As Elizabeth Bowles zooms down Route 65 in her black SUV, she’s pointing out possible “vertical assets” on the flat horizon of browned cornfields and the occasional Dollar General. Out here in the Arkansas Delta, the rural area west of the Mississippi River, a vertical asset could be a tall flagpole, or a granary, or the smokestacks of a paper mill—anything high enough for Bowles’s scrappy broadband company, Aristotle Unified Communications LLC, to rig up with telecommunications equipment so it can zap the internet to far-flung customers. “See that water tank up above the pine trees?” she asks. “If you put a radio antenna on top, you can hit everything that it can see.”

Bowles is showing off her whatever-it-takes strategy for narrowing the digital divide between people with reasonably speedy internet access and those without. This gap has remained stubbornly persistent for decades, even as the internet has become steadily more inextricable from daily life, business, health care, and education. Research group BroadbandNow estimates that 42 million Americans have no broadband access, while a depressing 120 million people in the U.S. are without any connection fast enough to even call the internet, according to an October 2020 study by Microsoft Corp. These disparities are particularly severe among Black, Hispanic, Indigenous, and rural communities.

The Delta is what government officials refer to as a “high-cost area,” a remote spot with a sparse population, high poverty rate, and topography that makes everything complicated. In denser towns, it’s more economical for Aristotle to deliver broadband over fiber-optic cables, the industry’s gold standard for speed and reliability. But Bowles says it gets way too expensive in these parts. At about $9 a foot, she notes, every mile we drive deeper into the Delta would cost $50,000 or more to snake fiber through.

To augment coverage, Aristotle is turning increasingly to Citizens Broadband Radio Service (CBRS), a wireless spectrum historically used by U.S. Navy aircraft carriers for radar transmissions. In recent years the Federal Communications Commission has opened a slice of this spectrum for commercial use, enabling Aristotle to beam broadband as far as 6 miles to distant Arkansans over signal stations—installed atop cell towers, barns, even a prison—that are sort of like massive Wi-Fi routers. The network is fast enough to stream movies and costs a fraction of what fiber costs to build. “We are one of the poorest states in the country, and the Delta is the poorest area of the state,” Bowles says. “If we can solve the problem here, we can solve it anywhere.”

Although Bowles is an evangelist for so-called fixed wireless systems such as CBRS, she’s adamant that no single technology can solve the whole problem. Fiber proponents believe unspooling cables to every address in America is the only “future-proof” option capable of handling pretty much any bandwidth-heavy application of tomorrow, a premise Bowles finds ridiculous given the price tag and the scale of terrain. Silicon Valley, meanwhile, has long gone after unproven moonshots to blast internet to the masses, from Facebook Inc.’s solar-powered plane project (killed in 2018) to Alphabet Inc.’s stratospheric balloons (scrapped this year).

The messy reality on the ground in places such as Arkansas suggests that a mix of physical and wireless networks would be cheaper and more practical than some one-size-fits-all solution. Now states are looking at the feasibility of everything from CBRS to Elon Musk’s Starlink satellites to 5G home internet depending on their geographic challenges. Like Bowles, Vickie Robinson, general manager of Microsoft’s Airband Initiative, a philanthropic program to bring 3 million more rural Americans online by next July, says her team advocates making use of whatever technology is on offer. “What’s going to give you the most bang for your buck? That should be the guiding principle,” says Robinson, who’s provided Airband grant money to Aristotle for several wireless deployments.

With the Biden administration pushing ahead with its infrastructure bill, including $65 billion in broadband-related subsidies in the plan that passed the Senate, there’s a great debate playing out over which technologies the U.S. should bet on. Tom Wheeler, a former FCC chairman under President Barack Obama, is a fervent fiber-first advocate, but he acknowledges that the country will have to use every tool to connect the most isolated citizens. “I don’t care if they’re using a string and tin can if they can get the right throughput,” Wheeler says of locations fiber can’t reach. “The question becomes, where do you draw the line?”

Around lunchtime during our Delta drive, Bowles pulls into Arkansas City, a dead-quiet town of 409 residents. Aristotle’s fiber presence ended nearly 8 miles back, the point at which its network switches to CBRS and hopscotches from antenna to antenna to reach customers wirelessly. In a grass field behind the local post office, Bowles and Rick Hales, the mayor of Arkansas City, who incidentally works for Aristotle as its director of community partnerships, walk me to a 180-foot radio tower the company lit up earlier this summer. “The first customer we turned on in the county was the county judge,” Hales recalls. “His wife called and said, ‘Rick, Netflix is running good!’ ”

Aristotle was able to expand this far sooner thanks to $31 million in grants stemming from the Trump-era federal stimulus package in response to Covid-19, which marked a major turning point in broadband investment. In past decades the U.S. government’s efforts to close the digital divide were mostly sporadic and poorly funded. In 1996, President Bill Clinton committed to wiring up every classroom and library by the new millennium; the Bush administration called for universal broadband by 2007; and President Obama set a goal of giving 20 million additional Americans coverage by last year. None of it was enough.

