The investment frenzy in AI infrastructure — ranging from high-performance computer chips to huge data centres and even electricity providers — bears striking similarities to the internet boom of the 1990s.
Companies such as Nvidia, whose share price has surged dramatically with rocketing demand for AI chips, are reminiscent of the tech darlings of that era. Back then, investors poured capital into the infrastructure of the digital age, betting on the proliferation of the internet by financing vast networks of fibreoptic cables, switches and networking equipment.
Companies such as Cisco, which dominated the market for internet hardware, experienced extraordinary stock price gains as businesses raced to build the backbone of the web. Just as the internet infrastructure boom laid the groundwork for the eventual rise of transformative companies such as Amazon and Google, today’s AI infrastructure investment is setting the stage for a new technological revolution.
However, a major recent development in AI may shift how this infrastructure race unfolds. Chinese AI research lab DeepSeek has unveiled a breakthrough that significantly enhances the efficiency of model training, meaning AI systems can achieve the same level of performance with far fewer computational resources.
This efficiency improvement raises a critical question: will demand for AI chips, data centres and power-intensive computing decline as a result? Some analysts argue that this could lead to a slowdown in AI infrastructure spending, as companies can now accomplish more with less. However, others believe the AI arms race will continue at full speed, just with greater efficiency — meaning infrastructure spending won’t shrink, but AI’s capabilities will accelerate even faster.
If the latter view is correct, the real investment opportunity may not be in the infrastructure itself but in the AI applications that become commercially viable much sooner. Self-driving cars, AI-driven drug discovery, advanced robotics and personalised AI assistants are just a few of the areas that could see rapid advancement.
This echoes what happened after the internet boom: while infrastructure companies such as Cisco were the initial winners, the real long-term gains went to businesses that leveraged the internet to transform entire industries — firms such as Amazon and Google.
Hence investors who can identify today’s AI equivalents of those companies stand to make huge returns (though this is easier said than done, with many of today’s internet winners obvious only in hindsight).
The discussion around AI’s investment potential often focuses on the US, and with good reason. Historically, most of the profits linked to digital advancements accrued to global tech giants with global reach and scale in a winner-takes-all outcome. By contrast, local businesses that use similar technological platforms tend to find the going tougher. Consider South Africa’s Takealot, for instance; though modelled on Amazon, the company has yet to turn a profit (acknowledging that Amazon generates a large part of its profits from Amazon Web Services rather than its e-commerce division).
Companies such as Nutun, which specialise in outsourced customer service, are at significant risk of disruption
In the application space, this will also hold true to a degree. For example, large global pharmaceutical companies stand to benefit more than smaller, localised peers as AI streamlines research & development and accelerates drug discovery, while global ride-hailing companies such as Uber, with their unmatched scale and deep pockets, could significantly cut costs by deploying self-driving vehicles.
However, South African companies may gain more from this phase than from the initial infrastructure expansion, much as the internet revolution improved efficiencies for many local businesses without necessarily producing home-grown global tech giants.
AI could enhance productivity in multiple local sectors. Companies with large physical distribution networks may benefit from self-driving logistics fleets, reducing operational costs and improving delivery efficiency. Telecommunications firms could use AI to optimise network traffic, pre-emptively managing congestion and improving service reliability. Banks, too, could see gains as AI-driven credit scoring models improve lending accuracy, reduce defaults and lower customer acquisition costs. The manufacturing, mining and energy sectors could deploy AI-driven predictive maintenance to minimise downtime and improve operational efficiency.
However, just as AI presents opportunities, it also poses risks to certain industries, such as customer service call centres. Companies such as Nutun (part of Transaction Capital), which specialise in outsourced customer service, are at significant risk of disruption as AI-driven chatbots and voice assistants become more sophisticated.
This trend is already playing out globally: in India, one of the world’s largest call centre hubs, AI-powered customer service platforms from companies such as OpenAI, Google and Microsoft are beginning to replace human agents in many routine interactions.
AI models can now understand and respond to customer queries with near-human accuracy, handle complex multi-turn conversations and even detect emotional cues to personalise interactions. As these systems improve, the cost advantages of using AI over human agents will become too compelling for businesses to ignore, putting pressure on outsourced call centre operations in South Africa and other emerging markets.
So while AI will make many South African businesses more productive, it will also disrupt traditional industries, forcing companies and workers to adapt to a rapidly evolving landscape. Investors should be mindful of both sides of this equation, identifying firms that stand to gain from AI-driven efficiencies while also recognising industries that could face existential threats.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.