From a Penn State epidemiologist (note that the quote below speaks of the "infection fatality rate", which is different from the "case fatality rate", the difference is explained in the article)[1]:
"Scientists working at the London School of Hygiene and Tropical Medicine, Imperial College London and the Institute for Disease Modeling have used these approaches to estimate the infection fatality rate. Currently, these estimates range from 0.5% to 0.94% indicating that COVID-19 is about 10 to 20 times as deadly as seasonal influenza. Evidence coming in from genomics and large-scale testing of fevers is consistent with these conclusions. The only potentially good news is that the epidemic in Korea may ultimately show a lower CFR than the epidemic in China.
...
"On balance, it is reasonable to guess that COVID-19 will infect as many Americans over the next year as influenza does in a typical winter -- somewhere between 25 million and 115 million. Maybe a bit more if the virus turns out to be more contagious than we thought. Maybe a bit less if we put restrictions in place that minimize our travel and our social and professional contacts.
"The bad news is, of course, that these infection numbers translate to 350,000 to 660,000 people dying in the U.S., with an uncertainty range that goes from 50,000 deaths to 5 million deaths. The good news is that this is not a weather forecast. The size of the epidemic, i.e., the total number of infections, is something we can reduce if we decrease our contact patterns and improve our hygiene. If the total number of infections decreases, the total number of deaths will also decrease."
This analysis is a little old, however (from March 9th, six days ago), and assumes that the rate of infection of COVID-19 is comparable to seasonal influenza. A recent This Week in Virology podcast[2] had another epidemiologist, Ralph Baric, of the University of North Carolina Chapel Hill, who had this to say:
"The R0, or the number of people who become infected from an infected individual, is estimated somewhere between 2.5 and 3.2. That's actually quite high. That means that for every case there is two and a half to three additional cases. Contemporary flu is much closer to 2 or 1.8, 1.6. So this is very explosive spread and rapid transmission. It's even more explosive because you have rare individuals or maybe not so rare individuals who are superspreaders, who can infect 15, 20 people just passing through a room."[3]
"In Canada there was one example of a superspreader who simply walked through an emergency room that was packed that was fairly packed with individuals and infected 19 people in the less than the 15 seconds they were in the emergency room as they walked through it."[4]
With such explosive growth, the health care systems stands a very good chance of being overwhelmed, causing substantially more deaths. For instance, those 10% to 17% of severe cases that might have otherwise survived had they gotten care in an ICU could well die because they can't get that care. In addition many other, non-infected people who need critical care may not be able to get that care either and die, raising the overall death toll even higher.
[1] - https://www.marketwatch.com/story/why-this-epidemiologist-is...
[2] - http://www.microbe.tv/twiv/twiv-591/
[3] - about 8 minutes in to the program
[4] - about 38 minutes in to the program