Why Are Start-Ups Losing So Much Money?

Financial losses for today’s start-ups are much more common than they were decades ago, and the losses are much bigger. VCs are making back less from their initial investments than at any point since the global financial crisis of 2007–9. According to a study by Jay Ritter, only 22 percent of start-ups were profitable in 2021, the year of peak IPOs, versus 80 percent in the early 1980s.¹ And today’s start-ups are not becoming more profitable over time. About 85 percent of America’s unicorn start-ups (those valued at more than $1 billion before doing IPOs) that have gone public were unprofitable in 2023, despite most having been founded more than fifteen years earlier. The percentage of unicorns going public has also barely increased over the last five years.

The continuation of large start-up losses over many years means that cumulative losses are much larger than they were decades ago. As of early 2024, twenty-three American unicorns had more than $3 billion in cumulative losses, the amount Amazon had the year it became profitable. Five of them (Uber, WeWork, Rivian, Teledoc Health, and Lyft) had more than $10 billion, with Uber well over $30 billion. Other members of this club offer crypto, AI, consumer products, business software, biotech, electric vehicles, and healthcare. Despite these compa­nies having significantly higher losses than Amazon, Amazon’s eventual success continues to be used as an excuse, as if all start-ups can do what it has done.

Some observers began to notice the high percentage of start-up failures and the poor returns for venture capital as far back as 2010.² Many explanations exist, from personality clashes to short attention spans, a lack of independent thinking, and an increasing number of ven­ture capital investors, including big companies such as Google occasionally playing this role.³

Many of these explanations focus on the misalignment of incentives between start-ups and venture capital (VC) firms. VCs, for instance, tend to emphasize growth over profits, but focusing too much on growth distracts start-ups from their most important activity: experimentation. Start-ups should experiment with different designs, customers, and business models, but they often don’t. Instead, in sectors such as ride sharing, food delivery, telehealth, crypto, fintech, and business soft­ware, many start-ups around the world are doing essentially the same thing.

Venture capital funds earn fixed fees from investors that enable them to profit even if start-ups do not.⁴ These fixed fees encourage VCs to hire people who are good at raising money and spinning narratives, but not at identifying good opportunities. VCs also convinced not only investors but also cities, states, and countries that start-ups are key to economic growth. According to this narrative, governments should be concerned with maximizing VC funding, but not profits or even the productivity of the underlying technologies. VCs often spun these nar­ratives, particularly those that emphasized rapid growth, through close relationships with the media, some of which may have involved pay­ments for placing articles.⁵

VC success at raising money complemented founders who were prone to hype their personal capabilities. One empirical analysis con­cluded that “acting like an expert even without experience can help secure venture capital funding.”⁶ The mutual hype, plus lots of money, also caused the funding rounds to become larger, and the founders to become richer. Flush with money, they became hard to remove. Part of this deal, however, was that the VCs pushed a growth-at-all-cost ap­proach on the start-ups, which made it more difficult for them to experiment and thus become profitable.

Although I basically agree with these arguments, this paper will add additional and complementary reasons for why today’s start-ups have had less success than their predecessors, reasons that are explained in more detail in my book Unicorns, Hype, and Bubbles. Today’s start-ups are losing money because they have either tried to commercialize an opportunity that didn’t involve much new technology, or they have tried to commercialize one that did involve new technology, but they acted as if there were no challenges to adoption. In the latter case, they assumed the new technology would rapidly improve and quickly diffuse, and thus they didn’t think carefully enough about the first applications and custom­ers.

Some of today’s start-ups have introduced breakthrough technologies, but the rate of improvement for many of these technologies—including virtual (VR) and augmented reality (AR), eVTOL (electric vertical takeoff and landing), hyperloop, delivery drones, blockchain, and even AI—has been slow. One reason rapidly improving technologies are less common than they once were—think integrated circuits and lasers—is that fewer science-based technologies have emerged from universities over the last twenty years, the subject of an­other article I published with two colleagues in American Affairs in early 2023.

