After studying the fundamentals and technical aspects, and conducting research with experts, Goldman Sachs once again bullishly recommends: Buy the dip in AI!

Is it time to bottom fish AI stocks? The best time to buy AI may have already arrived.

Since DeepSeek impacted the US stock market, the AI sector has cumulatively dropped by 17.5%. Goldman Sachs believes this has created an excellent buying opportunity, even though the current market holds a pessimistic view on the AI sector, all fundamental data—including corporate earnings expectations, valuation levels, and industry outlook—indicates that this is the best entry point since last summer.

Goldman Sachs believes that if DeepSeek's success can be applied or replicated, AI usage will accelerate, and more participants will enter this field; it is expected that the earnings of all AI themes will continue to grow over the next 12 and 24 months, surpassing the S&P 500 index.

Pessimism is "not uncommon," hedge funds have begun to return

According to Goldman Sachs, this is not the first time the AI industry has faced a crisis of investor confidence. Last summer, the broad AI sector "Power Up America" basket index (GSENEPOW) lagged the market by 19% due to investors questioning the timeline for AI investment returns.

Vincent Lin and Erin Tolar from Goldman Sachs' Prime Brokerage department pointed out that this year, GSENEPOW has experienced significant selling by hedge funds because investors have raised doubts about the outlook for AI power demand and reduced positions during the recent momentum pullback.

It can be said that the pessimism towards the AI sector is greater compared to other themes. However, the report states that in the past week, some buying activity has begun to be seen, primarily driven by long positions, with a short covering ratio of about 2:1.

The fundamentals of power demand remain strong

Goldman Sachs' utility research department noted that following the DeepSeek impact, there has been a noticeable increase in investor inquiries regarding the outlook for AI power demand and the pace of AI investment.

However, according to statements from utility companies during the latest earnings season, the fundamentals supporting power demand for data centers have not changed, and the attitudes of hyperscale computing providers regarding their power demand or timing preferences have not shifted.

Goldman Sachs predicts that by 2030, the annual compound growth rate of electricity demand in the United States will be 2.5%, with data centers (including AI and non-AI) contributing about 100 basis points. Even if all AI power demand growth is adjusted out of the model, the total annual compound growth rate of electricity demand in the United States is still expected to be 2% by 2030.

Natural gas: A key power solution for data centers

According to Goldman Sachs' John Mackay on natural gas and pipeline research, considering reliability and time to market are key considerations for data center developers, the market generally views natural gas as a key power solution for data centers in the medium term, supporting the outlook for increased natural gas pipeline capacity. Mackay pointed out that the focus remains on the potential locations for new data centers in the United States, the role of natural gas and renewable energy in the power structure, and the comparison of the growth in data center demand with the driving factors for the increase in infrastructure capacity.

Renewable energy generation, especially solar power, will continue to benefit

Data from Nick Cash in clean energy research indicates that by the end of this decade, electricity demand in the United States is expected to grow by about 2.5%, with this growth accelerated by data centers (approximately 0.9% of the estimated 2.5% growth is expected to come from data centers).

Cash believes that the growth in demand will require an increase in power generation, and solar power generation (mainly utility-scale) will be the biggest beneficiary during the period from now until the end of this decade, with solar expected to account for more than 50% of new generation capacity.

This is primarily based on two factors: first, solar power is the lowest-cost new form of electricity compared to other generation types; second, solar is also the fastest-growing new form of power generation, with a construction time of 12-18 months, whereas gas turbines take about 4-5 years, and new nuclear power plants require 8-10 years.

AI sector earnings are expected to continue to grow, and the power sector is the cheapest

Goldman Sachs expects that earnings from all AI themes will continue to grow over the next 12 and 24 months, exceeding the S&P 500 index.

Data also shows that AI power is the cheapest investment angle in the AI field, now back to summer 2024 levels.