After three years and little progress, Alphabet's Sidewalk Labs is giving up on its dream of building a tech-focused neighborhood in Toronto. Citing "unprecedented economic uncertainty" — while glossing over the fact that project had not been signed off on by local government officials, and was roundly disliked by a large, vocal group of critics — CEO Daniel Doctoroff pulled the plug on the project Thursday.

Get daily insights from the Protocol team in your inbox

But as one project fades, another rises. Two former Googlers, Jonathan Winer and Brian Barlow, spun off a new company from Alphabet, confusingly called the Sidewalk Infrastructure Partnership despite being separate from Sidewalk Labs. The new venture (SIP for short) has $400 million in funding from Alphabet and the Ontario Teachers Pension Plan, and Sidewalk's Doctoroff is also mentioned in the company's SEC filings. It plans to tackle large-scale infrastructure problems of the future with investments, including plastics recycling, highways for self-driving cars, and high-speed internet in rural and urban environments.

Whether money or tech prowess alone are enough to pull off the sorts of grand designs SIP is dreaming of is unclear. But if Sidewalk Labs' own history is any indication, it won't be easy — nor will it be cheap. Infrastructure projects of the scale SIP plans to contend with take considerable resources, time, and generally local and government approval. SIP's Winer told Fortune it plans to work with local officials in the areas it chooses to invest or operate in.

SIP has a "novel tech-powered approach to modernize the [U.S.'] infrastructure like never before," according to Fortune, which will change "everything from recycling to how America deploys driverless cars." To date, the only patent that Sidewalk Labs (the original one) has been awarded relates to connected trash chutes. Some have argued that the Google-powered city would never see the light of day, and the current pandemic certainly offers some cover to fold a project of this scale. Will SIP be any different?

Although $400 million is a lot of cash for a young startup to raise in its Series A, it doesn't really encapsulate the scale of traditional infrastructure investments. SIP's website references how the steam engine and automobile reshaped infrastructure investment in the U.S., and argues that "technology will again revolutionize infrastructure and transform the way the world lives."

But building infrastructure is expensive, especially in America, where contractors have to navigate an entrenched web of bureaucracy and history. The city of New York is in various stages of completing some of the most costly infrastructure projects the world has ever seen. It's working to connect the Long Island Railroad to Grand Central Station in a project that has cost over $11 billion, and completed just part of a proposed new subway line down Second Avenue. Both projects cost well over $1.5 billion per mile to complete. (By comparison, Paris forecast building 120 miles of tracks for a subway line extension in 2016 for about $25 billion. It is well underway.)

Meanwhile, the global cost of rolling out 5G cellular networks is expected to be in the trillions. In 2019 alone, the four major U.S. networks spent over $55 billion on capital expenditures — largely on upgrading their networks to 5G. Closer to home, Alphabet has already been spending heavily on its own connectivity efforts. Google Fiber and Loon, two Alphabet companies housed in the company's "Other Bets" reporting structure, are tasked with providing access to fiber-optic internet and cellular in remote areas, respectively. Although Alphabet doesn't break out how much each Other Bets company loses, the division routinely loses hundreds of millions of dollars per quarter. And before Alphabet had spun out so many companies, it looked as if Fiber investments alone could have cost the company a few hundred million in 2018.

Building private roads or rails isn't cheap, either. The Orchard Pond Parkway, a 5.2-mile stretch of private road in Florida that opened in 2016, cost $17 million to build. In 2006, Indiana leased a 157-mile stretch of highway to an Australian-Spanish consortium, which filed for bankruptcy in 2014. Florida's Brightline (now owned by Virgin Trains), which operates the only private train line in the US, forecast the total cost of building its line at around $2.5 billion.

None of this is to say that SIP, which is aware of the "political headwinds" Sidewalk Labs has faced, can't find new ways to get work pushed through governments. But given Sidewalk and Alphabet's struggles to date, the road ahead looks rocky. In much the same way that the Valley loves to remind startups working with physical products that "hardware is hard," it's worth remembering that infrastructure is expensive.