Bain & Company’s fifth annual Global Technology Report finds larger data centres could drive costs to between US$10 billion and US$25 billion in five years
The market for AI-related hardware and software is expected to grow between 40% and 55% annually, reaching between US$780 billion and US$990 billion by 2027, according to new research released by Bain & Company.
The fifth annual Global Technology Report provides insights on the new waves of growth in the technology sector because of disruptions from the fast-changing AI advancements.
Three areas of opportunities – bigger models and larger data centres, enterprise and sovereign AI initiatives, and software efficiency and capabilities – could enable the AI hardware and software market to come close to a trillion-dollar industry in the next three years.
“Generative AI is the prime mover of the current wave of change, but it is complicated by post-globalisation shifts and the need to adapt business processes to deliver value,” said David Crawford, Chairman of Bain’s Global Technology Practice. “Companies are moving beyond the experimentation phase and are beginning to scale generative AI across the enterprise. As they do, CIOs will need to maintain production-grade AI solutions that will enable companies to adapt to a landscape that is quickly shifting. Essentially, they need to adopt an ‘AI everywhere’ approach.”
As AI scales, so will data centres: Industry could face next wave of chip shortage
AI workloads could grow 25% to 35% per year through 2027, Bain estimates. As AI scales, the need for computing power will radically expand the scale of large data centres over the next five to 10 years. AI will spur growth in data centres, from today’s 50–200MW to more than 1GW, Bain reports.
This means that if large data centres cost between US$1 billion and US$4 billion today, they could cost between US$10 billion and US$25 billion five years from now.
These changes are expected to have huge implications on the ecosystems that support data centers including infrastructure engineering, power production and cooling, as well as strain supply chains.
In addition to the need for more data centres, the AI-driven surge in demand for Graphics Processing Units (GPUs) could increase total demand for certain upstream components by 30% or more by 2026, Bain predicts.
Just as the pandemic created a surge in PC demand, surging demand for AI computing power will strain supply chains for data centre chips, personal computers and smartphones.
These trends, when paired with geopolitical tensions, could trigger the next shortage of semiconductors, Bain warns. If data centre demand for current-generation GPUs were to double by 2026, not only would suppliers of key components need to increase their output, but makers of chip packaging components would need to nearly triple their production capacity to keep up with demand.
Emergence of sovereign AI presents both challenges and opportunities
Another area that Bain says will add an additional layer of complexity for technology companies is the emergence of ‘sovereign’ AI blocs. The post-globalisation movement in technology is spreading from the pandemic-era chip shortage to current data, security and AI privacy concerns. Governments worldwide – including Canada, France, India, Japan and the United Arab Emirates – are spending billions of dollars to subsidise sovereign AI.
They’re investing in domestic computing infrastructure and AI models developed within their borders and trained on local data. As the sovereign AI push picks up steam, those who emerge as leaders will be based on several determining factors.
“Establishing successful sovereign AI ecosystems will be time-consuming and incredibly expensive,” said Anne Hoecker, Head of Bain’s Global Technology Practice. “While less complex in some ways than building semiconductor fabs, these projects require more than securing local subsidies. Hyperscalers and other big tech firms may continue to invest in localised AI operations that will ensure significant competitive advantages.”
Similarly, as enterprises face rising challenges in managing suppliers, protecting data and controlling total cost of ownership, small language models with algorithms that use RAG (retrieval-augmented generation) and vector embeddings (numeric representations of data) could see demand increase as these handle a lot of the computing, networking and storage tasks close to where the data is stored.
More efficient software development needed to drive value
The arrival of Generative AI has added pressure on software development companies to demonstrate greater efficiency. Generative AI appears to save about 10% to 15% of total software engineering time, according to Bain’s survey of more than 200 companies from across industries. However, most companies aren’t making the most of these savings, Bain found.
“When implemented properly, generative AI could result in efficiency gains of 30% or more,” said Roy Singh, Global Head of Bain’s Advanced Analytics Practice. “Using Generative AI to achieve meaningful improvements in software development is possible but requires efforts that stretch beyond the introduction of coding assistants.
“When it comes to AI deployment, engineering teams should drive end-to-end efficiencies by incorporating other advanced techniques such as static analysis and covering the full software development lifecycle including product management, refactoring, code reviews, testing and build/release management,” added Singh.
The above pressures come as software companies see slows in revenue growth. The median annual revenue growth for a group of about 90 publicly traded Software-as-a-Service (SaaS) companies declined by 16% in the last two years, Bain’s analysis shows. As growth slowed, SaaS companies significantly scaled back spending on sales and marketing, while spending on R&D has proved more robust.
Software companies’ sales and marketing budgets have shrunk from 41% of revenue in 2022 to 33% of revenue in 2024 while spending on R&D shrunk by just 3% – declining from 21% to 18% – of revenue during the same period.
Software companies will need to ensure they’re producing what customers need, make the most of their R&D spend, and rein in inflating operating expenses. Software vendors, on the other hand, should be more disciplined in deciding what to build and sell and be clearer about their product strategy.
M&A in tech becomes more unpredictable
Bain’s research shows that persistent regulatory obstacles have prompted tech companies to shift their M&A activity away from deals intended to capture scale and toward deals intended to acquire access to new capabilities, products, or markets – which Bain refers to as ‘scope deals’.
From 2015 to 2018, the percentage of tech industry scope deals increased from 50% to 80%, holding steady ever since. Over the past six years, scope deals have accounted for nearly 80% of all tech industry M&A. That’s a bigger share than in most other industries.
Bain’s research shows that tech is still heavily scrutinised and there’s no sign that the popularity of tech-scope deals will give way to a return to massive scale deals any time soon. If anything, M&A in the industry has become more unpredictable, Bain concludes.
“The technology sector is no stranger to disruptions, and as a result, we are used to seeing massive changes across the industry leaderboard every 10 years. Recently, however, the most valuable technology companies have shown remarkable resilience, holding spots at the top for many years and expanding their share of market value. Their success relies on their ability to identify disruptive trends and successfully scale and commercialise them, creating ‘winner takes most’ dynamics. For this decade, whoever masters the AI disruption will win big,” added Crawford.
