What Are the Risks Jeopardising the $1 Trillion AI Boom?

Most of us are now using AI in some way. Whether it is content creation, carrying out detailed internet searches, or making enhanced images for social media posts, a growing number of people are turning to the power of AI. However, all of the content generated by casual media users is dwarfed by the amount that industry spends on using AI, and the systems that are now in place to ensure that AI is a major tool in business and industry.

The Spiralling Spend on AI

While AI systems tend to be decentralised, they operate through enormous data centers and these are getting larger every year. Alphabet Inc. – the holding company that effectively owns Google – are expected to spend somewhere in the region of $48 billion in the next year and that spend is only going to rise as the company embraces AI systems. And they are not alone. Microsoft are currently in line to spend at least as much as Alphabet, if not more on cementing their own AI data centers. 

Collectively, these two technology giants are going to spend nearly $110 billion on bespoke data control facilities. When you add in the spend for other major companies such as Meta, Amazon, and Apple, then the total value is likely to reach well beyond $300 billion next year alone! Forward projections are putting spend on AI infrastructure at somewhere around $1.4 trillion.

Is It All Worth It?

Of course, spending that kind of money in business is usually done as an investment; the companies are expecting to get a rich return on the money spent, but investment in AI systems is an infrastructure exercise, and there isn’t any real tangible result from the investment. That kind of thing makes investors and shareholders a little uncomfortable to say the least. 

Any business investment is usually backed up by a comprehensive pay-back calculation that will show how the new development will bring money back into the company, by either extending its product range, or by introducing some new progression. However, investment in AI is a lot less certain and where the final endgame will lead is simply unknown. There is no way of knowing how – or even if – all of that investment will be paid back.

One thing is certain; many of the companies involved in building this AI infrastructure in real terms are making significant profits from building the myriad new data centers. Chief amongst these is Nvidia – the Santa Clara microchip manufacturer – who have become a major contractor in the construction of essential parts.

Chips With Everything

Apart from obvious things such as physical buildings and services, the investment in data centers can be split roughly into two distinct parts; computer chips and the hardware and systems needed to drive them. 

The need for computer chips has had such an impact on data-related businesses that manufacturers such as Nvidia have seen their business stock rise by around 14%. That compares with an average of just 1% growth in businesses that are not associated with the high-technology industry. While there are plenty of chip manufacturers around the world, the lion’s share of the market goes to Nvidia, and the company has recently posted forward sales predictions of around $104 billion over the next year, with much of it on equipment for AI data centers.

Regardless of the fact that no one can really say where AI will end up, these data centers are growing in both number and complexity, and that leads to another issue; power. As the size of these centers increases, and developers pack even more electronic systems in to enable processing, more power will be needed but, ironically, so will the need for cooling increase.

When operating under a heavy load, a computer chip shouldn’t be at a temperature above 80°C to 85°C, or it is likely to suffer damage. This becomes an increasing problem when there are systems with many chips all operating together, and contributing to the overall ambient temperature. To prevent any possibility of damage, data centers are usually cooled to between a somewhat chilly 18°C to 27°C. You can probably imagine how much energy that takes.

Because of the complexity of data centers, they take many months and sometimes years to construct, once again demonstrating the commitment that industry has towards AI systems. They are pouring a huge amount of money and resources into building an infrastructure that will see AI thrive, but what are the threats to all of this?

Boom or Bust?

Undeniably, AI has a number of features that make it vulnerable. Let’s have a look at what factors could bring AI down.

Non-availability of computer chips. As we have already seen, the entire industry is dependent upon the constant supply of highly complex silicon chips, and were they to become in short supply then it would severely impact the data centers. There are several areas from which supply issues could arise:

• Cyber-attack. Chip manufacturers are usually well protected, by hackers are wily. A serious electronic attack on a manufacturing facility could impact chip design and manufacture for months.
• Lack of materials. Silicon chips aren’t just made from silicon; there are also a slew of rare-Earth materials used either in their manufacture or as support structures, and these are in increasingly short supply. If they are not available, neither are silicone chips.
• Market forces. Chips are highly valuable, and it isn’t beyond imagination that suppliers – such as Nvidia – could place a stranglehold on the market to maximise their profits. With few alternative suppliers, it’s a real possibility.

Power outages. The power infrastructure is quite robust, but it’s not beyond imagination that parts of the system could break down.  Reasons could include.

• Unintentional damage. Wildfires, earthquake, storm damage are all credible ways in which the power infrastructure could be damaged to the point where it causes significant outages.
• Intentional damage. Terrorist action springs to mind but there are other areas of social unrest that could lead to the power system being suspended. While it would take major damage to stop power supply completely, the chip manufacturing system is so well balanced that even a small dip in power could be a problem.

Lack of Investment. If shareholders start to get too jittery and pull funding on future purchasers, chip manufacturers may well scale back production to protect their own investments. By simple scaling back production for just a few months would have a huge, knock-on, effect on the chip market.

The Take-up of AI Systems. All of this development is based on the notion that AI will become increasingly important to industry. If that doesn’t happen, or something better (I have no idea what) comes along, then AI data centers could be left unused.

Nvidia have scaled up their development in recent years, going from the design and introduction of a new chip from every two years to every year. That kind of development of such a complex device requires significant investment, and it would only take a relatively small hiccup in the market to make them rethink their strategy.

AI is fast becoming a fact of life, and on the current trajectory is destined to power a growing number of areas of our lives. Unity Developers are using AI as much as possible to power and refine our development, but is it all going to end badly, with AI undone by its own life-blood – the very energy that it needs to survive?