If the parasitic capacitance of the node between cascaded LNA’s is high, the noise contribution of the second LNA could be significant at high frequencies.
In CMOS LNA, since noise current is proportional to gm, the cascode transistor's gm is degenerated by input transistor, so I suppose that explains why it contributes less. But how about bipolar LNA? The shot noise is 2qI. How does it work over here? Thanks.