Fundamental physical limits of technologies

Technologies leverage discoveries in fundamental sciences for practical purposes, enabling applications not previously attainable. Due to this foundation in fundamental sciences, all technologies have limits bounded by physical laws and mathematical frameworks. Especially with the dizzying accelerating advancements, technologies can make it seem like they can make anything possible. However, these fundamental physical limits determine the boundaries in performance of all technologies.

Identifying these boundaries is a fruitful endeavor. It enables three things:

1. It allows us to know how far we can go before we are limited by our understanding of nature.
2. It allows us to understand why our current technologies can’t perform better.
3. It allows us to understand what we can do to make our technologies better.

To delineate fundamental physical limits of technologies, we need to start with delineating technologies themselves. Information technologies can be categorized into three groups based on their application domain:

1. Computing technologies
2. Communication technologies
3. Sensing technologies

All of them are self-explanatory, but it is helpful to categorize them clearly so that we can clarify the thought processes that go into inventing new technologies in these domains. Ways these technologies handle information are also similar enough that a lot of the limits arising in pushing these technologies are coupled to one another. However, we will categorize these limits so that each limit is a dominant boundary for the way their performances are determined.

Fundamental physical limits

There are various fundamental physical limits arising from quantum field theory and general relativity. There is only so much information you can squeeze into a space before it collapses into a black hole. There is only so much you can localize a particle before intrinsic uncertainty from quantum mechanics stops you from deciding where it is. We will untangle these general bounds in fundamental physics and assign them into their corresponding technology domain.

Computing limits

Computing is concerned with the physical manipulation of information, where abstract logical operations are mapped to the state transformations of a physical system. This process involves encoding information into physical states, executing an algorithm through a controlled sequence of these transformations, and decoding the final state to obtain the computational result.

In computing, a computer’s performance can be physically defined by its computation rate, memory capacity, and energy efficiency (Lloyd, 2000). Conventional electronic computers are built from very-large-scale integrated (VLSI) circuits made from semiconductor devices. For finite resources, such as energy, space, and time, how can we maximize the computational power of such a system? One way is to scale it down. As each semiconductor device is scaled down, the energy required to control the device approaches the energy of the quantized energy levels of the corresponding quantum harmonic oscillator. At that scale, quantum-mechanical effects cannot be ignored (Feynman, 1960; Dowling & Milburn, 2003). If these semiconductor devices or unit cells are well isolated from extrinsic noise sources, such as thermal noise, and can be coherently controlled at scale, the unit cell of the computer becomes a qubit, and the conventional electronic computer becomes a quantum computer (Lloyd, 2000; Feynman, 1960; Dowling & Milburn, 2003).

The throughput of such a system is bounded by the quantum speed limit (Margolus & Levitin, 1998), the energy efficiency is unbounded for reversible computations (Bennett, 1973) and is bounded by the Landauer limit for irreversible computations (Landauer, 1961), and the memory capacity is bounded by the Bekenstein limit (Bekenstein, 1981). The quantum speed limit sets the minimum time for a quantum harmonic oscillator to evolve between two distinguishable states (Margolus & Levitin, 1998). For a potential ultimate computer made out of very-large-scale-integrated quantum harmonic oscillators, the computation rate is bounded by this evolution time, which scales with the energy of each device. What is the minimum energy required for each computation? For computations that are reversible, meaning that the inputs can be reconstructed from the outputs without any loss of information, the minimum energy required turns out to be unbounded (Lloyd, 2000; Bennett, 1973). However, many conventional algorithms perform irreversible computations, whose minimum energy is bounded by the thermodynamic entropy of the information lost during the computation. This is named the Landauer limit (Lloyd, 2000; Landauer, 1961). Finally, for such a system with finite energy and volume, what is the maximum amount of information that can be stored? As we pack more information into a given volume, the system’s entropy increases until it collapses into a black hole. The information capacity, or entropy, of such a black hole is well-defined, thereby setting the Bekenstein limit (Lloyd, 2000; Bekenstein, 1981).

Communication limits

In communications, the performance of a communication link is defined by its throughput and latency (Peterson & Davie, 2011). The throughput of a communication link is ultimately bounded by the channel capacity of the corresponding information channel. If we constrain our mathematical framework to classical probability theory, we can derive the Shannon limit to channel capacity (Shannon, 1948). If we relax some axioms of classical probability theory, we enter the domain of quantum information theory, where the Holevo limit to channel capacity arises, which is strictly higher than the Shannon limit (Holevo, 1973; Schumacher & Westmoreland, 1997; Banaszek et al., 2020). Therefore, the Holevo limit sets the fundamental limits on throughput, and the speed of light sets the fundamental limit on the latency of a communication link.

Sensing limits

In metrology, how well a sensor performs is defined by its sensitivity, given a finite number of resources, such as photons, available to probe a sample (Giovannetti et al., 2011). If we suppress the extrinsic noise sources, we are left with the intrinsic quantum noise of the probes. This quantum noise arises from Heisenberg’s uncertainty principle and, for classical probes, limits the sensitivity of a sensor to the classical limit, known as the standard quantum limit (Caves, 1981). We can surpass this classical limit by leveraging nonclassical probes. Nonclassical probes, such as entangled states, push the standard quantum limit in sensitivity toward the fundamental limit, known as the Heisenberg limit (Giovannetti et al., 2011; Bollinger et al., 1996).

Where we are and where we can go

It turns out our current understanding of nature leaves us plenty of room to improve our current technologies. A 1-kg, 1-L ultimate computer can perform ~10⁵¹ operations per second and can store ~10³¹ bits of information (Lloyd, 2000). In 2025, a state-of-the-art 1-kg, 1-L (normalized) computer can perform ~10¹⁴ operations per second and can store ~10¹¹ bits (NVIDIA, 2024). A 1-W, 1-km ultimate communication link can transmit ~10¹⁶ bits per second with ~10^-6 second latency (Giovannetti et al., 2004). In 2025, a state-of-the-art 1-W, 1-km fiber-optic link can transmit ~10¹⁰ bits per second with ~10^-5 second latency (Optical Internetworking Forum, 2025). An ultimate optical interferometer can sense ~10^-26 m of displacement, using 1 J of light at 1064 nm. In 2025, LIGO can sense ~10^-18 m of displacement, using 1 J of light at 1064 nm (Pitkin et al., 2011). Studying the fundamental limits of technologies and building proof-of-concept systems that operate at these limits allows us to understand how much our current technologies can be improved before we are limited by nature. It helps us understand why our current technologies can’t perform better and learn how to improve them. Therefore, this is an impactful pursuit that we should strive for.

How you can help

I tried to formalize these limits across three primary domains of technologies. However, there are still a lot of limits to further identify and I’d love to get anyone’s opinion who read this far. I recommend checking out this tool to play around with these limits in computing, communications, sensing and compare the ultimate versions of our technologies to the current state of the art. I also welcome any comments below the post here.