WE’RE ALL PAYING THE COST OF EXPANDING AI

A surge in demand in one area can have far-reaching effects across seemingly unrelated markets, and artificial intelligence highlights just how interconnected modern technology has become. While AI is often discussed as a software revolution, its growth depends on a physical infrastructure that consumes energy, materials, and specialised components at an unprecedented scale.

One of the most important components is memory.

Whether we are taking photos, streaming video or running an AI-powered app, memory allows our devices to process information quickly. Memory is also vital for the massive data centres—essentially warehouses filled with servers—that depend on vast quantities of high-performance memory to train and run AI models. These two worlds, personal devices and industrial-scale computing, are now competing for the same essential resources.

AI systems require vastly more memory per unit than consumer devices and supply chains are increasingly prioritising the needs of AI infrastructure rather than consumer electronics. As production capacity is redirected toward AI workloads, the Dynamic Random Access Memory (DRAM) and the NAND flash memory used in smartphones and PCs are becoming both scarcer and more expensive.

A physical challenge

Fabrication plants cannot maximise output across all chip types simultaneously, so every production decision carries a trade-off. As more capacity shifts toward AI memory, supply tightens for consumer electronics.

With memory manufacturers reallocating production toward high-bandwidth memory (HBM)—the specialised chips required for AI systems—International Data Corporation (IDC) has warned that the industry is facing a “structural” supply imbalance driven by AI demand, rather than a cyclical shortage. Meanwhile, Counterpoint Research has revised its 2026 global smartphone shipment forecast downward by 2.1%, citing rising memory costs.

According to IDC, in Q1 2026, global smartphone shipments fell 4–6% year-on-year as DRAM and NAND shortages disrupted supply and increased costs for original equipment manufacturers. Premium vendors with stronger supply chains are holding their ground, while volume players are being squeezed. IDC reports that price increases in some markets have reached as much as 40–50% as component shortages feed through to retail pricing.

Environmental considerations

The competition for memory is not only an economic issue; it also carries environmental consequences. Every new semiconductor fabrication plant requires enormous quantities of energy, water, and raw materials. Expanding memory production to meet AI demand is therefore not simply a matter of building more factories. It means increasing resource consumption across the entire technology supply chain.

AI data centres themselves are highly resource-intensive. Training and operating large AI models requires vast amounts of electricity, cooling infrastructure, and specialised hardware. As demand for high-bandwidth memory grows, manufacturers are investing heavily in new production capacity, creating additional environmental pressures associated with semiconductor manufacturing.

At the same time, rising memory costs may alter consumer behaviour. If devices become more expensive or receive fewer hardware upgrades from one generation to the next, consumers may hold onto smartphones and PCs for longer. Longer device lifespans can reduce electronic waste and lower the environmental impact associated with frequent replacement cycles. However, this benefit may be offset if manufacturers respond by accelerating cloud-based services that depend on energy-intensive data centres.

The result is a complex trade-off. AI promises efficiency gains across many sectors, but those gains depend on a physical infrastructure that consumes significant resources. The growing competition for memory illustrates a broader reality: digital technologies are not weightless. Every advance in artificial intelligence relies on materials, energy, and manufacturing capacity that must be shared across the wider economy.

As policymakers and industry leaders consider how to support continued AI growth, questions of sustainability will become increasingly important. The challenge is not only to allocate memory efficiently, but to ensure that the environmental costs of producing and using that memory remain manageable. Balancing innovation, accessibility, and resource consumption may prove just as important as improving computational performance itself.

Intelligence costs

Memory already represents a significant share of device costs—around 15–20% of the bill of materials for mid-range smartphones—and that share is increasing as prices surge. In some cases, component costs in low-end smartphones have risen by 20–30%, forcing manufacturers to either pass costs on to consumers or reduce specifications.

PC manufacturers have already begun adjusting their pricing strategies, and smartphones are expected to follow. Analysts anticipate increases in bill-of-material costs of 15% or more. Mid-range devices may ship with less RAM, entry-level models could lose features, and flagship devices may no longer advance at the same pace.

Even the largest buyers are not insulated. Companies with scale can manage volatility through long-term agreements and purchasing power, but they remain dependent on the same constrained supply chain. Over time, those limits will become unavoidable.

For smaller Android manufacturers, the situation is more acute. They face difficult trade-offs between raising prices, reducing specifications, or withdrawing from lower-margin segments altogether.

IDC notes that vendors are already responding by reducing baseline configurations: devices that might have shipped with 12GB of RAM are now launching with 8GB at the same price point.

