The rail cars that once carried iron ore around Republic Steel’s sprawling plant at the edge of downtown Buffalo, New York, were plowed under when the steel company abandoned the location in 1984. They were recently discovered when excavation began for the so-called gigafactory to be operated by SolarCity, the country’s leading supplier of solar panels. Now the rusted cars and a scattering of other relics from the days of Republic Steel greet visitors to the construction site, a reminder of the city’s past manufacturing might and a testament to the dream that North America’s largest solar-panel manufacturing facility can help revive it.

Buffalo is attempting an economic comeback fueled by the state’s Buffalo Billion initiative, a multi-year redevelopment plan spearheaded by Governor Andrew Cuomo. Included in the funding is support for a new genomic research center and an information technology center, but at the heart of the city’s ambitions is the solar factory, which New York is spending $750 million to build and equip. SolarCity, based in Silicon Valley, will lease it, essentially for free, and has committed to spending $5 billion on its Buffalo operations over the next decade. For Buffalo, it’s an attempt to reimagine its future around solar manufacturing. For SolarCity, it will solidify its position as one of the country’s most aggressive and fastest-growing solar companies.

The plan to build the massive manufacturing facility comes at a time when demand for solar power is booming in the United States. In 2008, the nation had about 1.1 gigawatts of photovoltaic power, the dominant type of solar energy; by the end of 2014 it had 18.3 gigawatts. Last year, homeowners, businesses, and energy companies added about 6.2 gigawatts, and they are expected to install another eight gigawatts this year. Much of that is in California, but solar power is taking hold in other states, boosted by a mix of federal tax credits and state and local incentives. Roughly a third of the electricity generation capacity added last year in the United States was solar, second only to natural-gas plants. (Even so, solar power still provides less than 1 percent of the country’s electricity.)

SolarCity has played a large part in the rapid expansion. By offering innovative financing schemes, it has spurred strong demand for rooftop panels on homes, the fastest-growing sector of the solar market. Instead of buying the expensive solar panels and paying for their installation, homeowners participating in one of SolarCity’s offerings can lease the system for 20 years, paying a monthly fee. Because it owns the panels, SolarCity benefits from the generous 30 percent federal investment tax credit for solar power; the homeowner is credited at retail electricity rates for any surplus power fed back to the grid. SolarCity is still unprofitable, but its revenue doubled from 2012 to 2014 as its leasing program proved attractive for homeowners—especially in locations with high electricity rates and lots of sunshine, such as California. The company expects to install enough panels this year to produce a gigawatt of power.

Not coincidentally, a gigawatt will be the capacity of the Buffalo factory when it is fully up and running, which is scheduled for the beginning of 2017. Until now, the company’s business has been built around marketing, financing, and installing solar systems. Instead of producing solar panels, it buys them, mostly from Chinese manufacturers. The Buffalo factory changes all that. “Our aspiration is to build many more of these factories over time,” says Peter Rive, the chief technology officer, who founded SolarCity with his brother nine years ago (their cousin Elon Musk is the company’s chairman). And though Rive says the company doesn’t want to “take its eye off the ball” in getting the Buffalo plant built and operating, he adds that shortly after that’s accomplished, “we want to create the largest solar facility in the world, never mind the Western Hemisphere.” Indeed, SolarCity stated earlier that its plan is to add “one or more significantly larger plants” with annual production capacity an order of magnitude greater than that of the Buffalo facility.

The company will make a new type of photovoltaic technology in Buffalo. The solar cells use crystalline silicon—the material used in conventional cells—with a thin film of another form of silicon and a layer of a semiconductor oxide. The hybrid solar-cell design, which SolarCity got when it bought a small company called Silevo in 2014, is designed to be more efficient than standard silicon cells in converting sunlight to electricity, as well as relatively cheap to make. But while SolarCity operates a 32-megawatt plant in Hangzhou, China, that Silevo built to make the solar cells, quickly scaling up those operations to the far larger plant in Buffalo will be an engineering feat.

Even if all goes well, the gigafactory could be facing a dramatically different solar-power market. At the end of 2016, the federal tax credit for solar power is due to drop from 30 percent to 10 percent for businesses and to disappear altogether for consumers who buy their own solar panels. By making residential solar power less affordable, the change could be devastating to the industry. And it will come just as the Buffalo factory is ramping up its manufacturing capacity.

