The American epidemic is still concentrated primarily in gay men, and is growing rapidly worse among young black gay men.

That realization is causing a rift in the AIDS community. Activists say the persistent H.I.V. infection rate proves that the government prevention policy is a flop. Federal officials are on the defensive even as they concede that the epidemic will grow if prevention does not get better, which they know is unlikely while their budgets are being cut.

And some researchers believe it is impossible to wipe out a fatal, incurable disease when it is transmitted through sex and carries so much stigma that people deny having it and avoid being tested for it.

Looking back, epidemiologists at the Centers for Disease Control and Prevention believe that new cases peaked at 130,000 a year in the 1980s, sank slowly during the ’90s and reached a plateau at 50,000 around the year 2000.

Larry Kramer, a longtime AIDS activist and the author of “The Normal Heart,” a play about the epidemic’s early days, said: “It means I don’t see an AIDS policy, and I don’t see anyone in charge. It’s so dispiriting that it’s hard to find something to say about it. How many times can you yell ‘Help!’ without ever getting anywhere?”

Both Dr. Kevin Fenton, chief of AIDS prevention for the C.D.C., and Dr. Anthony S. Fauci, chief of AIDS research at the National Institutes of Health, took issue with Mr. Kramer’s interpretation. While both agreed that 50,000 new annual infections was, in Dr. Fauci’s words, “a great concern,” both pointed to some areas where substantial progress had been made. They said that new studies were seeking ways to get more people tested and treated early in the course of the illness, which would make them less infectious and drive transmission rates down.

“The C.D.C. is absolutely not resting,” Dr. Fenton said. “It was a major accomplishment to drop infections from 130,000 to 50,000, and we’re dealing with an epidemic that is dynamic.”

But, he conceded, 50,000 is an “unacceptably high level,” and without better prevention efforts “we’re likely to face an era of rising infection rates.”

Philip Alcabes, a public health epidemiologist at Hunter College in Manhattan, noted that 50,000 is close to the number of Americans who die in road accidents each year — almost 40,000 — “and in some ways, we consider dying on the road an ordinary thing.”

By contrast, he said, nearly one million Americans a year die of heart disease and strokes.

“So it’s not clear that prevention is a failure,” he said. “The average adult’s chances of encountering H.I.V. infection — 0.02 percent a year — are rather low. It’s not reasonable to expect that a sexually transmitted virus will disappear in America, or anywhere else. But I agree with Larry Kramer that there has been a dearth of new policy ideas.”

For most risk groups, infection rates are stable, with 61 percent of cases contracted through gay or bisexual sex, 27 percent through heterosexual sex and 9 percent through drug injections.

But they are increasing rapidly in one subgroup: young gay black men. Black teenage boys who realize they are attracted to men are often too poor to move to gay-friendly cities like San Francisco or New York, researchers said, and often must keep their homosexuality hidden from relatives and friends, making it more likely they will have furtive, risky sex.

They often lack health insurance, meaning they do not get checkups where a doctor might suggest testing. And while new surveys find that they use condoms at about the same rates as young gay white and Hispanic men, sex tends to stay within racial groups and more older black gay and bisexual men are infected. Also, untreated syphilis, whose sores open a path for H.I.V., is more common among blacks.

The National Institutes of Health is supporting studies in the Bronx, Washington and other heavily black urban areas seeking new ways to reach these men, Dr. Fauci said. Results will be ready in two or three years.

Prevention has worked for two groups, Dr. Fenton said. The number of women infecting their children at birth or through breast-feeding has dropped to only 100 a year from about 1,300 two decades ago. In that respect, the United States is like Africa: scarce public clinics focus on women and children, and many poor women see a doctor only when pregnant.

Also, the number of infections through drug use has dropped 80 percent, although that may be a result of changing fashions among addicts: Fewer inject heroin and more smoke or inhale heroin, crack, crystal meth and cocaine or swallow prescription opiates like OxyContin. Only needle-sharing passes virus-tainted blood.

Chris Collins, director of public policy for amfAR, the Foundation for AIDS Research, said the decade-long persistence of 50,000 infections “shows that we’ve failed to target prevention services adequately and have not gotten treatment coverage in many communities that would bring down community viral loads.”

A recent study has shown that getting people on antiretroviral drugs early makes them 96 percent less likely to infect others, so there is a growing outcry for “test and treat” — shorthand for actively seeking out gay men and those injecting drugs and asking them to get tested, and then helping them find medical care if they have the disease.

Dr. Fauci and Dr. Fenton said there was no discussion now of making such tests mandatory — as, for example, syphilis tests once were for marriage licenses.

San Francisco and Vancouver, British Columbia, have lowered new infection rates, Mr. Collins noted. But how applicable those lessons are to the United States as a whole is debatable; both cities have very small black populations, and Vancouver’s success relies partly on a government-approved center where drug addicts can shoot up under the eyes of a nurse and without fear of arrest — an experiment unlikely to be repeated in the United States.

The new C.D.C. figures are based partly on a new blood test that can tell recent infections from old ones, said Joseph Prejean, who led the team that made the new estimates. The test, invented in 2005 and nicknamed the “BED test,” for the B, D and E viral subtypes it uses, measures H.I.V. antibodies in the blood relative to total antibodies. That ratio rises rapidly from infection to about six months, then levels off, he said.

Dr. Alcabes, who was once a harsh critic of C.D.C. estimates, said he believed the new numbers were as accurate as they could get. “They’ve done an enormous amount of number-crunching with stupefying amounts of detail,” he said.