A technician handles a cell at the 1366 Technologies solar manufacturing demonstration plant

Funding from ARPA-Energy has led to the creation of companies such as solar manufacturing firm 1366 Technologies (now CubicPV). Here, a technician shows off a solar cell at a demonstration plant in 2013.Credit: Pat Greenhouse/The Boston Globe/Getty

US President Joe Biden’s administration wants to create a US$6.5-billion agency to accelerate innovations in health and medicine — and revealed new details about the unit last month1. Dubbed ARPA-Health (ARPA-H), it is the latest in a line of global science agencies now being modelled on the renowned US Defense Advanced Research Projects Agency (DARPA), whose work a generation ago laid the foundation for the modern Internet.

With more DARPA clones on the horizon, researchers warn that success in replicating DARPA’s hands-on, high-risk, high-reward approach is by no means assured.

“The ARPA model has been successful, and we’ve learned a lot,” says Laura Diaz Anadon, who heads the Cambridge Centre for Environment, Energy and Natural Resource Governance at the University of Cambridge, UK. “But ARPA is not a magic bullet that will apply to everything.”

Enamoured with the innovation that DARPA fostered in the United States, governments around the world, including in Europe and Japan, have attempted to duplicate the agency within their own borders. Most recently, the United Kingdom announced plans to create its version, the Advanced Research and Invention Agency (ARIA), with an initial allocation of £800 million (US$1.1 billion). And the Biden administration has proposed launching a second US agency, the $500-million ARPA-Climate (ARPA-C), to spur technologies for fighting climate change.

Scientists who have studied the DARPA model say it works if applied properly, and to the right, ‘ARPA-able’ problems. But replicating DARPA’s recipe isn’t easy. It requires the managers who build and run an agency’s grant programmes to have the freedom to assemble research teams and pursue risky ideas in promising fields that have typically been neglected by conventional industrial research and development programmes. Critics aren’t yet sure how ARPA-H, ARPA-C and ARIA will fare.

Following the recipe

The US Department of Defense established DARPA in 1958, one year after the Soviet Union launched the world’s first satellite, Sputnik 1. The goal was to avoid falling behind the Soviets, and to ensure that the United States remained a world leader in technology. DARPA was instrumental in early computing research, as well as in developing technologies such as GPS and unmanned aerial vehicles (See ‘Following in DARPA’s footsteps’).

DARPA functions differently from other major US science funding agencies, and has a leaner budget ($3.5 billion). Its roughly 100 programme managers, borrowed for stints of 3–5 years from academia or industry, have broad latitude in what they fund, and actively engage with their teams, enforcing aggressive deadlines and monitoring progress along the way. By comparison, projects funded by agencies such as the US National Institutes of Health (NIH) typically see little engagement between programme managers and the researchers they fund, beyond annual progress reports. Projects funded by these agencies also tend towards being those that are likely to succeed — and thus typically represent more incremental advances, says William Bonvillian, a policy researcher at the Massachusetts Institute of Technology in Cambridge who has studied DARPA.

Following in DARPA’s footsteps

Beginning in the early 2000s, the United States and other nations began launching agencies that aimed to capture the magic of the US Defense Advanced Research Projects Agency (DARPA).

1958: The United States establishes what will eventually become DARPA, which initially focuses on space, missile defence and the detection of nuclear weapons.

1963: A DARPA project successfully launches the first of several Transit satellites, giving US Navy ships accurate location readings and laying the groundwork for the future GPS.

1966: DARPA initiates the ARPANET project in an effort to remotely and securely link computers. Years later, this effort gives rise to the Internet.

1988: A joint project between DARPA and the US Navy produces the first long-range, unmanned aerial vehicle, Amber, which flew non-stop for more than 38 hours.

2002: US Congress establishes the first DARPA clone, the Homeland Security ARPA.

2007: Former US president George W. Bush signs the America COMPETES Act, which authorizes the creation of ARPA-Energy (ARPA-E), with the aim of developing low-carbon technologies. Two years later, former president Barack Obama formally launches the agency.

2013: DARPA awards the first grants to firms such as Moderna in Cambridge, Massachusetts, and Pfizer in New York City to support the development of messenger RNA vaccines, helping to lay the groundwork for the jabs deployed to fight the COVID-19 pandemic.

2018: Japan launches its largest DARPA-inspired programme, Moonshot, with an initial budget of ¥100 billion (US$911 million) over five years.

2019: Germany launches the Federal Agency for Disruptive Innovation (SPRIN-D) with a budget of around €1 billion (US$1.2 billion) over 10 years.