Impacting the ecosystem

Developers are being forced to reconsider long-standing assumptions about performance progression. Annual gains in speed and capacity can no longer be taken for granted. Memory limits may constrain application design, and storage capacities could stagnate or even decline in certain segments.

Cloud-based AI will remain central to the user experience. While on-device intelligence offers advantages such as reduced latency and improved privacy, it is also resource-intensive. For most users, AI processing will continue to rely heavily on remote infrastructure. APIs, messaging platforms, and cloud services will therefore remain critical components of the ecosystem.

At the same time, market power is likely to consolidate further. Companies that control both hardware and supply chains will strengthen their positions, while smaller manufacturers face growing difficulty competing effectively.

The expansion of AI is not only advancing technological capabilities, it is also transforming the economic foundations of the hardware industry.

Lessons for industry and consumers

For entrepreneurs and developers, the implications are increasingly clear. The assumption of ever-improving device capabilities is no longer reliable.

Applications must be designed to operate efficiently within tighter memory constraints or offload processing to the cloud. Scarcity is becoming a defining feature of the environment, requiring flexibility in both product design and business models.

At the same time, pricing dynamics are shifting. IDC and Counterpoint both point to rising average selling prices alongside declining unit volumes – a combination that signals a more constrained, less elastic market.

AI is effectively introducing a new cost layer across consumer electronics. Smartphones will not automatically become more powerful or more affordable with each generation.

Experiences will continue to grow more intelligent, but the devices that deliver them may evolve more slowly – leaner in design, more constrained in capability, and, in many cases, equipped with less memory than users have come to expect.

Consumers need to take note that the devices in their pockets are part of a much larger ecosystem, one where shifts in industrial strategy and global demand can shape everyday experiences.

In the coming years, the challenge will not simply be producing more memory but distributing it in a way that supports innovation, accessibility, and sustainability. As AI continues to expand, the industry will need to balance the benefits of increasingly powerful systems against the environmental and resource demands required to support them. Finding that balance will be crucial—not just for technology companies, but for consumers, policymakers, and the broader economy that depends on digital infrastructure.

 

FACT BOX: THE ENVIRONMENTAL COST OF DIGITAL INFRASTRUCTURE

• Data centres consumed an estimated 415 terawatt-hours (TWh) of electricity in 2024—around 1.5% of global electricity demand. The International Energy Agency (IEA) expects this to rise to approximately 945 TWh by 2030 as AI adoption accelerates.

• AI systems are among the fastest-growing sources of data-centre electricity demand, requiring large quantities of specialised processors, memory, storage, and cooling infrastructure.

• Modern data centres rely heavily on water for cooling. United Nations researchers estimate that global AI-driven growth could significantly increase water consumption by 2030, placing additional pressure on local water supplies in some regions.

• Data centres and telecommunications networks together account for roughly 1% of global energy-related greenhouse-gas emissions, according to the IEA.

• Cooling systems can account for roughly one-third of a data centre’s energy consumption, depending on facility design and efficiency.

• Cloud computing can improve efficiency by consolidating workloads into large-scale facilities, but the rapid growth of AI is increasing overall demand for computing resources and electricity.

• Telecommunications networks face their own sustainability challenge. Growing mobile data traffic, 5G deployment, and cloud-based services continue to increase energy requirements across digital infrastructure.

• Extending the lifespan of smartphones, PCs, and network equipment can reduce electronic waste and lower demand for the energy- and resource-intensive manufacturing processes required to produce new devices.

• Industry analysts increasingly view energy availability, water access, and semiconductor supply as strategic constraints on future AI expansion, alongside computing power itself.

Sources:
International Energy Agency (Energy and AI, 2025)
International Energy Agency (Data Centres and Data Transmission Networks)
United Nations University Institute for Water, Environment and Health (2026).

iea.org/reports/energy-and-ai/energy-demand-from-ai
arxiv.org
iea.org/energy-system/digitalisation/data-centres-and-data-transmission-networks
reuters.com/business/energy/ai-double-data-centre-power-water-consumption-by-2030
iea.org/reports/energy-and-ai

 

ABOUT THE AUTHOR
Dario Betti is CEO of MEF (Mobile Ecosystem Forum) a global trade body established in 2000 and headquartered in the UK with members across the world. As the independent voice of the mobile ecosystem, MEF focuses on cross-industry best practices, anti-fraud and monetisation. The Forum provides its members with global and cross-sector platforms for networking, collaboration and advancing industry solutions.

Web: mobileecosystemforum.com      X: mef      LinkedIn: mobile-ecosystem-forum      Facebook: MobileEcosystemForum