The pandemic exposed the consequences of such buck-passing. “Nobody gets it on that visceral level until their employees can’t work or their kids can’t do homework,” Bowles says. That disparity was obvious in Arkansas City, where locals complain of past internet service from horrendously slow old-school satellite dishes and DSL connections, pricey cellular hotspots, or Vyve Broadband, a telecom that abandoned the town during the pandemic. (Vyve President Andrew Parrott says that a pilot wireless system struggled in the area’s “excessive foliage” and that the company decided to discontinue service following a severe storm in April that damaged its equipment.)

Geoffrey Wright, an assistant superintendent for Arkansas State Parks, whose office is a stone’s throw from Aristotle’s new signal tower, recalls a half-dozen students who didn’t have internet at home frequenting the picnic tables outside last fall to use the facility’s taxpayer-funded AT&T Wi-Fi for remote schooling. He’d bring out a couple of fans to cool them in the 90F heat. Wright says AT&T Inc. ran a fiber line to the parks center, but the company never offered residential service. “When the AT&T guy was here, I asked if he could run the line another 30 feet to my house,” Wright tells me. “He essentially said, ‘Nope, can’t do it. We won’t get enough customers for it to be worth it.’ ” An AT&T spokesperson says, “Our decisions to build and expand our networks are based on capacity needs and demand for our services, nothing more.”

Bowles is a pro at writing grant proposals—“I crank ’em out,” she says—but the disconnect between Washington and rural Arkansas makes the task a bureaucratic pain. The FCC’s broadband maps are notoriously inaccurate, because they rely on amorphous census blocks instead of address-by-address data. If internet providers report service at just one home in a vicinity, the whole area is considered covered, which has forced Bowles to challenge the government’s false positives to be eligible for certain subsidies. Microsoft, which analyzes national broadband data as part of its Airband Initiative, reported in 2020 that eight times as many Americans as the FCC then estimated didn’t have access to the internet at modern baseline speeds.

“The path to fiber is not dumping tens of billions of dollars into the market and building it halfway”

There’s also a strong sense that Beltway officials don’t grasp what it takes to install fiber in backcountry. Steven Porch, who doles out federal grants to state recipients as head of the Arkansas Rural Connect program, says that in rocky regions the trenching alone for cables could cost a quarter of a million dollars per mile, and stringing lines across telephone poles is risky with storms an ever-present threat. He needs to include fixed wireless as an alternative when considering how to budget for connecting the entire state. “Do I spend over $1 million to get fiber to 15 people?” he asks. “Will those 15 people sustain that network?”

Bowles says it’s been “all hands on deck” to build the 133 signal stations now dotting the state, tackling the raft of unique Arkansas hurdles, from intense humidity and pine needles interfering with radio signals to electricity shortages at rustic antenna sites to dealing with land rights for installing equipment. “One of the easements we’ve been waiting on is from a guy who’s been in Colorado elk hunting,” Hales says. “That’s not a big infrastructure problem—it’s a daggone huntin’ problem!” Even maintaining the network has been a particular challenge given the raccoons, hornets, and other creatures that burrow into its electrical systems. “We’ve fried quite a few squirrels,” says business development specialist Jonathan Duncan.

When Aristotle’s service went online in July 2021, Wright says his $60-a-month plan proved a godsend for family FaceTime chats and helped his wife manage her business from home. Jennifer Tice, owner of Mama Carol’s, a ribs-and-burgers joint across the street from the tower, became an Aristotle enterprise and residential customer. “I’m just glad my security cameras are working,” Tice says, referring to a set of web-connected devices monitoring the restaurant. “And my son, he loves it. He’s an Xbox gamer, and if it’s down, he’s like, ‘Call Rick and see why the Wi-Fi isn’t working.’ ”

Of course, as Tice’s kid suggests, Aristotle’s system is not perfect. An administrator for C.B. King Memorial, a special-education school, says internet connections have been slow and inconsistent, to the point where the school has been unable to upload billing files online. A weekday speed test shows the school at 35 megabits per second for downloads and 11 Mpbs for uploads, above the minimum broadband standard but below the 100/20 ratio government officials are moving toward. (Bowles says that the company’s network can deliver 100/20 speeds, but that C.B. King had purchased a slower plan and Aristotle’s technicians concluded the school’s old computer software is likely causing the problems.) Elsewhere, residents tell me they either weren’t aware of the service, can survive on DSL and 4G, or simply don’t see the point of the internet.

