I surfed a bit on the Ceramicspeed website. Lots of BS there. They produce no hard data on their findings only pictures and graphs without units or reference data. Pure marketing nonsense.
They make claims such as 9% improvement but don't say over what. They don't say whether those bearings are hybrids or pure ceramics although they hint at them being hybrid steel/ceramic bearings. They claim the bearings are hand made. Bearings are not tapestries, they're best made and assembled by machines.
They make claims such as 100% smoother, x-much harder etc etc. They claim their ceramic (Silicon nitride) balls are harder than other balls made from the same material. It goes on and on.
9% improvement: It is easy to achieve such a figure. You don't state the load under which the bearing works, you don't state the base reference point which could be a tricycle bearing made from square balls.
Hybrid bearings have steel races and ceramic balls. The ceramic is much harder than the ball and easily damages the race. Although the balls survive, the races, once damaged renders the bearing useless. Ceramic bearings have low durability in real-life bicycle use.
Bearing smoothness is measured in microns and expressed as Class. A class 100 bearing has imperfections that are no larger than 100 microns. A class 20 bearing has imperfections no more than 20 microns and so on. A good steel bearing is typically class 10 and an average ceramic bearing is class 5. This company claims its bearings are 100% smoother than the competition i.e. Class 2.5. Nonsense. It's own claim is Class 3, so why lie elsewhere on the website?
Generic bicycle bearing balls are typically class 100 and super smooth Shimano balls are class. Campagnolo balls are class 10. As you go smoother, the returns diminish but cost escalates exponentially.
Ceramic bearings were designed for fast-spinning applications such as turbo chargers which reach 100 000 rpm. At such speeds the centrilfugal force of the balls deform them and they start to lose contact with the races and then self-destruct. Lighter balls solve the problem. Turbos also generate very high heat and the ceramic copes with that. However, bicycles don't go that fast and class 20 or 10 is just fine. I cannot picture a scenario where replacing all the bearings in a quality groupset with ceramic ones will save 9 Watts. Without data I will continue to disbelieve it.
As someone pointed out, the power required to keep a body moving through air goes up by the square of the speed due to wind resistance. At about 40 kph the slightest aero adjustment to the rider will save a minute and many adjustments and mods, aero helmets and clothing saves many more minutes. At higher speed the savings are even more. These high speed savings are only applicable to solo time-triallists, not pack riders who mostly hide from the wind. For the price of a bearing replacement you can afford a good time trial coach and bike setup and buy an aero suit.
The argument that 2 seconds can make or break the race is only valid if the claims made by the bearing companies are correct but I've seen their little pseudo experiments at trade shows and if you know what to look for, it is clear that they are rigged.
Fortunately for companies like this, there is no species of human more superstitious than the professional or serious amateur athlete.