Since the dawn of venture capital, VCs have funded ideas that seemed downright absurd at the time. The first cable company competed against free, over-the-air television. FedEx competed with the US postal service. Wireless carriers sunk billions of dollars into towers and spectrum before they could sign up a single subscriber. But each of these investments paid off. And they all followed a similar pattern.

First, an entrepreneur would demonstrate that a business “worked” in a single market. And by “worked” I mean that the price, combined with consumer purchase behavior, was adequate to cover the cost of providing the service on an ongoing basis. Only then would the company expand into additional markets. This was a slow and steady process of raising capital, deploying capital, and then raising capital again. It took FedEx a decade to expand across the US. Now 45 years later the company is a global shipping powerhouse worth over $40BN.

But over the past few years this model changed. Perhaps it was due to competition from “cloners” who sprung up to copy successful businesses. Or perhaps it was due to the low interest rate environment, which has made investors less risk adverse. Either way, every idea is now the next Uber. And billions of dollars are invested into businesses before anyone knows if they really work.

To be fair, some of this might be simple decision bias. There are many businesses that work in the microcosm of San Francisco but don’t work in the rest of the world. Like Google Glass for instance. No traditional CPG company would pick San Francisco as a test market, unless it was for a computer programmer drink that tasted like pancake batter cut with formaldehyde. Yet many venture-backed products and services begin life in San Francisco. A cautious approach for any company would be to raise $5MM to test a few markets before going big, but many companies today raise $50MM or $100MM and go national in one fell swoop. Do not pass go, do not collect a traditional Series A.

While I am on the topic of going big, it is probably worth taking a closer look at Uber. To the best of my knowledge, Uber is not yet profitable in any market. I understand that Uber could be profitable if they turned off marketing, but I don’t know what that means in the context of a company that will never turn off marketing.

When I step back and look objectively at Uber I see evidence that the company may not have all the math figured out just yet. For instance, Uber recently increased the percentage it takes from drivers by 25%. That doesn’t fit the depiction of a supply-constrained marketplace awash in cash. The company’s CEO has also said that driverless cars will soon take costs out of the system, which could be a literal deux ex machina. FedEx was never dependent on future innovation to make the math work.

And don’t get me wrong – I love the Uber service. Greycroft spends tens of thousands of dollars each year on Uber because it is often cheaper than taxis. I hope it stays that way. However our Uber expense pales in comparison to the amount we spend each year on airfare, yet Uber is almost as valuable as all the US airlines combined.

The challenge with on-demand services is that, at the end of the day, they are still services. Someone has to show up to pick up your on-demand laundry, mow your on-demand lawn, paint your on-demand house, and cut your on-demand hair. The Internet is really good with businesses that have high fixed costs and low variable costs, because these businesses benefit from the scale that occurs online. However I am a little skeptical about services that have low gross margins, or in some cases negative gross margins. I have yet to see a magic panacea to solve for that.