Finally, there is a preponderance of superficial thinking in the start-up ecosystem, beginning with universities and America’s top business schools. VCs, entrepreneurs, and business schools are too busy selling the genius of entrepreneurs, the wow factor, the myths of frequent incumbent failure, of disruptive technologies, and of rapid diffusion of technology.

How can we fix this system? New laws won’t help, and it will take years for VCs and business schools to change, just as it has taken dec­ades for us to get here. We need new people, business models, and ideas.

Why Profits Matter

The fundamental performance measure for firms is profits. Profits signal that a company is providing more value than the costs it is incurring. Value reflects the prices customers are willing to pay, while costs (for today’s start-ups) are mostly salaries, and thus employees of unprofitable start-ups are not delivering as much value as they are receiving.

My emphasis on value is similar to Steve Jobs’s emphasis on custom­er experience. YouTube is filled with videos of Jobs describing how companies must focus on customer experience and only those technological details that provide a good customer experience. I have merely substituted value for customer experience because I want to link value and productivity, arguing that profits often come from an advantage in productivity, which results from a firm providing more value to customers than do others, while incurring lower costs.

Ideally, productivity figures show whether a company is getting better at providing value to users. In many industries, productivity focuses on how many products can be produced per hour of labor, be it automobiles, mobile phones, or houses. Even in services sectors, such numbers are meaningful. The number of haircuts a salon can do, the number of patients that doctors or hospitals can treat, or the amount of data we can send over fixed-line and wireless networks per hour of labor are some examples.

If you think we use too much energy or land, you can focus on energy or land productivity. Does a new business enable companies to produce cars or heat homes using less energy or less land? Higher land productivity can lead to more land being available for other uses, includ­ing housing or parks.

Sometimes we must think more broadly than the narrow figures of dollars per hour of value added. Are people living longer, particularly when we adjust for the number of doctors or nurses employed? Do they have more access to clean air and water? How long does it take people to reach their destination or how many cars or passengers can be moved per hour of labor or per unit of land by a city’s transport system? The last sentence is essential to any discussion of ride hailing or robo-taxis.

My concept of productivity and value draws heavily from Robert Gordon’s descriptions of how new businesses and technologies changed American life in his 2016 book The Rise and Fall of American Growth. For instance, Gordon argues persuasively that indoor plumbing had a bigger impact on life than did any subsequent innovation, enabling not only a happier but a healthier life through better hygiene.

Such discussions become very complex for some industries, however. It is hard to measure the output from management consulting, finance, and education because they are supposed to enable improvements in productivity in other industries, the ones that produce usable goods and services. The former sectors should enable car companies to produce automobiles and housing companies to construct houses using fewer hours of labor, and for us to live longer without adding more doctors or nurses. But they often don’t.

A Lack of Breakthrough Technologies 

Let’s look at several sectors that are globally popular among start-ups but don’t involve breakthrough technologies. Ride sharing is technologically similar to taxis and thus provides little productivity advantage. As Hubert Horan has written in dozens of articles on ride sharing, the losses are not unexpected because ride sharing companies use the same vehicles, drivers, and roads as taxis—only the dispatcher has been automated.

In fact, ride hailing might be at a disadvantage from a system-wide perspective because it increases congestion. As described in a February 2020 Wall Street Journal article, entitled “The Ride-Hail Utopia That Got Stuck in Traffic,” “traffic speeds in San Francisco’s downtown core fell 21 percent to 13.7 miles an hour in 2016, from 17.4 miles an hour in 2010.”⁷ The impact of ride hailing on congestion was known decades ago and is why most cities limited the number of taxis for almost a hundred years.

Some proponents of ride sharing will claim that these firms are moving to profitability, citing Uber. Uber is the only ride-sharing com­pany to become (modestly) profitable, but its profits came from raising fares after the pandemic lockdowns caused public transit operations to be scaled back or canceled. The number of transit rides in the United States in 2022 and 2023 was only 6.2 and 7.1 billion, down from nearly ten billion in 2019, while Uber’s average U.S. ride-hailing fare per trip jumped by 50 percent between the third quarter of 2019 and the end of 2023.⁸ Should work-from-home trends reverse, America’s cities will have trouble handling the traffic unless public transport is revived.

