A week ago, the Centers for Disease Control and Prevention confirmed what people have been suspecting: This flu season is one of the worst in recent memory. It’s on track to match the 2014-2015 season in which 34 million Americans got the flu, and about 56,000 people—including 148 children—died.

One reason behind the high toll is a mismatch between one of the flu viruses infecting people and one of the viral strains chosen almost a year ago for the global vaccine recipe, which gets rewritten every year. The dominant strain this winter is one called H3N2, which historically causes more severe illness, hospitalizations, and deaths than other strains. When the flu swept through Australia last summer, the effectiveness of the H3N2 component of the vaccine was only about 10 percent. The CDC doesn’t yet have a hard estimate for effectiveness in the United States but thinks it might be near 30 percent.

That mismatch is a bad piece of biological luck. But we should consider it a warning.

We’ve long known that our flu vaccines aren’t built to last, or to tackle every strain. But pharma companies don’t have an incentive to research drugs that will make them less money—not while current vaccines are good enough to make them $3 billion a year. To drive those new vaccines forward, medicine needs a Manhattan Project-style investment, pulling on resources outside the drug industry to force a new generation of vaccines into existence.

It’s well-known inside medicine, and little appreciated outside it, that flu vaccines aren’t as protective as most people assume. In January, the CDC collated data on flu-vaccine effectiveness from 2004 up through last year. There was no flu season in which the vaccine protected more than 60 percent of recipients. In the worst season, 2004-2005, effectiveness sank to 10 percent. That’s very different from childhood vaccines. As Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, lamented at a meeting last summer: “The measles, mumps, and rubella vaccine is 97 percent effective; yellow fever vaccine is 99 percent effective.”

The flu virus itself is to blame. The measles virus that threatens a child today is no different from the one that circulated 50 years ago, so across those 50 years, the same vaccine formula has worked just fine. But flu viruses—and there are always a few around at once—change constantly, and each year vaccine formulators must race to catch up.

The dream is to develop a “universal flu vaccine,” one that could be given once or twice in toddlerhood like an MMR vaccine, or boosted a few times in your life as whooping-cough shots are. That is a substantial scientific challenge because the parts of the flu virus that don’t change from year to year—and thus could evoke long-lasting immunity—are hidden away in the virus, masked by the parts that change all the time.

A handful of academic teams are competing to build such a new shot. They’re tinkering with the proteins that protrude from the virus, trying to take off their ever-changing heads so the immune system can respond to their conserved, unchanging stalks. They’re creating chimeric viruses from several proteins fused together, and they’re emptying out viral envelopes or engineering nanoparticles to provoke immunity in unfamiliar ways. Several of those strategies look promising in animal studies but haven’t been tested in humans. There are substantial hurdles to putting any formula into a human arm—including the fundamental one of figuring out what level of immune reaction signals that a new formula is protective enough.

And then, of course, there’s the fact that creating a new vaccine is expensive. It includes not just the cost of research and development, clinical trials, and licensing—generally accepted, across the pharma industry, to take 10 to 15 years and about $1 billion—but also the price tag for building a new manufacturing facility, which can top $600 million. Contrast that to the expenses of making the current vaccines, which use equipment and processes not changed in decades. A 2013 World Health Organization analysis pegged each manufacturers’ cost of refreshing the annual vaccine at $5 million to $18 million per year.

Now consider this: Right now, millions of people, roughly 100 million just in the United States, receive the flu vaccine every year. If those shots were converted to once or twice or four times in a lifetime, manufacturers would lose an enormous amount of sales and would need to price a new vaccine much higher per dose to recoup.

“What’s the business model here? Am I going to spend more than $1 billion to make a vaccine when I can only sell $20 million worth of doses?” Michael Osterholm asks.

The founder of the University of Minnesota’s Center for Infectious Disease Research and Policy, and a former adviser to the Secretary of Health and Human Services, Osterholm has been pushing for years to get people to notice that the market structure for the flu vaccine works against innovation. “Think about this,” he told me. “If you get a licensed product, which can take billions of dollars to achieve, how are you going to get a return on investment unless you are able to charge an exorbitant amount?”

This isn’t a hypothetical. Take the case of FluMist: As Osterholm’s CIDRAP group revealed in a 2012 report, The Compelling Need for Game-Changing Influenza Vaccines, the vaccine manufacturer MedImmune expended more than $1 billion to develop the novel nasal-spray flu vaccine. In 2009, its first year on the market, FluMist earned just $145 million. And in 2016 and 2017, a CDC advisory body recommended against using the spray at all, saying its rate of effectiveness had sunk to 3 percent.

Examples such as FluMist, Osterholm’s group wrote in their report, make it unlikely that any manufacturer will embark on a new flu vaccine or that VCs will fund them. “We could find no evidence that any private-sector investment source, including venture capital or other equity investors or current vaccine manufacturers, will be sufficient to carry one, yet alone multiple, potential novel-antigen influenza vaccines across the multiyear expenses of production,” they wrote.

As it happens, another sector of medicine is grappling with a similar problem. Since about 2000, pharma manufacturers have largely abandoned antibiotics because of a similar mismatch between investment and reward. Like vaccines, antibiotics are priced low and used for short amounts of time—unlike the lucrative cardiovascular or cancer drugs you’ll see advertised on TV and in magazines.

One answer to the funding gap has been a public-private research accelerator, CARB-X. It was founded in 2016 to dispense $455 million from the US government and a matching amount from the Wellcome Trust in England to support risky early stage research into new antibiotic compounds. Another proposal, put forward by the British Review on Antimicrobial Resistance but not yet enacted, would give roughly $1 billion in no strings “market entry rewards” to companies that get new compounds all the way through trials to licensure, counting on the cash grant to repay R&D expenses.

Osterholm thinks flu vaccines need research support, market rewards, sales guarantees, and more—a matrix of investment in research, manufacturing, and research leadership that he likens to the Manhattan Project, the all-in federal effort to build atomic bombs to bring an end to World War II. Only governments have the power to organize that scale of project, he thinks, and only private philanthropy, on the scale of the Gates Foundation or the Wellcome Trust, has the resources and the flexibility.

And he may be right. What’s clear is that the current flu vaccine market is broken. It’s important to think about that now, because this flu season marks the 100th anniversary of the worst flu known to history: The world-spanning 1918 influenza, which killed an estimated 100 million people in little more than a year. Flu pandemics arrive irregularly, and no one has been able to predict when the worst of them will come again. It would be smart of us to fix the vaccine problem before it arrives.