You’re looking at weight the wrong way.
When referring to the rim, you are taking about rotational weight at nearly the farthest point out (aside from tire) from the bike/rider. So weight has a noticeable difference than when it’s near the low center of the bike or on the rider. The wheels, bike and rider move independently and the constant reaccelerations of MTB make it ever more important.
To experiment yourself (I have), take a 5 pound weight (bottle of sand) and move it from your bottle cage, to your seat post, then to your jersey pocket. Go for a long punchy trail ride and see if you can feel how your bike handles differently in each configuration. Multiply that by the hundreds of accelerations and bike manipulations over obstacles and out of corners. This is why some riders hate Camelbaks or refuse to put on saddle bags, because it affects how they ride or how their bike handles. Now imagine that on the wheels. Where weight IS matters in MTB expecially. Yes, a 100 grams difference per wheel is minimal and could be approaching marginal gains (when you ignore the other benifits of carbon rims), but if you’re in the front pack over a 1:30-2:00 race, that may be the difference in energy savings that give you a little more at the end. In my case I lost nearly half a pound per wheel (200 grams) when I switched to carbon rims. Also, over my 9 hour MTB 100 with 90% singletrack and 10,000 feet of climbing, I am sure my carbon rims assisted both in weight and handling.
Let’s see what science says:
An increase in speed means an increase in kinetic energy. Since the kinetic energy depends on both mass and velocity, more mass would mean more energy required to speed up.
But does it matter where this mass is located? Does it take more energy to increase speed if you put the mass on the wheel? Yes. First, let’s look at mass on the frame of the bike. If I add something to the frame the total mass increases. This means that I would need more work to increase the kinetic energy. That’s pretty straight forward.
What if the extra mass is on the wheel? In that case, I must do two things to increase speed: increase the kinetic energy and increase the rotational kinetic energy of the wheel. If all of the mass on the wheel is located at the rim, I can write the rotational kinetic energy as:
In this expression, m w is the mass of the wheel, R is the radius of the wheel and ω is the angular velocity of the wheel. But if the wheel is rolling and not slipping then there is a relationship between the angular speed of the wheel and the linear speed of the bike (this is how a car speedometer works—or at least the way it used to work).
If I substitute in for ω, I can write the following for the total kinetic energy of the bike (translational plus rotational).
In the translational kinetic energy, m b is the total mass of the bike (including the wheels) but the rotational kinetic energy only depends on the mass of the wheels.
So let’s say I add 100 grams to the frame. This would increase the value of m b but not increase the mass of the wheel. The translational kinetic energy would increase by some amount and it would require more energy to accelerate (increase the kinetic energy).
Now let’s add 100 grams to the wheel (increasing m w). Since the wheel is part of the bike, this means that the total mass also increases ( m b). Both translational and rotational kinetic energy terms will have a 100 gram increase in mass. You will have double the increase in energy by adding mass to the wheel.
So yes, adding mass to the wheel is worse than adding mass to the frame—but only when accelerating.