Where's the data from? An open source model based on standard methods predicts a peak of 200,000-600,000 critically ill in the US, and 350,000-1.1 million either severely or critically ill (depending on the degree of mitigation): https://neherlab.org/covid19/
This is a totally unrealistic normal distribution until the end of the year. You get somewhat similar values in the simulator by assuming icu duration of 30 days, 14 days infectivity, and R0 of 3 (i.e. no mitigation).