True costs

Fears about what will happen when the tax breaks decrease are fueled by an unfortunate reality: in most locations and under most conditions, unsubsidized solar power is still far too expensive to compete with other sources of electricity. And rooftop solar is especially expensive. Subsidies and other government incentives are the reason the solar market is booming. If technologies were chosen purely on the basis of what it costs to produce power, “there isn’t a market for residential solar,” says Severin Borenstein, a professor at the Haas School of Business at the University of California, Berkeley, and an expert on electricity economics. Without government incentives for clean energy like solar, he says, “natural gas wipes everything else away.”

Much has been made of the fact that solar power is nearing grid parity—the point at which it is just as cheap as electricity from natural gas or coal. Most recently, a reportby Deutsche Bank calculated that solar power is already at grid parity in 14 U.S. states and that nearly all the others will be there by next year. But that doesn’t mean it is just as cheap to produce solar power as it is to generate electricity with natural gas. The Deutsche Bank report compares today’s cost of solar power with the retail price of electricity, which includes various charges, including fees for upgrading and maintaining the grid. That’s a sensible comparison for consumers deciding whether to install solar power. But it isn’t a true comparison of the costs to actually produce electricity. And that is the comparison that matters in determining the most cost-effective way to introduce more clean power and lower our carbon emissions.

The cost of the photovoltaic module—the chunk of silicon or other semiconductors that convert sunlight to electricity—has dropped impressively over the years. A silicon solar module sold for $4 per watt in 2008; in 2014 it was 65 cents per watt. But it has been more difficult to cut the other expenses—the so-called balance of system (BOS) costs, which include hardware like the inverters that are necessary to connect the panels to the grid and, most crucially, the labor to install the equipment. Installing heavy solar panels on the roofs of houses is particularly expensive. In such installations the BOS costs account for roughly 85 percent of the total expense of the system, according to a lengthy MIT report called “The Future of Solar Energy,” released in May. Or as Robert C. Armstrong, director of the MIT Energy Initiative and one of the authors of the report, puts it: “Even if you give away the [photovoltaic] materials for free, you still couldn’t produce electricity as cheaply as with coal or natural gas.”

Economists favor a measurement called “levelized cost of energy” to compare different sources of electricity. The calculation estimates the expense of installing a system and the average cost of producing electricity over its lifetime. When expenses are computed this way, big solar farms that directly supply electricity to utilities are the most cost-effective kind of solar power. According to the MIT report, solar-power plants in Southern California and Massachusetts would have levelized costs of 10.5 cents and 15.8 cents per kilowatt-hour, respectively (California gets far more sun, producing greater output). Meanwhile, a new natural-gas power plant can generate power at 6.6 cents per kilowatt-hour. The comparison for residential solar is even more discouraging: a Massachusetts home generates solar power at 28.7 cents per kilowatt-hour, and one in Southern California produces it at 19.2 cents, says the MIT report.

That is without government subsidies. With today’s incentives, including the tax credit, the numbers become far more favorable for solar power, even though it is still generally more expensive than producing power with gas-fired plants.

Stefan Reichelstein, a professor at Stanford University’s business school and director of the Steyer-Taylor Center for Energy Policy and Finance, and his colleagues have looked at how changing the tax credit in particular will affect solar economics. They found that even without the tax credit, large solar farms could be competitive with natural-gas plants by 2025 in states like California. But the story is very different for residential power. With a 30 percent credit, a residential solar installation produces power at less than the price of retail electricity in California (the state’s electricity rates are far higher than the national average). The same is true in other sunny states like Colorado and North Carolina, though not in a state like New Jersey. But drop the credit to 10 percent and no state is at grid parity. Take away the tax credit completely and—even assuming a continuing decrease in the cost of solar cells and installation—residential solar power remains far above grid parity in all states for many years to come.

No snow days

We will probably need vast amounts of solar power if we’re going to avoid the more dire effects of climate change. MIT’s Armstrong, for one, calculates that roughly 50 percent of the world’s electricity will need to come from solar power by 2050, requiring about 12.5 terawatts of photovoltaic capacity. We’ve barely begun the difficult and expensive transformation. Eventually, it will take vastly improved solar materials and better storage options such as batteries, as well as a realistic price on carbon emissions. But meanwhile, we need policies that are more effective in helping to make solar power a significant contributor to our electricity supply. As Armstrong says, “Money is not infinite. We need to get as much solar as we can for the money.”

