Neither panic nor euphoria is an answer to technological changes but simple understanding of its implications, its ramifications, both in breadth and depth. Neither gaping in awe nor recoiling in fear is a sensible response. Since, despite all the hype, it is reasonably clear that AI is here to stay, wisdom lies in drawing out opportunities for various businesses. Incidentally, this holds important lessons for education too.

Technological obsolescence is another way of describing what the economist Joseph Schumpeter called ‘creative destruction’, captured in a telling manner by the eclipse of cameras when smartphones began describing their camera features more than other parts. Thus, it should come as no surprise that AI will leave a trail of the ‘displaced’, but it is not a one-sided story, as it will create new businesses and modify some existing businesses. Here is my list.

Chip-making (and all that goes with it)

Chips are not unique to AI except that design, fabrication and manufacturing bring with them an additional layer of complexity not easy to manage. These two vital dimensions are common to chips per se but just gets some additional degree in AI. 

Technologically, chip-making is an extremely complex business and is not a function of just the size of investment. I don’t see any escape from reading Chris Miller’s extraordinary book on ‘Chip Wars’ to get a grasp of how elusive success can be in this line of business.

 

Chip-packaging

The growth of e-commerce and processed foods including liquids has made packaging a huge business. However, packaging of AI-chips is a specialized requirement as temperature control is a fundamental necessity, which opens up a whole area of thermal technologies or even bio-inspired alternatives. In sum, thermal management is a crucial area as becomes clear here. Thermal management is already an established business given the growth of EVs and the concomitant demand for batteries – BTMS (battery thermal management system) is a significant area as a career too.

 

Copper processing and its by-products

Copper is the principal metal used in making AI chips.

First, upstream investment in prospecting for copper and all that goes into setting up a processing plant. 

Incidentally, this means that Geology ought to be a subject of study.  

Naturally, processing of copper for the specific purpose of using in AI chips should grow into a huge business. Now copper processing is highly complex and will automatically set in demand for associated skills, including recycling of copper. A lot of useful, actionable information can be found here.

Copper cathode is the primary raw material used in manufacturing copper wire rods, molybdenum, sulphuric acid, each with its own applications. For example, sulphuric acid is used in fertilisers.

 

Processing of waste

Whatever the degree of efficiency, waste is a natural output, as natural as the desired output, which opens up opportunities in processing of waste, pushing more and more R&D into making something out of waste. The International Copper Association has a lot of useful information.

 

Land development      

As I have already written in ‘Land-based business models in data centres’ – demand for land will just increase and increase, thus automatically creating demand for land development.

 

Data centres – setting up and management

Any data centre will need a lot of hardware and infrastructure such as network (and its management), software to run it, power and cooling systems, installing both physical and digital security, a range of related supplies or accessories, topped up by management skills in integration.

 

Power generation & Power Grids – micro to national

Data centres and power go together, as is made amply clear by the continuous search for cheaper sources of power, which include countries as well. Iceland is a classic example. We will see demand for power generation which automatically means demand for all that goes into building power plants, be it fossil-fuel based or hydro or renewable energy. There is certainly also going to be a discussion on the appropriate grid model.  

 

Miniaturization

The trend of miniaturization began with storage devices but is now an inevitable requirement wherever applicable. This trend has already deeply influenced data centres in more ways than one. Part of this trend is driven by phenomenon such as edge computing – and also technologies such as miniature networks.

In a May 2026 article provocatively titled ‘Could the next AI data center be attached to your house?’, Scientific American covers some interesting developments.

 

Monitoring of data centres – CCTV etc

Although technically a part of management, I want to separately mention this for one reason – the sheer size of a data centre. It is a foregone conclusion that there will be a network of CCTVs which will be accessed through a Control room. CCTV makers should feel thrilled at the prospect.

 

Code-writing & System maintenance

If codes are written by an AI agent under the instructions of ‘prompting’ and ‘verified’, it is a fair presumption to make that they will be efficient codes, which means, as a corollary, that there will less number of codes to maintain, exerting a downward pressure on maintenance costs.

 

Code integration – a system, not a series of patches 

This too is implicit at least until such time an AI agent ‘writes’ an entire system. It is more realistic to expect such AI-codes written for many discrete and connected tasks and functions, which opens up the need to put it all together so that a system functions as one, and not as a haphazard collection of patches of codes, however well-written each is.

In a sense, this is no different from software integration which began with the proliferation of software, though it is closer to ‘app integration’ since many apps feature within a single software system. This also means that ‘interface management’ becomes equally important, as any software developer ought to know.     

 

Pithy communication

We need a new breed of managers who can say more with less. The new generation of coders that begins to grow with experience in using some AI agent, is likely to become ‘better’ and therefore will ‘expect’ to be treated differently than they have been used to. Hence, we need managers who treat them with respect and entice them to innovate.

 

Skill development  

Prompting is the first skill to be developed because these AI agents are vulnerable to the quality of prompts. Now, this means that the prompter ought to have a thorough understanding of what the code is expected to accomplish, which means a non-technical, a more business-focused understanding of what the code is being written for. Whether we admit it or not, a large body of developers today will fall into the category of people with just a technical understanding which invariably leads to inefficient codes and thus more expensive maintenance.