Airbnb has been much more successful than ride sharing companies despite their similarities. Airbnb enables homeowners to profit from their homes while Uber enables car owners to profit from their cars. The big difference is that Uber’s service increases congestion on roads, while Airbnb has a much smaller negative externality. Some have criticized Airbnb because it raises the cost of living for residents by making them compete with short-term rentals. Adding new housing, however, despite oft-lamented restrictions, is still often easier than adding new road or transit capacity. And in tourist towns, Airbnb is a lifesaver for many residents, providing them with an income even when a town has no industry.

Fintech has been a mixed bag. Crypto doesn’t solve many problems, and often creates new ones. This has been true for many financial innovations of the last fifty years, because few understand that finance’s purpose is to support other sectors. On the other hand, some peer-to-peer loan start-ups have, at least occasionally, been profitable, such as Oportun and Green Sky, and to a lesser extent Lending Club and Sprout Social. They match lenders with borrowers for loans that require no collateral. When interest rates were close to zero, they worked because lenders could obtain higher rates than elsewhere, rates that borrowers were willing to pay because no bank would lend to them. But the efficacy of this approach began to decline when rates rose and lenders could put their money elsewhere.

The most profitable fintech start-up in America at one point was Block, once called Square. Square developed a two-inch-by-two-inch square device that plugs into a phone’s charge point to read a credit card. The device enables any business to accept credit card payments, thus introducing small but significant competition against older services such as PayPal. Square was the most profitable unicorn until the pandemic hit and it was overtaken by Zoom and Moderna.

Then Square became greedy. It decided to enter the crypto and BNPL (buy now pay later) segments, and changed its name to Block, in emulation of blockchain. By 2022, its profits had turned to losses. Allegations that Block was inflating the numbers of its peer-to-peer money service surfaced in 2023, causing further problems. One of the few unicorns with a good idea couldn’t resist the allure of easy money.

Online education start-ups have also incurred losses because they aren’t adding significant value. Massive open online courses (MOOCs), which made many courses accessible to virtual students at low or no cost, were once heralded as an educational breakthrough. But they don’t necessarily make the education better, and that is a big problem. They are the same courses, simply available to larger audiences, while also lacking many benefits of in-person learning as well as credentialing pathways. They also have low revenues and no profits. Even after ten years of operation and the recent pandemic, the revenues of Coursera, Udemy, and Udacity together were about $1.1bn in 2021, far less than 0.5 percent of the revenues for America’s public and private universities.

Within health care, start-ups offering telehealth and insurance are also big losers, partly because they set out to fix only fragments of a dysfunctional system. They offer telehealth-only services when most health care interactions take place in person, and these telehealth services are often prone to overprescription of drugs such as amphetamines. They mostly offer direct payment solutions despite most health care interactions being funded through insurance. They mostly target the “worried well” population while ignoring the vast majority of patients who have acute problems that are much harder to solve. When they do target the sick, for instance, with so-called digital therapeutics that try to use apps to end drug addictions, they use questionable techniques. Furthermore, many solutions were only relevant during the pandemic, a once-in-a-century event. The result is that these start-ups fail to address most real problems and the vast majority of patients’ needs.

Looking across multiple businesses, many VC-backed unicorns often do not solve and sometimes do not even attempt to address the real issues facing an industry. Health care start-ups, for example, ignored the problems of multiple health care databases. Business software start-ups are also guilty of the same problem on a larger scale, leaving the incum­bents to deal with issues of integration. Online education start-ups did not solve the credentialing problem, or more fundamentally solve the problem of less-than-relevant educational content being offered by the incumbents. Rethinking what universities should offer does not appear to have crossed their minds.

Underwhelming unicorn start-ups have another common factor: many of them are platforms that connect users and different suppliers. But the failure of these start-ups suggests that platforms are often not as beneficial to customers and investors as their proponents have claimed.

