Chip designers could implement a 50 Ohm input by a bias resistor, common-gate amplifier's input transconductance, etc. But this way is a waste of sensitivity or/and power.
Discrete inductance is technically noiseless because its series resistance is very low, it has got acceptable Q factor, and same is true for integrated capacitors. But IC developers cannot implement inductance on the chip, because it could be huge in area, the chip cost would be huge too.
And the IC manufacturers also don't like bigger external component numbers, because cheaper applications are better, so they try to minimalize the components, and for input matching usually a simple LC network is enough.
If you build an LC matching network from external/discrete inductance and internal capacitor (cheaper actually then use discrete capacitor) and you match the antenna with a bigger resistive input impedance on the chip then the LC network will gain/boost your RF signal. Without adding noise to it! It is at the very begin of the receiver chain, so it will determine the system's noise figure.
This is why a simple inductor is so elegant. Good for the manufacturer, enough for the customer.
10nH btw comes from the tolerance of the cheap SMD inductors, the GPS center frequency and the GPS bandwidth.