As AI, cloud computing and data centres demand more electricity, the next generation of computing may be defined not by how much power a machine can use, but by how much useful work it can deliver from every watt.
Computers have spent decades becoming faster. Processors gained more cores, chips became more powerful, and new hardware allowed machines to handle increasingly complex workloads. But the technology industry is now facing a different question, How much more performance can we get without continuously increasing energy consumption?
That question is becoming especially important as AI expands. The International Energy Agency says electricity use by data centres grew by 17% in 2025, while electricity consumption from AI-focused data centres grew even faster. At the same time, the energy used for individual AI tasks has been falling rapidly as hardware and software become more efficient.
Why Computing Efficiency Is Becoming More Important
For years, the easiest way to improve a computer was to make it more powerful. A faster processor could complete more calculations in less time, while additional processing cores allowed a system to handle more tasks simultaneously.
That approach still matters, but power consumption has become a major constraint. More powerful processors generally require more electricity and produce more heat. In large data centres, thousands of processors operating continuously can turn small improvements in efficiency into significant differences in electricity use, cooling requirements and operating costs.
This is why the industry is increasingly looking at performance per watt. Instead of asking only how quickly a computer can complete a task, engineers can also ask how much energy it needs to do that work.
The Green500, for example, ranks supercomputers according to computational performance per watt rather than simply ranking them by raw performance. Its June 2026 list shows how energy efficiency has become a distinct measure of computing capability.
That change in measurement reflects a broader shift: the fastest computer is not necessarily the most useful computer if it consumes far more energy to achieve its performance.
The Future May Be About Performance Per Watt
Energy efficiency does not mean making computers slower. The goal is almost the opposite: increase useful performance while keeping energy requirements under control.
This can happen in several ways.
Processors can be redesigned to perform particular tasks more efficiently. Software can reduce unnecessary calculations. AI models can become smaller and require less computation. Memory and data movement can be improved so processors spend less energy moving information around. Cooling systems can also become more efficient.
In May 2026, a senior TSMC executive said energy efficiency was increasingly becoming a primary consideration in chip design as customers seek performance improvements without increasing power consumption. TSMC is also exploring technologies such as advanced packaging, 3D chip stacking and photonics as alternatives to relying only on traditional transistor scaling.
This suggests that the next major improvements in computing may not come from simply pushing processors to higher speeds.
They may come from redesigning how computers perform the work in the first place.
Smaller and More Specialized Chips Could Make Computing Smarter
One way to improve efficiency is to stop asking a general-purpose processor to do everything.
A general-purpose CPU is designed to handle a huge variety of tasks. That flexibility is useful, but specialised processors can sometimes perform particular workloads much more efficiently because their hardware is designed around specific operations.
This approach is already common in AI computing. Dedicated accelerators can perform certain mathematical operations far more efficiently than a traditional processor. Similar ideas are being used in mobile devices, embedded systems and other low-power computers.
The trend is also moving toward hardware-software co-design, where the processor and the software are developed with each other in mind rather than independently.
The goal is simple: remove unnecessary work.
That could allow a device to deliver the performance it needs while consuming significantly less power.
AI Is Creating a Bigger Reason to Build Efficient Computers
AI makes this shift particularly important because modern AI systems can require enormous amounts of computation.
Training large models requires powerful computing infrastructure, while running those models for millions of users creates a continuous demand for processing. Data centres therefore have strong financial reasons to improve efficiency: every reduction in energy required for a task can matter when that task is repeated millions or billions of times.
But there is an interesting twist.
AI itself is becoming more efficient. The IEA reports that energy use per AI task has fallen dramatically as hardware and software have improved. However, new applications such as AI-generated video, advanced reasoning and agentic systems can require substantially more computation than simple text generation.
So efficiency gains are happening at the same time as demand is expanding.
That creates a race between doing each task more efficiently and doing many more tasks.
Efficiency Will Matter Beyond Data Centres
This shift is not limited to massive technology companies.
Efficiency is particularly important for devices that have limited battery capacity or operate in places where constant access to electricity is difficult.
Smartphones, wearable devices, sensors, cameras, vehicles and industrial equipment can all benefit from processors that deliver more useful computing without requiring significantly more power.
For example, an efficient processor can help a battery-powered device run longer without increasing the size of its battery. In industrial systems, lower power consumption can reduce operating costs. In remote sensors, efficiency can determine whether a device needs maintenance every few months or can operate for years.
Companies are already developing processors specifically around this idea. In February 2026, Efficient Computer announced a $60 million funding round to develop energy-efficient general-purpose processors, arguing that power, thermal limits and battery life are becoming important constraints as computing moves beyond the cloud and into physical devices.
But More Efficiency Does Not Automatically Mean Less Energy Use
There is an important limitation to this story.
Making computers more efficient does not guarantee that total electricity consumption will fall.
If computing becomes cheaper and easier to perform, people and businesses may simply use more computing.
Consider a simple example. If a task becomes ten times more energy-efficient but the number of times people perform that task increases twenty times, total energy consumption can still rise.
This is already visible in AI. The IEA says efficiency per task is improving rapidly, while demand for AI services and more energy-intensive applications is also increasing.
That means efficiency is not a single solution to computing’s energy challenge.
It is one part of a much larger equation involving hardware, software, demand, electricity generation, cooling and how technology is used.
The Computer of the Future May Not Look More Powerful
The most interesting change may be that future computers do not necessarily need to look more powerful to be more capable.
A processor that performs the same task using half the energy is an improvement even if its headline speed has barely changed. A smaller AI model that delivers an adequate result locally can be more useful than a much larger model that requires a remote data centre for every request.
The industry is therefore beginning to think beyond raw performance.
The question is becoming:
How much useful work can a computer accomplish with the resources available to it?
That could influence everything from chip architecture and memory design to software development and data-centre construction.
The future of computing may not be a race to build machines that consume more power to become faster. It may be a race to build machines that waste less power while becoming smarter, faster and more capable.
And in a world where computing demand continues to grow, that could be one of the most important upgrades of all.
