In the early days of 5G research, Massive MIMO (Multiple-Input Multiple-Output) was often presented as a "free lunch" for energy efficiency. The logic was simple: by using hundreds of antennas to focus signals into narrow beams, we could theoretically reduce transmit power to near zero while maintaining high data rates.
However, as researchers like Emil Björnson, Jakob Hoydis, and Mérouane Debbah pointed out in Designing Multi-User MIMO for Energy Efficiency: When is Massive MIMO the Answer?, this perspective ignored the "cost of the box." Every antenna added to a base station requires its own radio frequency (RF) chain, including mixers, power amplifiers, and digital-to-analog converters. This hardware consumes a significant amount of "circuit power" that remains constant regardless of how much signal is actually transmitted.
The Concave Nature of Efficiency
The fundamental insight across this research area is that energy efficiency (EE)—measured in bits per Joule—is not a monotonic function of the number of antennas. Instead, it follows a bell-shaped or concave curve.
Initially, adding antennas provides massive gains. The beamforming becomes sharper, allowing the system to serve more users simultaneously and reducing the required transmit power per bit. During this phase, the spectral efficiency gains far outweigh the incremental power cost of the new hardware.
However, as shown by David López-Pérez and Nicola Piovesan in their comprehensive survey, there is a saturation point. Eventually, the static energy consumed by the electronic circuitry of the 101st or 200th antenna exceeds the marginal improvement in data rate it provides. Beyond this "sweet spot," adding more hardware actually makes the network less energy-efficient.
Balancing Transmit and Circuit Power
To find this optimal point, researchers had to move beyond idealized models. Emil Björnson and Luca Sanguinetti demonstrated in Optimal Design of Energy-Efficient Multi-User MIMO Systems that the most efficient way to operate a massive array is counter-intuitive: you should actually increase transmit power as you add antennas.
Why? Because if you have already "paid" the high energy price to turn on hundreds of transceiver chains, you must ensure the data rate is high enough to justify that overhead. If the transmit power is too low, the fixed circuit power dominates the denominator of the efficiency equation, dragging the bits-per-Joule down.
Diminishing Returns in Reconfigurable Surfaces
This principle of diminishing returns isn't limited to active antenna arrays. It also applies to Reconfigurable Intelligent Surfaces (RIS)—passive or semi-passive sheets of elements that reflect signals toward users.
In Large Intelligent Surfaces for Energy Efficiency in Wireless Communication, Chongwen Huang and Alessio Zappone found a similar trade-off. While more reflecting elements improve the signal-to-noise ratio, each element adds to the system's static power dissipation. Similarly, Wanli Ni and Yuanwei Liu noted that the computational complexity of managing these elements eventually hits a wall, where the overhead of calculating the optimal phase shifts outweighs the signal gains (STAR-RIS Enabled Heterogeneous Networks).
Hardware Impairments as a Bottleneck
The "sweet spot" is also moved by the quality of the hardware itself. In Massive MIMO Systems with Non-Ideal Hardware, Emil Björnson and Marios Kountouris explored how low-cost components introduce distortions.
They discovered an "averaging out" effect: while hardware impairments usually impose a hard limit on capacity, the sheer number of antennas in a Massive MIMO system allows the base station's distortions to cancel each other out. This suggests that Massive MIMO is actually a tool for enabling the use of cheaper, less power-hungry hardware, provided we find the right balance between the number of components and their individual quality.
Adaptive Strategies for Real-World Deployment
Because the optimal number of antennas depends on the current traffic and user density, static configurations are rarely efficient. This has led to the development of "Adaptive RF Chain Activation" (ARFA) schemes.
As Nhan Thanh Nguyen and Kyungchun Lee argue in Hybrid Beamforming and Adaptive RF Chain Activation, the most energy-efficient systems are those that can dynamically turn off unnecessary hardware. By activating only the subset of RF chains needed to meet current demand, networks can stay at the peak of the energy efficiency curve even as user conditions change.
This shift from "always-on" to "always-available" architecture, supported by the findings of Serveh Shalmashi regarding D2D underlays (Energy efficiency and sum rate when massive MIMO meets device-to-device communication), represents the modern consensus: Massive MIMO is the key to green networks, but only if we respect the physical and economic costs of the hardware that powers it.