Investing.com -- DeepSeek, a Chinese AI model, is disrupting power and energy infrastructure stocks. Analysts at BMO Capital Markets note that DeepSeek's high efficiency and low energy consumption are key factors.
DeepSeek servers use 50-75% less power than Nvidia (NASDAQ: NVDA )'s latest GPUs, which is significant for energy-intensive data centers.
The implications of DeepSeek's energy efficiency ripple across the energy sector. Current AI infrastructure, particularly in the United States, relies heavily on power-intensive technologies such as Nvidia’s H100 GPUs, which demand approximately 150 MW of power for a 100,000-unit cluster.
In comparison, a cluster using Nvidia’s H800 chips, reportedly employed by DeepSeek, could operate at nearly half that capacity, consuming just 75 MW for an equivalent configuration.
This reduction in power intensity could lead to substantial shifts in demand for electricity, a concern for companies deeply leveraged to the data center boom.
U.S.-based power companies have long banked on the growing appetite for energy driven by AI's expansion.
However, DeepSeek's model introduces uncertainty into these expectations, given its potential to reduce the energy needs of large-scale AI operations.
BMO analysts said that while the U.S. needs to increase energy supply, DeepSeek's advancements will likely accelerate the demand for more efficient AI infrastructure.
DeepSeek V3's low cost ($6 million) creates competitive pressure, forcing companies to re-evaluate their spending in a cost-conscious environment.
The reliance on Nvidia’s H800 chips—a result of U.S. export restrictions on its H100 model—also underscores the geopolitical dynamics at play.
These restrictions were intended to curb China’s access to cutting-edge technologies but have inadvertently highlighted alternative approaches to achieving competitive AI performance at lower power costs.
As power-intensive technologies face challenges from more efficient systems like DeepSeek, companies such as GE Vernova and Bloom Energy—both closely tied to energy infrastructure—may see volatility in their stock performance.
While analysts at BMO Capital Markets maintain an “outperform” rating for GE Vernova, they express caution over the broader uncertainty tied to evolving AI technologies and their implications for power markets.
DeepSeek signifies a shift for the energy sector. As AI infrastructure becomes more energy-efficient, the energy sector must adapt.
This may not reduce overall energy demand for power companies, but it will likely change the nature of this demand, requiring a reassessment of traditional growth expectations.