Investment

As the AI Bubble inflates, can operators extract returns?

As the AI Bubble inflates, can operators extract returns?

Artificial intelligence has become the most ubiquitous term in an industry obsessed with buzzwords, with the astronomical levels of hype matched only by the investment being poured into the technology.

However, while companies – and particularly telecoms operators – are scrambling to demonstrate that they are investing in AI, many are doing so without a clear strategy. And while many are implementing AI systems, they have no obvious metrics for qualifying their value – indeed, there is a growing sense that telecom operators are not seeing a return on their investments into AI, which has fuelled speculation over the AI investment bubble.

Pressure to Invest

Cole Brodman, CEO of AI & machine learning software firm Opanga, concedes that across the industry, every business and telecoms operator feels a pressure to do ‘something’ with AI from their boards and leadership. He notes that currently much of the investment seems to be fairly exploratory, with proofs of concept, dashboards and generic tools that have not yet delivered measurable operational outcomes, and argues that AI must be connected to the right data set to deliver real value in terms of network or customer experience improvements.

“Areas like spectral efficiency, congestion reduction, energy savings…could and should be different”, says Brodman. “AI operations [can] help pinpoint network issues - those are real problems that AI can solve. While many operators today may not be seeing returns, that's not because AI lacks value. It's just that they're not applying it to what we feel is the right parts of the network where ROI is quantifiable and real, where customer experience can be impacted positively and where energy consumption can be reduced. We're using AI to essentially drive what we think is missing in the in the operator networks, which is true visibility to the right type of experiences and the right types of returns.”

Cliff de Wit, CIO at ADG, argues that the initial assumption that companies jumped on the AI bandwagon in order to claim that they are doing something in the field is very real. He notes that this is particularly common among larger firms that have the resources to set up new teams focused on AI – they tend to hire data scientists then leave them to their own devices and hope for the best. This approach has largely failed because no matter how transformative as AI is, it still needs to follow some basic principles – there needs to be a clear problem for it to solve, and there must be clear criteria for measuring the return on investment.

“Smaller companies who haven't actually gone headlong into it, throwing a bunch of stuff against the wall hoping it would stick, are being much more pragmatic now”, explains de Wit. “This is the lesson from the mature markets that's bleeding into the emerging markets: that you should approach this the same way you approach any other transformative project. That you should understand the problem you're trying to solve, define the measures and the outcomes and measure that, and then the results are quite pleasing in many cases. While there's hype, I think sanity may be coming back into the equation as we move through the hype cycle.”

Bursting the Bubble

However, Matt Walker of MTN Consulting is less convinced that the hype will calm down before it’s too late, and seems sceptical of a quick fix to the chronic lack of returns on AI investment. “The problem is probably understated, or at least too much of public discourse ignores it. When webscalers [i.e. the likes of Alphabet, Amazon, Microsoft and Meta, also known as hyperscalers] were focused on building out a data centre footprint to support cloud services, the right metaphor may have been “land grab”: they spent loads of money, but they at least got “land” (or footprint) in exchange. The current mania is more like a gold rush. Nobody knows if they’ll stumble upon something useful, or reach it faster than the next guy.”

Walker explains that in his view, investment by webscalers has been in bubble territory for several quarters, partly driven by inflated expectations about the near-term benefits of AI to growing company revenues and profits.

“The key problem is that there is a "winner take all" belief held by management teams and investors”, says Walker. “At this point, most people realize that true AGI [artificial general intelligence] is far into the future, so it's not really a race to AGI. It's a race to develop models good enough to secure support from a wide range of user groups, especially enterprise verticals. They are hoping to lock in users as early as possible.”

Walker notes that the key webscalers in the US driving capex skyward have all faced - and crushed - numerous rivals along their way to their current dominant positions, and they don’t wish to have the same fate befall them in the AI race. “Bubbles are driven by mass psychology - the fear of missing out is real. The current AI bubble began before Trump got into office, but his arrival and willingness to mix politics and business have made it several times worse - and the pop of the bubble will be worse because of it.”

So are we really shifting from boom to bust? According to Walker, there are several signs already flashing red: off-balance sheet financing to hide the impact of spending; circular financing arrangements; lack of profitability metrics, or at least failure to report any convincing data to show the investments are worthwhile; companies getting funded without business plans. The explosion in capex would require a significant uptick in revenue growth which is not yet apparent anywhere; additionally, very little public discourse on AI mentions China, but models such as DeepSeek and Alibaba Qwen are showing strong results with far lower levels of investment.

“Emerging markets have a strong incentive to work with the Chinese model developers. Avoid excessive reliance on individual companies at this point, take your time and don’t believe the hype”, cautions Walker. “Many people are burying their heads in the sand and just hoping that they aren’t stuck holding the bag.”

Cutting Through the Hype

De Wit counters that the bubble is not as clear-cut as the dotcom bubble, although he concedes that we’re at the top of the hype cycle for AI, heading from a peak of inflated expectations into a trough of disillusionment. Emerging technologies always feel like they can do everything, but this needs to manifest in reality. De Wit argues that there are tangible results of AI achieving strong results, but that customers need to be pragmatic about their expectations, particularly in the context of emerging markets such as Africa, where technical engagements need to define the ‘what’ before the ‘how’. He notes that slightly less mature  customers, particularly at C-suite level, will openly state that they know they need to be doing something with AI, but they don’t know how to do it. ADG works with them to identify whether there are AI use cases that can address their core problems, which can then be converted into projects with measurable ROI.

Brodman concurs that every operator in the world feels like they have to do something with AI, so a lot of them are investigating it – but notes that in emerging markets, the cost of investment and the cost of failure is very high because the economic stress is more acute than in other places, from both a network perspective and a business model perspective.

“That's where AI really has to pay off. The hype train has certainly been fuelled by a lot of these higher income markets chasing a lot of use cases, but network AI [is] practical, inexpensive to deploy, [and] immediately cost reducing. It’s exactly the type of AI and automation that emerging markets should invest in because your returns are going to be orders of magnitude greater than the cost.”

There are valid concerns across the industry that AI investment isn’t delivering returns, and operators in emerging markets should pay heed to these by tempering their expectations and focusing their investments on AI use cases that will actually impact their operations. AI is beginning to face criticism for being a solution in search of a problem – amidst the hype and fear of missing out, operators must selectively identify issues that can be effectively addressed by AI implementation, rather than simply pumping more money into the technology and keeping their fingers crossed.



More Articles you may be Interested in...