Key takeaways:
- Fidelity International’s global team of research analysts have been looking closely at how companies are starting to monetise AI, and what difference that investment is having on the ground
- They also report on the wider scope of the AI build-out, such as datacentres, and who is set to benefit
- The knock-on effect on jobs and growth remains a key consideration as the analysts assess what AI means for their sectors
A year ago, most of Fidelity’s investment analysts predicted AI would have a limited impact on companies’ profitability in 2025. The logic was that it would take more time for most businesses in non-tech sectors to adopt the technology and start seeing material benefits.
However, as we also reported at the time, many more analysts expected those benefits to start coming through beyond that one-year time horizon. And our latest survey of the research team suggests this shift is starting to happen. Nearly half of our analysts say that they expect AI will have a positive impact on their companies’ profitability in 2026, up from around a quarter last year.
All of the above rests upon one question: will AI models deliver what they are promising to deliver?
CHART 1
“What has changed is the urgency of adoption,” says Lee Sotos, who covers the US banking industry. “A year ago, the use cases by banks seemed more like generalised bullet points, but now they are detailing real-world solutions.”
How companies are monetising AI
While most use cases so far have focused on reducing costs, Sotos notes that banks are becoming more creative in adding capabilities that generate revenues.
“So far we are seeing new sales and trading capabilities with a capital markets focus,” he says. “But also in areas like prompting retail bankers for potential new product sales when a client comes in, or wealth management prompts on the best new ideas for a client.”
The ability to give customers more personalised offerings is emerging as one of the main AI use cases for banks in other parts of the world, helping deliver better operational efficiency, better customer experience, and better fraud detection.
Some of these improvements, such as targeted marketing or detecting fraud with biometrics, rely on technologies that can be labelled AI but are less readily associated with the recent rise of large language models (LLMs).
LLMs are also starting to have an impact.
“Chatbots are already helping to replace front-end people,” says Gaurav Jangale, a banks analyst focusing on the Asia Pacific region, citing conversations with C-suite members at leading banks.
Tightened operations
Financials is one of the top three sectors identified by analysts as expected to benefit from AI over the next 12 months.
CHART 2
In the top spot is communication services, with every analyst who covers the sector expecting AI to benefit at least some of their companies in the year ahead.
“Over the last 12 months I’ve gained more conviction through management discussions that AI can make telcos’ operations more efficient,” says Kazayuki Soma, a Japan-focused analyst.
“For example, with base stations, which are the servers on the ground enabling mobile telecommunication, AI can switch them on and off more flexibly based on usage, meaning more efficient power consumption.”
Elsewhere, analysts are starting to see the use of AI to drive efficiencies in sectors ranging from oil and gas and mining to consumer retail.
Alex Dong, who covers consumer staples and sportswear in China, reports efficiencies ranging from designing sportswear to operational gains in fast-food restaurants.
Large tech companies like Google, Meta, Amazon, and Microsoft are also exploring how to use AI to drive consumer revenues.
“I can see a very easy way to transition existing advertising inventory and content and meld them with AI-generated content and AI-generated feeds,” says Jonathan Tseng, who covers the semiconductor industry.
Given the size of these companies’ user bases and the scale of their existing revenues, this is likely to be a material piece of the AI revenue puzzle, although it will be hard to isolate the impact AI alone is having.
Assuming AI’s popularity with businesses continues to grow, we can expect the technology to have knock-on effects beyond the companies that use it.
Second-order effect #1: The datacentre build-out
One such effect will come from building the datacentres and other infrastructure the technology relies on.
“The datacentres supporting AI are energy intensive and will require significantly more copper,” notes Sam Heithersay, who covers Australian metals and mining companies.
“It seems evident to me that natural gas will have to be used to generate power as green sources have been overwhelmed by demand,” adds energy analyst Randy Cutler.
Shreeji Parekh, who covers North American capital goods producers, concurs, reporting a renewed interest in plants that can generate steady power.
“A datacentre’s 24/7 load profile better matches baseload – gas, nuclear, coal – than intermittent renewables, even with battery storage,” she explains, noting that renewable battery storage available today maxes out at just four hours.
Second-order effect #2: Jobs and wages
By and large, companies’ near-term use cases for AI appear to be mostly focused on driving cost savings. And the biggest cost-saving prize is the wages companies pay people.
In the US alone, annual wage spending is over US$13 trillion. If AI cannibalises even a small part of this, that implies a lot of further scope to monetise the technology. And there are already signs of AI reducing reliance on labour as a factor of production.
“Companies I cover have been growing revenue by 15 to 20 per cent, without any increase in employee numbers,” reports China healthcare sector analyst Lizheng Zhu.
Historically, weak job creation and rising unemployment has been seen as a bad sign for an economy. And while there are implications for government finances in putting taxpayers out of work, there are also those who argue that this time could be different.
A hit to humans’ earning power, through job losses or downward wage pressure, would indeed reduce the ability of those affected to fund consumption. However, richer members of society would experience a wealth effect as AI-driven efficiency boosts the value of their stock portfolios, potentially mitigating the effect on GDP.
What next for AI in 2026?
It’s plausible that not all the AI projects companies have told our analysts about will deliver a desired return on investment.
And ultimately, all of the above rests upon one question: will AI models deliver what they are promising to deliver?
There’s no definitive answer to that, at least not yet.
“Large enterprises are adopting agentic AI systems and adapting them to simplify existing business processes,” says Tseng. “And that takes time. People and processes take time to change.”
AI pessimists meanwhile fear that the capital expenditure on AI infrastructure risks destroying value by creating capacity that will sit unused.
And there are question marks over whether the companies that build and maintain the AI models can evolve into profitable businesses, given some of them have funding requirements that are multiples of their current revenues.
“If the spending commitments materialise, I don’t see any realistic scenario they can break even,” analyst Josh Han An Xin says of one of the big AI model companies he covers.
Yet staying on the sidelines also carries the risk of being left behind.
“So far,” adds Tseng, “AI models continue to improve and productisation is proceeding rapidly. If that works, then everything else will work. Trying to claim that you conclusively know that AI doesn’t deliver value based on old data and old models is like looking at the Wright Flyer and deciding that mass air travel will never take off.”
