Web applications using Google services and technologies.

Some projects have been killed by more successful and ambitious offerings within Google. One is Google Video, which was made obsolete by YouTube. Google Catalog Search fell victim to Google Book Search. Part of being innovative means being willing to fail. But it also means being able to succeed, and Google’s list of new products and new industries to tackle keeps growing.

Even re-creating products already invented by others, Google benefits from the ability to collect more data when people use Google’s version. Google’s new Chrome browser, for example, has the potential to increase Google’s knowledge of what sites people are visiting. It has its own RSS reader—a software product that allows people to automatically collect feeds from news sources and bloggers—even though other companies already have perfectly good RSS feed products. But it allows Google to collect still more data about people’s habits on the Internet and apply it to their computer algorithms.

Google executives tend to argue that the existing products, especially those created by Microsoft, are in dire need of improvement. “These apps are coming into a market owned by apps that are now twenty to thirty years old,” says Girouard. “It’s clearly the case that we have crazy new apps; that the fit and polish is not there. It’s breakthrough technology, but the degree of fit and polish, we will work on.”

As for tolerating a string of failures, most good companies in Silicon Valley know that you cannot innovate without being willing to make mistakes. And people tend to forget that Bill Gates built an empire on me-too products, most of which seemed like failures for years until the company got it right. Even Microsoft’s operating system was a derivative of one created by another company, which Microsoft licensed from it. Many of the major software applications Microsoft created were based on products invented elsewhere. And, like Microsoft, Google is wealthy enough to spend the money on experimentation.

A more substantive criticism may be that Larry and Sergey have a tendency to believe that they can do a better job than any competitor, creating many things that they could just as well buy. Google’s supernetwork is an example. Although it’s built with off-the-shelf PCs, their design is modified for efficiency, as is the network connecting them. But does Google really need to invent everything anew?

One thing Larry and Sergey have trouble with is figuring out how to complete a product. Their computer scientist approach is to release products before they’re perfected, using the Internet to collect opinions from millions of users to tweak the design. In general, it’s a brilliant approach, something that could not be done as efficiently before the Internet. And it’s a huge contrast to Microsoft, which creates its products behind closed laboratory doors and releases products that are too complicated for most people to fully exploit. Bill Gates has admitted that this ends up being the case because Microsoft relies on “experts,” mainly product reviewers, to design products. Every product reviewer makes a list of features that a product should have, and creates tables checking off the items on the list and comparing which products have the most check marks. And every reviewer has a different list.

Google’s products invariably start off with lousy reviews because of everything they lack compared with the competition. But the company carefully adds features that are really needed by most users, and ends up with a clean, easy-to-use product, if not the most versatile. The Gmail program, for example, was panned by a lot of reviewers when it first came out. It’s now widely respected for its simplicity and dependence on Google’s search capabilities, which rely on searching for names or words in a long list of e-mails rather than requiring users to organize everything into categories to make them easier to find.

The difficulty with releasing a product in “beta” form, before it’s finished, is that a product is never really finished. Most software developers just get as far as they can under the deadline they’re given, then extend the deadline before releasing the product to the public, in order to get in other features they feel certain are necessary. At Google, some products never seem to come out of beta. Gmail, for example, has been on the market for five years, and is still listed as a beta product. Google has never explained why this is the case.

It’s become something of a joke in the industry. Says board member Hennessy, “Sergey is frustrated with the fact that even when a product is ninety percent there, it’s still in beta. It’s hard to figure out how to get it to where you want it to be.” But he adds that making that decision is a problem at every innovative company. Google is just more public about it.

And overall, he’s extremely impressed with the company and its products. “Google’s execution, certainly since the IPO, has been as good or better than any company I’ve ever seen,” he says. “It’s driven by analysis and the approach of getting all the data in front of you, so you don’t spend as much time to make a decision: Is this a big issue or a small issue? Is there a quick fix or is this a major problem? That kind of thing goes on a fair amount.”

By tapping the power of technology and imagination, Larry and Sergey have started morphing Google into a new kind of company; one more diversified, perhaps less focused, certainly more ambitious. But this is also dangerous territory, technological landscapes in which they have neither experience nor a head start. Sometimes they wield Google as a blunt instrument. It’s not the best way to win friends. This time, Google is moving into new markets in which its competitors are aware of the potential and are determined to get it right. And they know who the enemy is.