The recent rise of AI technology, particularly exemplified by the Chinese company DeepSeek, has triggered a seismic shift in the stock markets across Europe and North AmericaNotably, this upheaval did not spare the traditional tech giants and energy suppliers that have been historically linked to the growth of artificial intelligenceThe stock prices of these companies have taken a substantial hit, evidencing the volatile market dynamics at play.

Specifically, the eve of the Lunar New Year saw energy suppliers such as Constellation Energy suffer a dramatic stock plummet of 21%, with Vistra, another major player in the electricity sector, witnessing an even steeper decline of 28%. The implications of these shifts indicate a brooding uncertainty in the sector, raising important questions about the future energy requirements predicated by AI developments.

Wesley Alexander Hill, an assistant director at the International Tax and Investment Center, stated in a Forbes article that DeepSeek's emergence has fundamentally altered the competitive landscape for AI in both China and the U.SHe emphasized that the previously held belief that AI would invariably lead to rising energy demands has been rendered obsoleteHill’s statement challenges the established norms that have guided energy policy discussions in many countries, prompting policymakers to reconsider their assumptions.

Adding to the skepticism, analysts from financial services firm Jefferies have also questioned the previously confident predictions that American electricity needs would skyrocket due to AI advancementsTraditionally, the dominating narrative suggested that as AI technology scales, businesses would invest heavily in infrastructure, leading to the need for significantly more energy consumptionThis was corroborated by a 2024 report from Lawrence Berkeley National Laboratory, suggesting that data centers accounted for approximately 4.4% of the entire U.S. electric consumption in 2023, a figure that could increase by one to three times by 2028.

Energy corporations, however, have displayed a marked enthusiasm towards these developments

Advertisements

Executives from titans like ExxonMobil and Chevron have made headlines by publicly discussing their plans to enter the power market, with intentions to supply energy to AI data centers through natural gas power generation and carbon capture technologiesExxonMobil, for instance, announced ambitions to construct a 1.5-gigawatt gas-fired power plant dedicated to serving data centers.

Yet, the low-cost training methodology developed by DeepSeek has raised eyebrows within the industry, as it casts doubt on the necessity for massive investments in AI power generationDeepSeek was able to train its sophisticated models using only 2048 NVIDIA H800 chips at a mere cost of $5.6 million—achieving performance levels comparable to OpenAI's leading model, GPT-4. By contrast, the expenses incurred by OpenAI and Google for training models of similar scale were approximately tenfoldJohn Quigley, a senior researcher at the Cleiman Center, highlighted that DeepSeek has managed to surpass current industry standards while utilizing only a minimal fraction, around 2%, of the chips, hardware, and energy that traditional methods would employ.

This efficient approach, however, does not absolve the industry of concerns about future power demandsComputer scientist Raghavendra Selvan from the University of Copenhagen warned that while DeepSeek might facilitate the development of larger models on more expansive datasets, this could inadvertently lead to increased overall energy consumptionHe referred to this phenomenon as a manifestation of the Jevons paradox, where improved efficiency in resource utilization leads to a heightened overall consumption.

The Jevons paradox, first articulated in 1865 by British economist William Stanley Jevons, underscores the notion that advancements in the efficiency of resource utilization might paradoxically lead to greater overall consumptionA historical example is the enhancements made to the steam engine by James Watt, which resulted in an increase rather than a decrease in coal consumption, due to the widespread adoption of the more efficient technology across various industries.

Industry insiders anticipate that DeepSeek's innovative, low-cost models might trigger a parallel effect in the market

Advertisements

Quigley posited that in the long term, as AI costs continue to drop, the broader utilization of AI applications would likely boost electricity demandsHe advised that grid planners and policymakers should prioritize rational layout guidance in market developments, focusing on energy generation facilities tailored for AI applicationsThere’s a clear call for greater attention to chip advancements, improvements in AI computational efficiency, and the potential of clean energy and storage solutions.

As these discussions unfold, the emergence of DeepSeek is also providing fresh new insights for energy companiesIn recent weeks, major Chinese energy conglomerates such as China National Petroleum Corporation, Sinopec, and China Southern Power Grid have announced their incorporation of DeepSeek’s model into their operationsFor instance, the Kunlun model by China National Petroleum Corporation has achieved privatized deployment of the DeepSeek model, which is now enhancing capabilities in areas such as intelligent Q&A and reasoning within the energy and chemical sectorsSinopec, on the other hand, has integrated DeepSeek into its Great Wall model application system, pledging to leverage this integration for improved efficiencies in seismic data processing, reservoir development optimization, chemical product research, and customer service models—thus pushing the oil and chemicals industry towards greater intelligence and digitization.

As AI technology progresses at a staggering pace, the real-world implications of these developments are yet to unfold in their entiretyWill energy demands drastically change in light of DeepSeek’s innovations? Or will we find ourselves in a situation characterized by increasing consumption, despite technological efficiencies? As the industry grapples with these pivotal questions, stakeholders must remain cognizant of the implications these advancements may have on energy utilization, market projections, and the broader environmental context.

Advertisements

Advertisements

Advertisements