2020: The United Kingdom announces plans to launch the Advanced Research and Invention Agency (ARIA), which will receive £800 million (US$1.1 billion) to cover its first few years.

2021: US President Joe Biden calls for the creation of ARPA-Health and ARPA-Climate in his first budget outline.

The DARPA model doesn’t work if programme managers aren’t given the space to fail, says Bonvillian. When the US government applied the model to developing national-defence technologies through the Homeland Security ARPA in 2002, he adds, this was the problem. The effort eventually collapsed. “If you don’t get the culture right on day one, you have got a problem,” says Bonvillian.

Researchers also point out that a successful ARPA needs a customer for the technologies it develops. In the case of DARPA, the US military was ready to purchase many promising inventions. ARPA-Energy (ARPA-E), which was launched in 2009 under former president Barack Obama to advance low-carbon energy technologies, addressed this challenge by helping grant recipients to develop plans for commercialization from the outset — a model that Bonvillian says DARPA has also now imported.

ARPA-E had the independence it needed to function well, researchers say. Still running today, the agency, housed within the US Department of Energy (DoE), has invested $2.8 billion in nearly 1,200 projects, which have attracted another $5.4 billion in private-sector investments and led to the creation of 92 companies. Last month, one of those companies, 1366 Technologies in Bedford, Massachusetts, announced plans to build a $300-million facility for manufacturing solar cells in India. The company, now known as CubicPV, received $4 million from ARPA-E in 2009 to develop a cleaner, faster, cheaper way to manufacture the silicon semiconductors that go into solar panels.

Because it can take decades for new technologies to have commercial and societal impact, whether ARPA-E will transform the energy industry remains to be seen. But scientists have documented preliminary signs of its success2,3, as measured by patenting, publishing and, in some cases, attracting venture capital for technologies originally funded by the agency.

“The answer is yes, the [ARPA] model works, or at least it did in this case,” says Anna Goldstein, an energy researcher at the University of Massachusetts Amherst who has analysed ARPA-E’s worth. But that does not mean the model will solve all problems, she warns.

The new generation

Researchers have responded to Biden’s latest ARPA proposals with trepidation. Some scientists have questioned the need to create ARPA-C, rather than expanding ARPA-E. They point out that the two have similar missions, even though DoE secretary Jennifer Granholm has said they will not overlap. As planned, ARPA-C would seek to foster “game-changing” energy and climate solutions, including technologies such as small, modular nuclear reactors and low-energy buildings — innovations that also fall under ARPA-E’s purview.

Questions also abound about ARPA-H. The Biden administration proposed that it should be housed within the NIH, which critics worry could stifle innovation.

In a guest editorial published in Science last month1, NIH director Francis Collins and other administration officials acknowledged that the NIH tends to fund incremental research rather than bold new technologies that could transform the marketplace, and agreed that ARPA-H’s organization should be “flat, lean, and nimble”, with a culture that values “bold goals with big potential impact”. They cited potential breakthroughs in everything from vaccine development, to drug-delivery systems, to wearable medical devices.

A US Air Force RQ-4 Global Hawk unmanned aerial drone lands on a runway

DARPA investments have led to the creation of technologies such as unmanned aerial vehicles.Credit: US Air Force Photo/Alamy

The Biden administration is saying all the right things, says Bonvillian, although he still worries about whether ARPA-H will have the independence and the authority that it needs to operate within the biomedical-research behemoth. He also says the NIH will need to embrace the kind of interdisciplinary research that has been fundamental to technology development at agencies such as DARPA and ARPA-E. “If they set up an ARPA that is all biology all of the time, like NIH is, then they are going to radically limit its effectiveness,” he says.

Others worry that the scope of ARPA-H’s mission is too broad. Health care is a huge field. Given that there is already plenty of private investment in new drugs and medical therapies for prevalent diseases, Goldstein says, ARPA-H might be better placed to have an impact on neglected diseases that affect people living in impoverished and underprivileged communities. This area receives much less funding from other sources.

“The trick is setting the scope broad enough so that programme managers can wander intellectually and follow their noses, but not so broad that you try to boil the ocean,” says Eric Toone, a chemist who helped to set up ARPA-E and now works for Breakthrough Energy Ventures, a venture-capital firm based in Kirkland, Washington. This is also a potential concern with Britain’s ARIA, whose scope has yet to be defined, Toone adds.

Toone also recommends starting out small and letting new agencies grow over time. “The challenge you have with too much money is people’s expectations wind up in funny places.”