Tensor Flow data that goes into the unit

In the Week 1 of the Advanced learning specialization Topic => Data in TensorFlow
In the Coffee Roasting example , the input (x) has values [[200 ,17]] , that temperature 200 and 17 minutes as input .
The layer 1 uses units =3 . What data would go into each unit . If each unit will have input as [[200 17]] , and use same activation as sigmoid function , how will they produce different output [0.2, 0.7, 0.3].

Am i missing something here ?? Is my understanding incorrect

All the data goes to every unit and as every unit has different weights and biases (parameters), they produce different outputs.

One more point to note here is that, at the start, we initialize the parameters to the same values but then they go change on every iteration.