The reality that the boom in solar power has depended on government subsidies doesn’t mean such incentives should end. To the contrary, it makes it obvious just how important they are to achieving the goal that society cares about: an overall reduction of carbon dioxide emissions at the lowest possible cost. But they must be carefully designed to be as fair as possible. This means, says Borenstein, that subsidies should not favor inefficient versions of clean-energy technologies, such as rooftop solar over utility-scale plants. “We need to take our thumb off the scale,” he says.

Take the practice of net metering, the policy in most states that effectively allows residents to sell solar power back to the grid at retail electricity prices. Nearly all homes with rooftop solar are connected to the grid, a necessity given the intermittent nature of solar power. These homeowners are essentially using the grid for power storage and backup, and they reap a small windfall from the high retail price of electricity in many states, including California and New York. Advocates for solar power argue that these installations add power to the grid, help offset demand during the day, and provide other benefits that stabilize the grid. Still, says Borenstein, net metering is clearly a subsidy that favors those with residential solar and adds costs to operating the grid—expenses paid for by other users.

The result has been a contentious debate in many communities and states over setting limits on the amount of solar power that qualifies for net metering. A broader danger is that incentives for solar power in general will increasingly be perceived as unfair or too costly at a time when it is clearly not yet ready to compete without subsidies. Even those who criticize the hodgepodge of existing state and federal incentives for solar, such as Borenstein and Armstrong, aren’t eager to see the tax credit suddenly change next year. “Turning off the credit abruptly could damage the [solar] industry,” says Armstrong. “And that would be a shame.”

Indeed, the change in tax policy will be a critical test of just how reliant the solar industry is on subsidies. SolarCity’s Rive believes it will cause the market to “stagnate for a couple of years.” He acknowledges that SolarCity will no longer be able to compete in several states that have low electricity rates. But he predicts his company will “be okay” given its relatively low-cost products. And he suggests that the Buffalo gigafactory could provide a strong competitive advantage.

SolarCity expects the solar panels built at the factory to be able to convert 22 to 23 percent of the sunlight that hits them into electricity, compared with about 15 to 16 percent for conventional silicon technology. That means homeowners could install fewer panels to produce the same amount of power, or they could install the same number of panels and produce more power. Either way, it could help keep the company competitive.

The new technology could be a very big deal for Buffalo, too. Overall, the gigafactory is expected to create 3,000 jobs in the city, half in the factory and another 1,500 with contractors and suppliers. SolarCity has also committed to employing 2,000 workers in the state over the next five years in sales and installation of its solar panels. It’s not exactly re-creating the jobs of the steel industry that once dominated the area, but it’s a start. (When Republic Steel closed in 1984, it had 2,500 workers, though Bethlehem Steel a few miles away in Lackawanna had far more.)

As one of the cloudiest cities in the United States, Buffalo is not a particularly attractive area for solar power. Rather, SolarCity is making its manufacturing debut there because of the state’s generous incentives and the city’s industrial infrastructure and experience. Ironically, Buffalo offers another huge benefit: the electricity rate for manufacturers averages just 4.79 cents per kilowatt-hour, which is possible because of cheap hydroelectric power generated from Niagara Falls. If the company wants to make the transition from being essentially a solar-services company to a manufacturer playing an important role in the country’s economy, Buffalo is a good place to be.

Earlier this summer, hundreds of people in the local business community packed a downtown hotel to hear SolarCity describe its plans and learn about opportunities to supply services to the gigafactory. It was the first formal meeting between executives from SolarCity, with their Silicon Valley enthusiasm and unbounded ambition, and many of the business leaders in a factory town that has suffered through decades of economic disappointments. SolarCity will produce 10,000 panels a day at the factory, one of its executive told the audience, and “bringing it down due to weather is not an option.” (No doubt the assertion made many remember that seven months earlier seven feet of snow had fallen over a few days, paralyzing the city.)

Buffalo, once the eighth-largest city in the United States, prides itself on a recent economic renaissance. The downtown and a renovated waterfront are bustling with activity and dotted with construction projects. But this renewal is heavily supported by the state’s investments. And the success of SolarCity, itself a huge recipient of the state’s funding, could be critical in determining the city’s economic future. It will also be a defining test for the role government incentives will play in further driving the expansion of solar power.