This paper presents a multistage adaptive robust generation expansion planning model, which accounts for short-term unit commitment and ramping constraints, considers multi-period and multi-regional planning, and maintains the integer representation of generation units. The uncertainty of electricity demand and renewable power generation is taken into account through bounded intervals, with parameters that permit control over the level of conservatism of the solution. The multistage robust optimization model allows the sequential representation of uncertainty realization as they are revealed over time. It also guarantees the non-anticipativity of future uncertainty realizations at the time of decision-making, which is the case in practical real-world applications, as opposed to two-stage robust and stochastic models. To render the resulting multistage robust problem tractable, decision rules are employed to cast the uncertainty-based model into an equivalent mixed integer linear (MILP) problem. The re-formulated MILP problem, while tractable, is computationally prohibitive even for moderately sized systems. We, thus, propose a solution method relying on the reduction of the information basis of the decision rules employed in the model, and validate its adequacy to efficiently solve the problem. The importance of considering multistage robust frameworks for accounting for net-load uncertainties in generation expansion planning is illustrated, particularly under a high share of renewable energy penetration. A number of renewable penetration scenarios and uncertainty levels are considered for a case study covering future generation expansion planning in Europe. The results confirm the effectiveness of the proposed approach in coping with multifold operational uncertainties and for deriving adequate generation investment decisions. Moreover, the quality of the solutions obtained and the computational performance of the proposed solution method is shown to be suitable for practical policy-making generation expansion planning problems, seeking to evaluate the impact of uncertainty on future system-wide performance.