Perplexity CEO Aravind Srinivas has suggested that smartphones could eventually render data centres obsolete, citing the potential for powerful, personalised AI tools to run on device hardware during a recent podcast appearance.

In a conversation with host Prakhar Gupta, Srinivas argued that the future of AI lies in localised processing, where devices can handle complex tasks without relying on remote data centres and cloud computing.

This shift would mark a departure from the current paradigm, where AI systems rely on transmitting data to and from large data centres, leveraging their processing power to function.

Apple has already begun exploring this approach with its Apple Intelligence system, which utilises specialised chips in the company's latest products to enable faster and more secure AI processing.

Microsoft has also incorporated on-device AI processing in its Copilot+ laptops, demonstrating the growing trend towards localised AI capabilities.

However, this technology is currently limited to high-end devices, as standard equipment often lacks the processing power required to support demanding AI applications.

According to Jonathan Evans, director of Total Data Centre Solutions, the prospect of powerful AI running on local devices is a long-term possibility, contingent on advancements in efficient and capable processing technology.

Despite the potential for disruption, the data centre industry continues to experience high demand, prompting questions about the sector's future development.

Traditionally, data centres have been large facilities housing extensive computer systems that support a wide range of digital services, including video streaming, online banking, and AI processing.

Virtually any online service relies on data centres, with major companies owning or leasing space in these facilities to support their operations.

A notable exception to the typical data centre model is a small, washing machine-sized facility in Devon, UK, which has been used to provide heat to a public swimming pool while performing computational tasks.

This innovative approach has sparked interest in alternative data centre designs, challenging the conventional wisdom that these facilities must be large and resource-intensive.

In November 2025, a British couple demonstrated another creative application of small-scale data centres, using one to heat their home from a garden shed.

Similarly, a university professor has been using a GPU under his desk to both drive AI applications and warm his office, illustrating the potential for localised, low-impact data processing.

Meanwhile, tech companies continue to invest heavily in large-scale data centre construction, with around 100 new facilities planned in the UK alone, despite concerns about their environmental impact and energy consumption.

Nvidia CEO Jensen Huang has referred to data centres as "AI factories," emphasizing their critical role in supporting the development of AI technology.

The AI sector has long adhered to the notion that increased computing power is essential for advancing AI capabilities, although this assumption is being reevaluated.

A growing number of voices in the tech sector are questioning the need for massive, remote data centres, suggesting that alternative approaches could be more effective and sustainable.

Evans proposes that smaller, edge data centres located near population centres could reduce latency and improve response times, offering a more efficient and adaptable solution.

Mark Bjornsgaard, founder of DeepGreen, agrees that smaller data centres are the future, envisioning a network of local facilities that can provide heat as a by-product of their operations.

Bjornsgaard suggests that every public building could potentially house a small data centre, creating a distributed and resilient infrastructure for AI and other applications.

He likens London to a single, undeveloped data centre, highlighting the potential for urban areas to become hubs for localised computing and AI processing.

Amanda Brock, head of OpenUK, shares this perspective, predicting that the traditional data centre model will eventually become obsolete as more efficient and sustainable alternatives emerge.

Brock advocates for repurposing derelict buildings and vacant shops into small data centres, promoting a more community-focused approach to AI and computing infrastructure.

Some companies are exploring even more innovative approaches, such as developing data centres in space to support future AI and computing needs.

Avi Shabtai, CEO of Ramon Space, sees space-based data centres as an opportunity to reimagine data structures and create more efficient, scalable, and flexible computing systems.

On a more terrestrial level, Brock and Srinivas agree that the future of processing will likely involve a shift towards handheld devices, set-top boxes, or local routers, reducing the need for large, centralised data centres.

This transition may be facilitated by the development of smaller, more targeted AI tools that require less computing power and can be more easily integrated into local devices.

The current emphasis on Large Language Models, which are powerful but often prone to errors, may give way to more bespoke AI solutions that are trained on specific datasets and designed for particular tasks.

The limitations of Large Language Models are partly due to their broad scope and tendency to make mistakes, highlighting the need for more focused and accurate AI applications.

As AI ethics campaigner Ed Newton Rex has noted, an AI tool designed for a specific purpose, such as cancer detection, does not require the same breadth of capabilities as a general-purpose AI model.

Businesses are increasingly opting for customised AI solutions, which are often more expensive but provide greater accuracy and control over data usage.

These tailored AI tools tend to perform better and require less computing power, making them more suitable for local deployment and storage.

Dr Sasha Luccioni, AI and climate lead at Hugging Face, has observed a shift towards smaller, more bespoke AI models that can run on local devices and are tailored to specific business needs.

This paradigm shift is driven by the need for more efficient, accurate, and sustainable AI solutions that can be integrated into local infrastructure.

However, the proliferation of small data centres raises concerns about national security, prompting questions about the potential risks and benefits of decentralised computing infrastructure.

According to Prof Alan Woodward, a computer security expert, smaller data centres may actually reduce the risk of large-scale security breaches, as they present less attractive targets for attackers.

Woodward notes that large data centres can be single points of failure, citing recent examples of major outages at Amazon Web Services facilities.

Luccioni adds that reducing reliance on large data centres can also have environmental benefits, as these facilities consume significant resources and contribute to energy consumption and e-waste.