A typical jaguar Waymo usually carries up to 4 people (usually just one), weighs 2.1 tonnes and is 5.2m long.
So the double decker bus should be more environmentally efficient with maybe 5 passengers making independent journeys, though operational costs for paying a driver probably make the Waymo more financially efficient in that situation.
I suspect robotaxis will become very important in suburban areas with distributed jobs where buses don’t work well. In busy cities there simply isn’t the roadspace to replace buses with robotaxis (or any other form of car) for commuting.
It will depend on the area. Where I am in San Francisco we have articulated buses designed for 100+ people running every few minutes above BART. (We probably need more BART stations.) There’s simply not enough road space for smaller buses (let alone cars) to carry commuter traffic downtown.
(I’ve never worked downtown though so always drove to work in desolate Silicon Valley office parks.)
Even in my smaller UK home city (250k people) you need double decker 70+ capacity buses running every 15 minutes on two routes from city centre to university to handle demand because the jobs are not so distributed.
A more reasonable taxi for single riders is going to weigh around 1.5 tons (e.g. the prototype Tesla Robocab).
You also incorrectly assume that the efficiency depends linearly on weight. It doesn't, it depends more on aerodynamics, speed, and the number of stops. London buses (11 tons) average around 2.2 kWh/mi, Tesla Model 3 is at 0.21 kWh/mi while weighing 1.8 tons.