Daily Maverick
Dailymaverick logo

Opinionistas

This is an opinion piece. The views expressed are not that of Daily Maverick.....

Is AI a bubble, and if so could it be about to burst?

Natale Labia
By Natale Labia
20 Aug 2024
Facebook
5
Amid all the hype, scepticism is beginning to emerge. Investors are questioning whether AI mania has outpaced reality. Is there a viable path to long-term profitability?

There are few things as divisive as the topic of artificial intelligence. Proponents argue that AI is a transformative force for productivity and profitability, while cynics warn of its potential existential risks to humanity. One thing is indisputable: AI has had a profound impact on financial markets, underpinning the relentless surge in stock prices over the past two years.

Rarely have financial markets been so dependent on one investment narrative. Investment firm Bespoke Investment Group estimates that without the AI boom US stock markets would be roughly 20% lower. 

This poses a systemic risk to financial markets, and indeed the broader global economy. Market valuations have priced in AI significantly enhancing the productivity and profitability of not only the companies like Google and Microsoft that are at the forefront of its development, but also the myriad other companies that are meant to eventually deploy this technology. If not, they will crash. 

The AI investment boom


The surge in market valuation driven by AI has been remarkable, with Nvidia at the centre of this growth. Nvidia designs the sophisticated and powerful chips that power the vast data centres at the core of generative AI and large language models. They are effectively selling the shovels in the gold rush. 

Mega-cap tech companies are investing tens of billions of dollars in a race to acquire increasingly powerful computing capacity, leading to insatiable demand for Nvidia’s high-priced superchips. According to Bloomberg, over the past year, Nvidia’s share price has soared by 162%, with its market capitalisation reaching an astonishing $3 trillion.

Similarly, companies developing AI models and selling AI-based software products and solutions have seen significant increases in their valuations. For example, Microsoft has heavily invested in AI, including its well-known $13-billion (about R232-billion) investment in OpenAI, the creator of the ChatGPT chatbot. Since its initial investment in OpenAI five years ago, Microsoft’s share price has tripled, enabling it to compete with Apple for the title of the world’s most valuable company and extend its lead over its rival, Google.

However, amid the hype, scepticism is beginning to emerge. Investors are questioning whether the AI mania has outpaced reality. Is there a viable path to long-term profitability? Could the AI bubble be about to burst?

The high costs of AI development


First, the costs associated with building and training AI models are staggering. According to a report from TechSite, OpenAI is projected to spend $5-billion more than it generates in revenue this year alone. This shortfall is being covered by OpenAI’s cash reserves and external investments. Although OpenAI is valued at $80-billion, its cash reserves are dwindling, and another fundraising round will likely be necessary within the next year, possibly with Microsoft once again digging into its deep pockets.

This issue is not unique to OpenAI; the entire sector is grappling with immense capital expenditures. Microsoft’s capital expenditures have soared to a record $56-billion, with the majority earmarked for AI-related investments. This figure has more than trebled over the past four years.

These expenditures stem from two main sources. First, the enormous investments required for chips and data centres to run the large language models that underpin AI technology. Second, the escalating costs of electricity to power these data centres and vast amounts of water needed to cool them. These running expenses already run into billions of dollars, not to mention their disastrous environmental impacts. The next generation of large language models are forecast to require several times more electricity and water.

The situation draws parallels to a Cold War-type arms race. Companies are locked in a competition they cannot afford to lose, yet increasingly look like they cannot afford to win either.

Challenges beyond costs


Beyond the financial challenges, AI technology faces other significant hurdles. 

First is the scarcity of high quality human-produced data on which AI models are trained. These models have already consumed a vast portion of the content humanity has created, and there is limited new data available. Most of the text found online has likely already been used to train models like GPT-4, with all the training information available likely to be exhausted as soon as 2026.

As a result, AI systems are increasingly relying on synthetic data — text generated by one model to train another. The effectiveness of this approach in improving or enhancing AI models remains uncertain. They may just get stupider.

Another challenge is the growing restrictions on the data these AI models can access. Whereas the entire internet was once available to AI for training, there are now increasing limitations due to concerns about copyright infringement. Companies like the New York Times are suing AI companies, arguing that using copyrighted material for training AI models does not constitute fair use, especially if the AI generates output resembling the original material. That is just plagiarism.

Finally, despite the initial surge in interest for tools like ChatGPT and similar AI-powered bots, subsequent usage and willingness to pay have been less promising. Currently, the practical applications of AI in daily life – those that genuinely simplify tasks for users – are still not intuitive. Mastery is required to effectively utilise the technology, limiting its accessibility and utility.

Will the bubble burst?


If AI technology proves more challenging to monetise than anticipated, and if current stock market valuations do indeed reflect a bubble, the question arises: what will remain after the bubble bursts? The dot-com bubble of 2001-2002 wiped out trillions of dollars in stock market value. However, the underlying knowledge and technology survived, leading to the consolidation of companies that are now at the forefront of AI development.

For investors, this historical lesson is a cautionary tale. Even the eventual winners of the tech industry took decades to recover their previous market highs. Microsoft, for instance, only surpassed its 1999 share price peak in late 2016. 

This bubble might burst, vast value may well be destroyed. Yet not all will be lost. Out of the ashes will emerge something new. It is possible that the technology currently being developed – both hardware and software – will endure and may eventually enhance productivity and profitability. 

However, it may well not do so in the ways currently expected. Who knows what the time frames will be, or which companies and shareholders will be the ultimate beneficiaries? Trillions of dollars, and much of the broader global economy, rest on these questions. DM

Categories:

Tags: