Feel Free to Ask audEERING®
Tuesday, the 3rd of septmeber, Nvidia experienced yet another significant decline in its stock price, following the black Monday that took place earlier in August. With a 9.5% drop, the company saw a staggering $280 billion wiped from its market capitalization—the largest one-day loss in U.S. history. The rapid rise of Nvidia and massive investments by tech giants reflect the immense expectations for AI, yet growing limitations, risks, and over-reliance on a few key players signal the need for a more balanced and sustainable approach to the technology’s future.
Massive investments and Nvidia’s dominance
The global race to dominate artificial intelligence has triggered an influx of billion-dollar investments from tech giants like Microsoft, Meta, and Amazon, primarily funneled into energy-intensive data centers to support their AI ambitions. However, these companies are also making moves to reduce their dependence on Nvidia’s chips, seeking alternatives to safeguard against potential supply chain risks and market dominance.
Nvidia, with its cutting-edge chips, is leading the charge, while once-prominent players like Intel have fallen behind. According to Ex-Intel board member Lip-Bu Tan, the company has been held back by bureaucracy, risk aversion, and an overly large workforce, underscoring the challenges for legacy firms in this rapidly evolving market.
Technological limitations of AI
Despite these aggressive investments, the limitations of AI technology are becoming increasingly clear. A growing concern is the degeneration of AI models, which arises from their increasing reliance on their own outputs as training data. As the internet offers only a finite amount of quality, fresh data, AI systems risk diminishing returns as they iterate on themselves, which can lead to performance degradation over time.
Furthermore, recent research highlights another critical weakness in AI, specifically in generative language models. Studies, such as one from Darmstadt University, revealed that while these models can follow basic instructions and mimic human-like responses, they show no evidence of “differentiated thinking abilities.” This suggests that their understanding remains surface-level, lacking the deeper, more nuanced reasoning capabilities expected in advanced AI. Additionally, another study with participants from the Jülich Research Center found that current AI systems struggle with reliable decision-making, especially in more complex or abstract scenarios, further emphasizing the need for improved models and more robust data sources.
Europe lagging behind in the AI race
Despite Europe’s strong technical talent, it continues to trail behind the U.S. and China in the AI race. One key indicator of this gap is the number of AI-related patents. According to the World Intellectual Property Organization (WIPO), between 2014 and 2023, China registered over 38,000 AI patents, while the U.S. significantly contributed as well. In contrast, Germany filed only 708 AI patents during the same period, with Siemens as the highest-ranking European company, coming in at 18th place globally.
This discrepancy underscores Europe’s challenges in keeping pace with global competitors. Experts highlight that Europe’s fragmented market and less innovation-friendly patent policies further exacerbate the issue, putting the continent at a disadvantage in the rapidly evolving AI landscape. The ability to patent AI innovations will ultimately determine which countries lead in technological development, as those holding the patents have control over how AI technologies are deployed.
So, how can we take a more balanced view of this situation?
To better understand the dynamics of AI’s development and the fluctuating expectations surrounding it, it’s helpful to draw on two key analytical frameworks.
The first is Amara’s Law, named after the futurist Roy Charles Amara, which reminds us that the impact of new technologies tends to be overestimated in the short term and underestimated in the long term. This idea is particularly relevant to AI, where early promises have often led to inflated expectations, but the true, transformative potential of the technology may still be far ahead of us. Amara’s insight encourages a more measured approach, suggesting that while the current enthusiasm may cool, the long-term influence of AI could ultimately surpass initial predictions.
The second useful framework is the Gartner Hype Cycle, a model that maps the typical lifecycle of new technologies. According to this model, technologies like AI often experience an initial surge of hype, followed by a “peak of inflated expectations” and then a sharp decline into a “trough of disillusionment.” After this phase of disappointment, a more realistic understanding of the technology’s true capabilities emerges, leading to a “plateau of productivity.” This cycle helps explain why AI, once hailed as an immediate game-changer, is now being met with more cautious optimism.
It also highlights that many technologies, despite initial setbacks, eventually find their place as essential tools once the excitement dies down and more practical applications are developed. For instance, AI’s ability to optimize logistics and supply chain management has already proven transformative for numerous industries.
Importantly, the race for AI dominance is far from over. This leaves room for the EU and Germany to still play a significant role, provided they act swiftly and strategically. With coordinated efforts and targeted investments, there is still time for Europe to become a key player in shaping the future of AI.