Moore’s law: the famous rule of computing has reached the end of the road, so what comes next?

The semiconductor industry has long operated under a principle that seemed almost magical in its consistency: computing power would double at regular intervals, costs would plummet, and devices would shrink to ever more compact forms. For decades, this phenomenon shaped technological progress, driving innovations from personal computers to smartphones, from scientific research to artificial intelligence. Yet recent developments have confirmed what engineers and researchers have suspected for years: the fundamental physical barriers that govern transistor miniaturisation have finally imposed their will, marking the end of an era that defined modern computing.

Understanding Moore’s Law: history and key principles

The origins of a revolutionary observation

The principle that would come to define the computing industry emerged from a remarkably straightforward observation. The co-founder of Intel noticed a consistent pattern in semiconductor development: the number of transistors that could be placed on an integrated circuit was doubling approximately every two years. This observation, made in the mid-1960s, would prove prophetic for half a century, guiding research and development priorities across the entire technology sector.

What began as an empirical observation quickly became a self-fulfilling prophecy. Manufacturers used this principle as a roadmap, investing billions to ensure that each new generation of chips would meet these expectations. The result was unprecedented technological acceleration that transformed every aspect of modern life.

The mechanics behind exponential growth

The doubling of transistor density relied on several interconnected factors:

• Advances in photolithography techniques, allowing engineers to etch increasingly fine patterns onto silicon wafers
• Improvements in materials science, enabling better insulation and conductivity at microscopic scales
• Innovations in chip architecture, maximising the efficiency of available space
• Economies of scale that made mass production increasingly cost-effective

These improvements created a virtuous cycle. As transistors became smaller, chips became faster, cheaper, and more energy-efficient. This enabled new applications, which in turn justified further investment in miniaturisation technologies.

Widespread impact across industries

The consistent improvements predicted by this principle revolutionised numerous fields. Scientific simulations that once required room-sized supercomputers became possible on desktop machines. Weather forecasting models achieved unprecedented accuracy through enhanced computational capacity. Perhaps most significantly, the rise of machine learning systems depended entirely on the availability of powerful, affordable processors capable of handling vast datasets.

Yet as the computing industry approaches the physical limits of silicon-based technology, questions about the sustainability of this trajectory have moved from theoretical speculation to practical concern.

The limits reached by Moore’s Law

Physical barriers at the atomic scale

The fundamental challenge facing continued miniaturisation is straightforward: transistors are approaching the size of individual atoms. At these scales, the classical physics that governed earlier generations of chips gives way to quantum effects that make reliable operation increasingly difficult. Electrons can tunnel through barriers that should contain them, creating leakage currents that waste energy and cause errors.

Diminishing returns on investment

Even when engineers manage to shrink transistors further, the benefits are no longer proportional to the effort required. Each new process node demands exponentially greater investment in research, development, and manufacturing equipment, whilst delivering increasingly modest improvements in performance and efficiency. The cost of building fabrication facilities capable of producing cutting-edge chips has soared into the tens of billions, limiting the number of companies capable of competing at the technological frontier.

Heat and power constraints

As transistor density increases, so does heat generation. Modern processors already operate at temperatures that require sophisticated cooling solutions, and further increases in density would push thermal management beyond practical limits. Power consumption has similarly become a constraining factor, particularly for mobile devices and data centres where energy costs represent a significant operational expense.

These physical and economic realities have forced the industry to acknowledge that traditional scaling cannot continue indefinitely, prompting a fundamental rethinking of how computing advancement will proceed.

Consequences of the end of Moore’s Law

Slowdown in general-purpose performance gains

The most immediate consequence is a marked deceleration in the performance improvements that consumers and businesses have come to expect. General-purpose processors are no longer becoming dramatically faster with each generation, as gains from miniaturisation have largely plateaued. This represents a significant shift for an industry built on the assumption of continuous, exponential improvement.

Economic implications for the semiconductor industry

The economics of chip manufacturing have fundamentally changed. The massive investments required for marginal improvements have led to industry consolidation, with only a handful of companies capable of producing state-of-the-art processors. This concentration raises concerns about:

• Supply chain resilience and geopolitical dependencies
• Innovation pace as competitive pressure diminishes
• Pricing power and accessibility of advanced computing resources
• Barriers to entry for new competitors and disruptive technologies

Shift in research and development priorities

With traditional scaling reaching its limits, the focus of innovation has necessarily shifted. Rather than simply making transistors smaller, engineers are exploring entirely new approaches to computing. This reorientation affects everything from university research programmes to corporate investment strategies, as the industry searches for alternative paths to continued progress.

