For the past three years, the artificial intelligence race has largely been defined by chatbots. OpenAI launched ChatGPT. Google expanded Gemini. Meta developed Llama, while Microsoft embedded AI across Windows and Office. But behind the competition to build smarter models, another battle is accelerating, one that could ultimately prove more important than the software itself.
The world’s largest technology companies are now competing for computing power, investing hundreds of billions of pounds in AI data centres, specialised chips and electricity infrastructure needed to train and run increasingly powerful AI systems. Recent developments suggest computing capacity, often referred to as ‘compute’, has become one of the industry’s most valuable resources.
AI Demand Is Growing Faster Than Infrastructure
The shift has become increasingly visible over recent months. During Google’s latest earnings call, chief executive Sundar Pichai acknowledged that the company remained ‘compute-constrained in the near term’, adding that Google Cloud revenue could have been higher if sufficient infrastructure had been available to meet customer demand.
The pressure extends beyond Google’s own business. According to the Financial Times, Google reportedly limited the amount of Gemini computing capacity available to Meta after demand exceeded available resources, highlighting how even the world’s largest technology companies are beginning to compete for limited AI infrastructure.
Oracle has experienced similar demand. The company recently increased its fiscal 2027 revenue forecast to $90 billion after reporting record growth in cloud contracts driven by AI workloads. Remaining performance obligations, a measure of contracted future revenue, climbed 325% year on year to $553 billion, reflecting strong long-term demand for AI computing capacity.
Rather than competing solely through software, cloud providers are increasingly competing on how much computing infrastructure they can build and deliver.
Data Centres Have Become Strategic Assets
Behind every AI chatbot sits an enormous physical network. Training and operating large language models requires thousands of graphics processing units (GPUs), high-performance servers, networking equipment, cooling systems and reliable electricity supplies.
That has transformed data centres from relatively invisible technical facilities into strategic assets capable of determining which companies and countries can scale AI most effectively. Governments are beginning to recognise that shift.
Countries including South Korea have announced major investments in AI infrastructure, while the United States continues expanding domestic semiconductor manufacturing and cloud capacity through public and private investment.
Rather than simply attracting technology companies, large AI infrastructure projects are increasingly viewed as long-term investments in economic competitiveness, research, advanced manufacturing and digital sovereignty.
Power Could Become AI’s Biggest Constraint
The next challenge for artificial intelligence may not be building smarter models. It may be powering them. Modern AI data centres consume enormous amounts of electricity while requiring sophisticated cooling systems to keep thousands of processors operating continuously.
As AI adoption expands across industries, access to energy, land and resilient power grids is becoming almost as important as access to advanced semiconductors. That changing reality is influencing government policy as countries increasingly link AI strategies with energy security and national infrastructure planning.
The result is that AI investment is no longer confined to software companies. Utilities, construction firms, semiconductor manufacturers and cloud providers all stand to benefit from growing demand for compute.
Why Investors Are Following the Infrastructure
For investors, the biggest opportunities may increasingly lie beneath the AI applications consumers interact with every day. Companies building cloud platforms, semiconductor technologies and hyperscale data centres are positioned to benefit regardless of which chatbot ultimately attracts the largest number of users.
Oracle’s expanding cloud business, Nvidia’s AI chips and Microsoft’s continuing investment in cloud infrastructure all reflect a broader trend: the infrastructure supporting artificial intelligence is becoming just as valuable as the models themselves.
Bloomberg recently reported that Google agreed to purchase additional computing capacity through a multibillion-dollar arrangement with SpaceX, illustrating how demand for AI infrastructure is beginning to reshape investment decisions across the technology sector.
The Next Phase of the AI Race
Consumers may continue judging AI by the quality of chatbots and digital assistants. Behind the scenes, however, the industry’s competitive advantage increasingly depends on something far less visible.
The companies able to secure sufficient computing power, build resilient data centres and access reliable energy supplies may ultimately gain the greatest advantage as AI adoption accelerates worldwide. The next AI breakthrough may still arrive through software.
But the companies that lead the next decade of artificial intelligence are likely to be those that invested first in the infrastructure making those breakthroughs possible.
Leave a comment