• This year is the ‘Year of AI Investment’ and, it seems, every year for the rest of this decade will follow suit
• For the first time CIOs are specifying AI and ML as their top investment priority 
• And don’t forget storage – that’s booming as well
• The way to get an AI pay rise in 2026? Easy, just ‘leverage’ your ‘digital charisma filters’ – job done 

The AI bandwagon is really gathering pace and rolling along now like the Cannonball Express. The subject dominated the MWC24 event in Barcelona and, elsewhere, hype, expectations and share prices are rising to the extent that some analysts now wonder how long it can continue before the bubble bursts. Meanwhile, others (the majority) remain extremely bullish and say there is no chance that the AI market will experience a repeat of the 2000 bursting of the dot.com balloon.

In the US, the Nasdaq stock market index, where so many high-tech companies are listed, has hit new highs on the back of exuberant optimism that AI will engender a tsunami of economic growth and wealth creation. Watching from the sidelines, open-mouthed, as chipmaker Nvidia’s stock price, helped by its recent financial report, rose by more than 78% since the start of this year on top of having tripled in value last year, investors are scrabbling to buy into just about any AI company in the hope that it will emulate the colossus of Santa Clara and make them all a fortune.

At this point it may be timely to remind ourselves that more or less exactly 24 years ago, on 10 March 2000, in fact, that the Nasdaq composite index peaked at 5,048.62, more than double its value of just a year before. It was downhill from there on for dot.com. The Nasdaq fell by 75% from March 2000 to October 2002, and the value that had been building for close to five years since the bubble first started to grow was all but wiped out, tipping many dot.com companies, big and small alike, into bankruptcy and liquidation. As we all know only too well, history does have a habit of repeating itself.

Today, though, traction in the AI sector is now so great that the subject is now routinely covered, all too often misleadingly, in the popular non-specialist media. Unfortunately, many of the general public (and even some in the industry) do not understand that in almost every case, what is being exemplified is just one manifestation of the technology, generative AI (GenAI), which is a subset of traditional AI. Generative AI can generate text, images or other data using generative models in response to prompts.

GenAI models learn the patterns and structures of their input training data which, usually, is scraped en masse from the internet with almost no distinction between dross and pure gold. It is then used to generate new data that has similar characteristics. This is very different to traditional artificial intelligence (often also referred to as “Narrow AI”) and is dedicated to the ‘intelligent’ performance of a specific task, the textbook example of which is playing chess. Here, a program is designed to react to a set of inputs made by the person with which the AI is interacting – its opponent in the game of chess. As the AI knows all the rules of chess it can analyse them and predict what an opponent is likely to do next and then make its own move based on that prediction. The AI did not invent chess but can make informed decisions on how to play it. Traditional AI is not creative but can complete tasks extremely well, and usually very quickly, within the parameters of specific rules.

For its part, GenAI can definitely create something from information it has ‘learned’, having been trained on masses of data sets scraped from the web. It then makes something new from that data having learned their underlying patterns. It can output results that are very hard to differentiate from work produced by individual humans. GenAI is booming. The big question is – will it last?

AI frenzy could continue unabated for the next three to five years 

On Friday, 21 February, Nvidia, which develops graphics processing units (GPUs), processors originally designed for gaming applications but which have proven themselves indispensable to AI, became the fifth publicly traded company ever to exceed a valuation of $2tn. It took just nine months from the time it first broke the $1tn barrier. By comparison, it took Apple, Google and Microsoft two full years to manage the same feat.

So important are Nvidia’s chips that the CEO of Meta (formerly Facebook), Mark Zuckerberg, is investing $10bn to buy 350,000 of Nvidia’s fastest semiconductors, the H100 series (at $30,00 each). They will be key to his ambition to build “the world’s fastest supercomputer”. That surely is prima facie evidence of runaway and, perhaps, overly optimistic attitudes to GenAI. Financial faith in companies that produce large language models (LLMs) able to provide general-purpose language generation and classification and call it innovative AI is growing daily. At some point, someone is going to get burned, but not quite yet.

The $10bn investment, payable by Meta this year, will go straight to Nvidia’s sales line, which helps explain the company’s massive valuation and validates Nvidia’s expectation that superfast training on gigantic sets of data will be the order of the day for years to come. It means Nvidia can continue to manufacture ever faster and more innovative chips and continue reaping the astonishingly lucrative rewards. Meanwhile, it is gambling that GenAI systems will continue indefinitely to live in the cloud and run in gigantic datacentres. Certainly, that situation could well pertain for as long as the next 10 years but it won’t last forever. Indeed, some AI companies are already working on a sort of distributed computing model for AI training via parallel processing data across tens of thousands or even hundred of thousands of remote, laptop computers.

The consensus amongst analysts seems to be that AI is going through a manic phase but that doesn’t mean it’s just a craze or a tech bro fad that will all but disappear in a major market valuation correction in the months and years to come. For example, more and more companies around the world are coming round to seeing AI as a solution to the global labour shortage that can largely be replaced by AI-enabled automation. However, on the downside, a recent McKinsey report warned that the widespread implementation of AI will result in at least 12 million workers “pivoting” (now there’s a word!) to “different careers” by 2030 – and that’s just in the US. Globally, the number of jobs lost is likely to be incredibly high. 

Goldman Sachs estimates that spending on AI will reach $200bn by 2025 and account for up to 4% of the total economy and, what’s more, could cause GDP worldwide to rise by 1.5% by 2032. And for the successful AI companies, estimates are that their share price could rise by at least a further 30% over the next 12 to 18 months. Meanwhile, Deepwater Asset Management reckons AI frenzy could run unabated for the next three to five years, during which time the Nasdaq could double its value. As mentioned, that would be getting into potentially dangerous territory and we have been there before. 

Technology, trends come and go. They have their (often short) time in the sun and are then eclipsed and superseded by the next ‘big thing’. Others, though, grow to become technological fixtures affecting everyone, such as the internet and smartphones. It seems then that GenAI is not just a trend – it’s a revolutionary, evolutionary and transformational technology that has already changed the world and will continue to do so to an incalculable extent. It’s here, it’s massively disruptive, potentially very scary and, for good or ill, we are going to have to learn to live with it. 

And look on the bright side: According to research house Gartner, by 2026, thanks to AI, 30% of workers will “leverage digital charisma filters to achieve previously unattainable advances in their careers” (if they actually have a job, of course). 

If that’s the bright side, though, don’t tell me about the dark side…

– Martyn Warwick, Editor in Chief, TelecomTV