This fundamental change in the technological landscape has accelerated the development and adoption of novel computing paradigms and architectures.

New technologies and emerging innovations

Advanced materials beyond silicon

Researchers are investigating materials that could overcome silicon’s limitations. Graphene, with its exceptional electrical properties, offers potential for faster, more efficient transistors. Similarly, carbon nanotubes and other exotic materials promise performance characteristics impossible with conventional semiconductors, though practical manufacturing at scale remains challenging.

Innovative transistor designs

Even within silicon-based technology, new architectural approaches are extending performance gains:

• Three-dimensional chip stacking, which increases density by building upwards rather than shrinking laterally
• Gate-all-around transistors that provide better control over electron flow
• Neuromorphic designs that mimic biological neural networks for specific applications
• Photonic interconnects that use light rather than electricity for data transmission

Specialised processors for targeted workloads

Perhaps the most significant trend is the move towards application-specific integrated circuits optimised for particular tasks. Graphics processing units revolutionised parallel computing for gaming and later proved ideal for machine learning. Tensor processing units take this specialisation further, designed explicitly for artificial intelligence workloads. This approach sacrifices general-purpose flexibility for dramatic performance improvements in specific domains.

Whilst these innovations represent significant progress, they operate within the framework of classical computing, whereas the most radical departure from traditional architectures lies in an entirely different computational paradigm.

The advent of quantum computing

Fundamentally different computational principles

Quantum computers operate on principles that bear little resemblance to classical computing. Rather than processing information as binary bits, quantum systems use qubits that can exist in superposition states, simultaneously representing multiple values. This property, combined with quantum entanglement, enables certain calculations to be performed exponentially faster than any classical computer could achieve.

Current state and practical limitations

Despite tremendous progress, quantum computing remains largely experimental. The systems require extreme conditions to operate: temperatures near absolute zero, isolation from environmental interference, and sophisticated error correction. Maintaining quantum coherence long enough to perform useful calculations represents a formidable engineering challenge that has yet to be fully solved.

Current quantum computers excel at specific problems:

• Cryptographic calculations and security applications
• Molecular simulation for drug discovery and materials science
• Optimisation problems with vast solution spaces
• Certain machine learning and pattern recognition tasks

Integration with classical systems

The future likely involves hybrid architectures that combine quantum and classical processors, leveraging each for their respective strengths. Quantum systems would handle specific computational tasks beyond classical capabilities, whilst traditional processors manage general operations and interface with users and applications. This complementary approach reflects the broader trend towards heterogeneous computing environments.

These developments, alongside other innovations, are reshaping expectations for how the computing industry will evolve in the coming decades.

The future of the computing industry after Moore’s Law

Emphasis on energy efficiency and sustainability

With performance gains from miniaturisation diminishing, energy efficiency has become a primary design objective. Data centres already consume significant portions of global electricity production, and the environmental impact of computing infrastructure is under increasing scrutiny. Future advances will likely prioritise reducing power consumption and heat generation, potentially through novel cooling technologies, more efficient architectures, and renewable energy integration.

Software optimisation and algorithmic innovation

As hardware improvements slow, software efficiency gains increasing importance. Better algorithms, more effective use of available processing resources, and optimisation for specific hardware configurations can deliver performance improvements that complement or exceed hardware advances. This shift places renewed emphasis on programming skills and computational thinking.

Diverse architectural approaches

The post-Moore’s Law era will likely be characterised by diversity rather than uniformity. Different applications will employ different computing paradigms:

• Traditional CPUs for general-purpose tasks requiring flexibility
• Specialised accelerators for graphics, artificial intelligence, and specific workloads
• Quantum processors for particular classes of problems
• Neuromorphic chips for pattern recognition and adaptive systems
• Photonic computing for high-speed data processing and communications

This heterogeneous landscape represents both challenge and opportunity, requiring new approaches to system design, programming, and resource management.

The conclusion of an era defined by predictable, exponential growth marks not an endpoint but a transformation. The computing industry faces a future characterised by innovation across multiple dimensions: novel materials and manufacturing techniques, architectural diversity, specialised processors, and entirely new computational paradigms. Whilst the convenient predictability of doubling transistor counts every two years has ended, the imperative for technological advancement remains. The transition from a single, dominant scaling law to multiple overlapping strategies reflects the maturation of the field, demanding creativity and adaptability from engineers, researchers, and businesses alike. The decades ahead promise continued progress, albeit through paths less straightforward than those that brought us to this juncture.