Still, considering that Aristotle only shuttered its dial-up service this past January—which the Little Rock-based company had been offering since 1995 for 50¢ per hour—the faster it can offer more connectivity options, the better for Arkansas. Other states will have to figure out their own blend of solutions. In Alaska’s icy expanses, that might mean low-Earth-orbit satellites like the ones Musk and Amazon. com Inc. are pursuing will fill in fiber gaps. In urban areas, shorter-distance 5G technology, similar to what Boston-based Starry Inc. deploys, could offer a boost where other solutions aren’t scalable. Alphabet is investing in optical systems that transfer huge amounts of data through the air using lasers. Microsoft is betting that television white space, which uses idle TV channel signals to beam internet service, could play a role, too, though the technology is nowhere near ready for prime time. “If you want ubiquitous coverage, we have to use every technology available to us,” Bowles says.

If she had her way, Bowles estimates, Arkansas could close the digital divide in five years. It’d be extremely difficult, she admits, but far faster to roll out than fiber alone and drastically less expensive. An oft-cited FCC analysis from 2017 projected that the U.S. could achieve 100% fiber coverage for approximately $80 billion. Microsoft Airband’s Robinson stresses that that forecast was based on the agency’s faulty data. “It would be a dangerous proposition to use this finite amount of money [from Congress] to get fiber everywhere,” she says. “It won’t work, because you just don’t know where all the gaps exist.”

Of the $65 billion of broadband subsidies in the Senate’s infrastructure bill, about $42 billion is allocated for equipment and service deployments, a figure Bowles finds “absurd” if the U.S. plans to push for a fiber-everywhere approach nationally. “The path to fiber is not dumping tens of billions of dollars into the market and building it halfway,” she says. Researchers at Tufts University have said a fiber-only network would cost the Biden administration at least $240 billion.

The financial distributions are particularly important, because infrastructure is only part of the problem. One important shift in the Senate’s bill is that it includes $14 billion in internet subsidies for low-income Americans, a focus on affordability that’s crucial to narrowing the gap. The more efficiently a national network can be built, presumably the more money would be available for these sorts of subsidies. For one of its grants, for example, Aristotle spent $1.9 million to deploy CBRS and a ring of fiber around the town of Hazen, bringing broadband to about 1,860 households. Bowles says a fiber-to-the-home alternative would’ve cost $5.5 million, taken at least four months longer to construct, and covered just over 600 homes.

The federal government is in the process of updating its broadband map, which is likely to change the calculus of where infrastructure spending flows. If Arkansas pandemic stimulus allocations offer any preview, the $87 million in Trump-era grants were split 50% to fixed wireless and 46% to fiber, with some of the leftover funds going to coaxial cable. A December 2020 state report said those funds are expected to cover 30,385 Arkansas homes out of the 44,874 households in the coverage footprint, a sign of how expensive this gap is to bridge even with a mix of broadband tools.

Inevitably, opinions on how to close the digital divide are available at gigabit speeds. Musk has said Starlink’s satellites can scale the gap, and cable and 5G players might claim they’ll eventually be able to deliver broadband at fiber speeds. Shirley Bloomfield, the chief executive officer of NTCA-The Rural Broadband Association, an advocacy group representing 850 small telecom companies, has reviewed everything out there, but she still thinks fiber is the only real future-proof solution today. Fixed wireless networks, after all, feed off fiber backhauls, the sort of physical broadband highways that link up the world’s internet exit ramps.

“We are pound foolish if we don’t use the right technology the first go-around,” she says of fiber. “I’m not saying you have to build it out to every igloo. But at the end of the day, and maybe it’s not the first year, but certainly in the first three to five years, it’s going to become a much better investment.”

The big concern among fiber hawks is that any alternative will consign the most disadvantaged populations, predominantly minorities and rural communities, to a slower digital lane. Bowles, however, sees a toolkit approach as the fastest way to getting more folks on the ramp to fiber speeds. Eventually, she says, the mapped lines of transition from fiber to fixed wireless will move deeper, pushing fiber backhauls closer to the homes only wireless internet can reach now. Until Aristotle can afford to go under the pine trees rustling across the Arkansas Delta, Bowles has no choice but to go over them.

Read next: Documenting the Last Pay Phones in America