Artificial intelligence (AI) has recently become the most important trend in the tech sector, thanks to global interest. Companies are either integrating AI into their technologies or developing entirely new AI-dependent innovations. This surge in interest has also captivated users, leading platforms like ChatGPT to become the fastest-growing service of the past decade, with over 200 million active users weekly as of September.
This success attracted numerous investors eager to capitalize on the potential returns once AI goals are met. Thus, the AI investment era began, with companies like Microsoft betting on ChatGPT, followed by various investment funds. Entrepreneurs also saw this as an opportunity to present their AI models and businesses.
However, as the excitement settled, Barclays Bank made a surprising statement that AI might not be the smartest investment in 2024.
High Depreciation Costs in Ai Companies
Barclays issued a report detailing its outlook for AI companies over the next two years. The bank noted that depreciation and wear on AI equipment are higher than usual, which rapidly eats into profits, ultimately lowering net returns from AI investments.
Barclays' analysis focused on the cost of acquiring AI chips and building AI servers, which are currently very high. These costs must be spread over the expected lifespan of AI equipment, a standard practice across industries. However, AI hardware tends to have a shorter lifespan than other corporate assets. This means that no matter how significant the investment in AI infrastructure, the depreciation must be accounted for over a shorter period, possibly just one to two years, before companies need to purchase new chips to continue operating AI technologies.
Ross Sandler, an internet companies analyst at Barclays, mentioned that some AI companies are trying to extend the useful life of their servers to five or six years in order to reduce depreciation rates, thus increasing profits. However, this could prove difficult given NVIDIA's aggressive pace in releasing new AI chips.
A New AI Chip Every Year
Ted Mortenson, CEO and technical strategist at Bird, a firm specializing in investment fund management, explained that NVIDIA's rapid development cycle for new chips makes it challenging for companies to slow depreciation. Mortenson described NVIDIA's approach as a headwind against attempts to boost profits for companies investing in AI.
NVIDIA seems to be following a similar model to its graphics card business, planning to release a new AI chip every year. Each new chip would offer greater performance, creating a new standard that forces companies to upgrade to the latest generation.
Barclays predicts increasing depreciation costs, pointing to Alphabet, the parent company of Google, which currently faces $22.6 billion in resource depreciation. This figure is expected to rise to $28 billion by 2026, a 24% increase. Similarly, Meta’s depreciation is projected to hit $30.8 billion compared to $21 billion today, marking a 47% rise.
Sandler emphasized that this trend is affecting all tech companies involved in AI, regardless of their size.
 |
| Some AI companies are trying to extend the useful lifespan of their servers significantly to 5 or 6 years. |
Where’s the Return on Investment?
Mortenson believes the real question companies must address isn’t about the costs of AI technologies or their implementation, but rather about the return on investment (ROI). He questioned the fate of Wall Street's $200 billion-plus investments in AI and the expected ROI.
He noted that it is still too early to expect significant profits from AI technologies, predicting that returns might begin to materialize by 2026. The question is: Can companies survive until then?
If the ROI for AI technologies turns out to be high, then concerns about high investment and depreciation costs will disappear, making AI a worthy investment.
Another issue is the relationship between tech giants and NVIDIA. NVIDIA continues to profit by raising the costs for other companies. Each time it introduces a more powerful chip generation, it forces companies to upgrade and incur additional costs.
While this strategy works with consumer graphics cards, it could frustrate major corporations and potentially attract regulatory attention. The question remains: Will NVIDIA maintain this "toxic" relationship, or will it develop a new approach to address corporate concerns and reduce their losses?