Slow Adoption of New Technologies

Some technologies do potentially offer value, but large-scale adoption is still years away, and thus start-ups need to focus on niches and not the main market. For instance, the tech sector is no closer to solving the problem of nauseousness from VR or AR than it was twenty years ago, despite the expected “iPhone moment” when everyone knows the technology has arrived. This is why Microsoft has had so much trouble delivering AR headsets to the U.S. Army, despite the $183,000 per headset implied in its contract. Microsoft can’t seem to make the Hololens work well without making soldiers sick.⁹ If this kind of money can’t make it work, what chance is there of AR or the metaverse becoming big markets?

The progress is also slow for the size of the headsets, with even Apple’s Vision Pro not any smaller than previous headsets. The reason for the lack of progress is because there is a tradeoff between size and field of view, and thus making headsets smaller also reduces the field of view (FoV), a result users don’t want.

A similar lack of progress can be seen for other technologies of the 2010s that barely diffused. This includes smart homes, smart cities, other wearables, satellite internet, delivery drones, space tourism, and eVTOLs (electric vertical takeoff and landing). For instance, delivery drones cannot deliver packages to the balconies of high-rise apartments, which is the dominant form of dwelling in big cities. Nor can they navigate power and telephone lines, a feature of suburbs or smaller cities, and they are very noisy.

Some of these technologies have devoted followers, but slow progress. Smart watches have the largest current market, and I am the most optimistic about them, but their progress is still slow, and only a few doctors believe they are useful. Elon Musk’s satellite internet ser­vice, Starlink, also has devoted followers, but just over four million global subscribers (by comparison, there are 8.9 billion cellular phone subscrip­tions in the world), and its speeds are barely improving, a big reason why the U.S. government rejected its proposal to provide rural services.

Investors began to recognize these problems in 2021 as share prices for many providers of these technologies began to decline, Apple being a notable exception. By 2022, many leading business media outlets began to say that interest in these types of technologies was gone. “Lofty Tech Visions Are Going Away” and “Enthusiasm for the Metaverse Remains Overheated” were titles from the Financial Times. “Big Tech Stops Do­ing Stupid Stuff” and “The Craziest Moments from the Longest Tech Boom (So Far)” were titles from the Wall Street Journal.

Due to the slow rates of progress, investors and innovators need to carefully identify those customers who have the largest willingness to pay or have the lowest requirements. Thinking the mass market will take off is a big mistake. Elon Musk seems like one of the few innovators to be realistic in his approach, admitting that Starlink will always be a niche and targeting customers like airlines.

AI is now arguably experiencing its own bubble, and the debate about improvements is heated, particularly for generative AI. For gen­erative AI, the number of new products being released causes many to think the progress is rapid. But when it comes to reducing the number of hallucinations, progress has proven difficult. Some point to continued hallucinations in new products such as Sora while others claim that hallucinations are inevitable. For instance, an article in Nature found that “Larger and more instructable language models become less relia­ble,” and IEEE Spectrum agrees: “The prevailing methods to make large language models more powerful and amenable have been based on continuous scaling up (that is, increasing their size, data volume and computational resources) and bespoke shaping up (including post-filter­ing, fine tuning or use of human feedback). However, larger and more instructable large language models may have become less reliable.”¹⁰

Superficial Thinking

Behind these poor product design decisions and business strategies were rising hype and what I call the “wow factor,” which is the practice of presenting optimistic goals without explaining how they will be achieved: Ride sharing will eliminate all private cars and thus all parking lots, freeing millions of acres in space. Cloud kitchens will replace restaurants because we are too busy (and too rich) to cook for ourselves. Neo-banks will replace traditional banks because they don’t need a building, and algorithms can make better decisions than humans. Crypto will prevent governments from enabling inflation. Telehealth will elimi­nate doctors’ offices.

One way that VCs hyped start-ups with the wow factor was by pushing the myth of genius entrepreneurs, and the media was initially cooperative in advancing this narrative because they weren’t very interested in debating the economics of the businesses. For instance, Forbes gave “30 under 30” awards to start-up founders, some of whom, including Charlie Javice, Sam Bankman-Fried, and Martin “Pharma Bro” Shkreli, were later indicted for fraud and found to be running scams.¹¹ I saw the superficiality of these awards in Asia as a judge in 2023; many of the candidates had reputedly created “successful” block­chain or NFT companies, but genuinely successful companies in these spaces simply do not exist.

If an award didn’t provide enough evidence that the founders were geniuses, funding was often considered a measure of success for the start‑ups themselves, as opposed to profits or happy customers. Unicorn valuations largely reflected the amount of money they had raised, demonstrating how inputs are often confused with outputs. VCs even created the terms “decacorn” and “hectocorn” for start-ups valued at more than $10 billion and $100 billion, respectively—another way to avoid the issues of profits, economics, and other traditional measures of performance. Instead, they appealed to the “fear of missing out” on the “most innovative era in human history.”

A third method of pushing the wow factor was to claim that disruption was so common that it must be easy, a message pushed by the proponents of disruptive technologies and those claiming that improvements are much more rapid than they were a hundred years ago. One source of this myth is the small number of companies (fifty-two) on the 1955 Fortune 500 list that are still on the list almost seventy years later. The problem with this narrative of start-ups disrupting old companies is found in the details. When the Fortune 500 companies’ roots are traced back to their beginnings, the average company in 1955 was sixty-three years old, but by 2020 this had increased to one hundred years old—not a sign of newly founded start-ups disrupting old-line companies. The roots of many of today’s leading banking and insurance companies can be traced back to before America’s Civil War. Perhaps more importantly, the Fortune 500 companies of 2020 were often not even operating in the same industries as those of 1955, but America’s loss of manufacturing and other capital-intensive sectors is not necessarily a sign of its technological prowess or progress.

Fixing the System

If all you heard was VC hype, you could be forgiven for thinking that start-ups have driven American growth in recent years. Far from engines of American innovation and technological progress, however, the start-ups of the last few decades have yielded, at best, slow-burning losses for investors. The start-up system is perhaps the least steady pillar of the broader American private-sector innovation system, and cannot be relied upon to produce the breakthroughs in the areas where it should be most useful.

Unprofitable and uninspiring start-ups are not just remnants of the zero-interest-rate environment, but the products of decades of deleterious change not only in venture capital and entrepreneurship programs at universities, but also changes in America’s system of basic and applied research. A 2023 article in American Affairs by myself and colleagues outlined those changes in the latter, with a big move from corporate labs to academic research, relying on huge university labs that produce hundred-author papers for hyperspecialized journals, employ hundreds of PhD and post-doc students, and require a vast bureaucratic system of proposals and letters of recommendation.

The result is the declining emergence of successful science-based technologies, which in turn has reduced the number of good opportunities for start-ups to exploit. Meanwhile, university-run start-up incuba­tors—contrived to replicate the results of past convergences of technical and entrepreneurial genius at universities like Harvard, MIT, and Stanford—steer students toward app design and away from business economics and innovative scientific research.

Fixing the current system of start-ups will require us to rethink our assumptions about new businesses and new technologies, eliminate some sacred cows, and basically remake a system that, at present, overflows with money but not self-reflection. It will require converting what is now an arena for founders chasing windfall gains into a system for targeted and profitable investments in breakthrough technologies. It will, in other words, require deeper changes not only in the start-up system but in systems of training entrepreneurs and engineers.

At every level of a start-up, founders, managers, and workers often lack the capability to run a profitable business; likewise, VC investors often lack the patience to allow firms to experiment with products and strategies. This lack of judgment is excused by the standard, scattershot approach of early-stage VC investors, which countenances the eventuality that most invested firms will fail. Periodic hype cycles amplify this behavior and lead to the proliferation of new start-ups in the sector du jour.

Many recognize a lack of critical thinking in today’s graduates and workers, a problem consistent with the inability of VCs and entrepreneurs to effectively analyze the economics of new businesses and technologies. This crisis seems bound to worsen as we consider the historically poor academic performance of current American K–12 students. Solving this problem will require widespread changes to our educational systems, but so long as hype cycles roll on and the occasional unicorn dazzles the market, few recognize the underlying issues.

This article originally appeared in American Affairs Volume VIII, Number 4 (Winter 2024): 33–43.

Notes