BELLEVUE, Wash. — Mike Nichols has a poster on his office wall. It shows the young Muhammad Ali glaring down at a fallen Sonny Liston, the bruising heavyweight who had seemed invincible — until Ali beat him, in 1964, in one of the biggest upsets in sports history, and then beat him again a year later.

“The triumphant underdog,” Nichols says, nodding toward the wall.

The inspirational fight poster is fitting, because Nichols, a general manager at Microsoft, is a lieutenant in an underdog corporate army here. Its daunting mission is to take on the Google juggernaut.

Microsoft’s assault on Google in Internet search and search advertising may be the steepest competitive challenge in business today. It is certainly among the most costly. Trying to go head-to-head with Google costs Microsoft upward of $5 billion a year, industry executives and analysts estimate.

As the overwhelming search leader, Google has advantages that tend to reinforce one another. It has the most people typing in searches — billions a day — and that generates more data for Google’s algorithms to mine to improve its search results. All those users attract advertisers. And there is the huge behavioral advantage: “Google” is synonymous with search, the habitual choice.

Once it starts, this cycle of prosperity snowballs — more users, more data, and more ad dollars. Economists call the phenomenon “network effects”; business executives just call it momentum. In search, Google has it in spades, and Microsoft, against the odds, wants to reverse it.

Microsoft has gained some ground. Its Bing search site has steadily picked up traffic since its introduction two years ago, accounting for more than 14 percent of searches in the U.S. market, according to comScore. Add the searches that Microsoft handles for Yahoo, in a partnership begun last year, and Microsoft’s search technology fields 30 percent of the total.

Yet those gains have not come at the expense of Google. Its two-thirds share of the market in the United States — Google claims an even higher share in many foreign markets — has remained unchanged in the past two years. The share losers have been Yahoo and smaller search players.

The costs for Microsoft, meanwhile, keep mounting. In the latest fiscal year, ended in June, the online services division — mainly the search business — lost $2.56 billion. The unit’s revenue rose 15 percent, to $2.53 billion, but the losses still exceeded the revenue.

Microsoft is a big, rich company. But investors are growing restless at the cost of its search campaign. In May, when David Einhorn, the hedge fund manager, called for Steven Ballmer, Microsoft’s CEO, to be replaced, he pointed to the online unit as a particular sore spot.

Qi Lu, president of Microsoft’s online services division, sees the situation this way: “To break through, we have to change the game. But this is a long-term journey.”

Microsoft’s leadership knew years ago that becoming a real competitor to Google would take patience as well as dollars. In 2007, Ballmer met with Harry Shum, a computer scientist who led Microsoft’s research lab in Beijing at the time. Ballmer, as Shum recalls, told him that the company wanted to make a concerted push in search and bring in leading technical experts and business managers.

“You spent 10 years in research, and now you’ll spend the next 10 years in search,” he remembers Ballmer saying to him.

Nouns to verbs

When Lu and Shum talk of changing the game, they mean making search smarter. Today’s search, they say, primarily finds topics, or noun phrases — a person’s name, a city, a product, a disease and so on.

“Search is still essentially a website finder.” Lu says. “It’s all nouns. But the future of search is verbs — computationally discerning user intent to give them the knowledge to complete tasks.”

The phrase that Microsoft uses is “decision engine,” as opposed to search engine.

New classes of information will help. Social network data, for example. Microsoft has an exclusive partnership with Facebook, and in May it included a feature for linking the “Like” tags of a person’s Facebook friends to that person’s search results in Bing. It might show, say, that 15 of your Facebook friends “liked” a certain restaurant. It is a first step, Lu says, in including trusted opinions in search — and not just the popular ones that conventional search does so well.

Location data, especially from the growing share of searches on smartphones, offer another rich stream of information. If the engine knows where you are, it can add another layer of context and knowledge to the search.

The ability to write increasingly responsive, full-featured applications for the Web — using the new HTML5 programming language — should also make search more intelligent. The goal, Lu says, is that someday you will speak a phrase into your smartphone — “dinner for two on Friday and movie after” — and the software will go to work. It will connect to your personal data — your location, your dining and film preferences. It will then connect to dining and restaurant reservation applications, like Yelp and OpenTable, and movie reservation applications like Fandango.

Then the engine will begin a dialogue: “Here’s what is available. Where would you like to eat and when?”

In short, Lu describes a vision of a search engine that is part intelligent software assistant and part mind reader.

In Bing, the most visible evidence of the decision-engine concept is the ability to aggregate and present specific kinds of information in a search result. Microsoft has invested in travel services, for example. Type “flights to San Francisco” into Bing, and below a few blue-shaded links to ads is a Bing Travel flight database. Click on departure and return dates, and it performs a full search of all flights and predicts whether the fare is likely to rise or fall in the days before the departure date.

But while Microsoft may be ahead in some facets, Google is innovating as well — and acquiring specialized technology to fold into its search engine. In April, the Justice Department approved Google’s $700 million purchase of ITA Software, which collects and organizes online data for airline flights. Last year, Google paid an undisclosed sum for MetaWeb, a start-up that used a vast database to better decipher the meaning, and not just the words, of search queries.

“Both these companies are making important steps to make search more intelligent,” says Oren Etzioni, a computer scientist at the University of Washington. “It’s an arms race.”

At Google, the Microsoft talk of a “decision engine” is regarded as a clever turn of phrase that merely describes the long-range ambition of search and information retrieval, as the field was called in the years before the Internet.

“The goal has always been the same,” explains Amit Singhal, a computer scientist who leads Google’s search team. “The progression is from data to useful information to knowledge that answers questions people have or helps them do things. Knowledge is the quest.”