3月24日，高盛分析师团队下调了机架级AI服务器(Rack-level AI Server)销量预测，
Source: Wall Street Journal.
Goldman Sachs believes that factors such as the impact of the product transition period, uncertainty in supply and demand, and challenges in production complexity will lead to a slowdown in market growth, reducing the forecasted shipment volumes for rack-level AI Servers for 2025 and 2026 from 0.031 million units and 66,000 units to 0.019 million units and 57,000 units (calculated using 144-GPU equivalent). At the same time, Goldman Sachs has lowered the Target Prices for several AI Server supply chain companies in Taiwan and downgraded Quanta's rating from "Buy" to "Neutral."
Goldman Sachs has downgraded the AI Server shipment forecast, anticipating a slowdown in Industry growth.
On March 24, the Goldman Sachs analyst team downgraded the sales forecast for rack-level AI Servers, expecting the shipment volumes for 2025 and 2026 to be adjusted from 0.031 million units and 66,000 units to 0.019 million units and 57,000 units (calculated using 144-GPU equivalent). This adjustment is primarily due to the impact of the product transition period and the uncertainty in supply and demand.
Goldman Sachs believes that while the second quarter of 2025 will be a strong quarter for Taiwan's ODM/thermal supply chain, it holds a more conservative view on the full-year shipment volume, anticipating that the product transition period may again impact shipment volume in the third quarter of 2025.
Therefore, Goldman Sachs has correspondingly lowered the Target Prices for related companies in Taiwan's ODM and thermal supply chain (Quanta, Foxconn, FII, Wistron, Asustek, and Double Wing), with decreases ranging from 7% to 21%, and downgraded Quanta Computer's rating from "Buy" to "Neutral." Moreover, Goldman Sachs believes that ASIC AI Server-related companies generally outperform GPU AI Server suppliers, a trend that may continue.
Goldman Sachs pointed out:
The demand for high-performance AI Servers will not be completely replaced by rack-level forms, as some customers still prefer to use Main Board solutions to achieve design flexibility. There is optimism regarding the outlook for AI inference and general Servers, driven by the application of AI technology and the recovery of the upgrade cycle.
Goldman Sachs has significantly lowered the Global AI Server shipment volume.
Goldman Sachs' analyst team expects that AI training Servers will remain the main driver of market growth, primarily driven by the demand for computing power from continually upgraded advanced AI models. However, the growth rate of shipments is expected to be lower than previously anticipated, mainly due to the following factors:
Impact of product transition period: As the GPU platform transitions to the next generation of products in the second half of 2025, shipments may slow during this transition period;
Challenges in production complexity: The production complexity of full rack systems is high, increasing the uncertainty in ramping up capacity;
Demand-side variables: With the release of more efficient AI models (such as DeepSeek), there is controversy over the market's demand for intensive computing capabilities;
Tariff risks: ODM manufacturers need time to diversify production bases and improve yield rates, while cloud computing service providers and enterprise customers also need time to assess additional costs.
Therefore, Goldman Sachs has also adjusted its revenue forecast for AI training Servers, expecting related revenue to grow by 30% year-on-year in 2025, reaching $160 billion; and to grow by 63% year-on-year in 2026, reaching $260 billion. Previous forecasts were $179 billion and $248 billion.
Rack-level AI Servers: Shipments are expected to be 19,000/57,000 racks (calculated based on 144-GPU equivalent) in 2025-26, with a market size of $54 billion/$156 billion (previously forecasted at $88 billion/$182 billion).
High-power AI Servers: Shipments are expected to be 423,000/423,000 units (calculated based on 8PU equivalent) in 2025-26, with a market size of $106 billion/$104 billion (previously forecasted at $90 billion/$66 billion).
AI inference Servers: Sales are expected to grow by 41%/39% in 2025-26, with their value increasing by 105%/33%, thanks to the continuous expansion of application fields.
General-purpose Servers: Sales are expected to grow by 6%/4% in 2025-26, with revenue expected to increase by 9%/7% year-on-year, benefiting from the gradual recovery of the replacement cycle and the launch of new CPU platforms.
Overall downward adjustment of supply chain stock price expectations.
Based on the above outlook for the AI server market, Goldman Sachs has lowered the earnings forecasts and Target Prices for several AI server supply chain companies in Taiwan, China, with Quanta's rating downgraded to neutral.
Goldman Sachs believes that within the overall coverage of the Greater China Technology Industry, Quanta's Target Price has relatively limited upside potential. Goldman Sachs predicts that due to uncertainties in the demand for rack-level AI servers and the upcoming production transition, which may lead to delivery delays, there is uncertainty in the demand/supply of current generation rack-level AI servers. In this framework, Goldman Sachs has lowered the Target Prices of Quanta, Hon Hai, FII, Wistron, AVC, and Auras by 7%-21%.
For investors, this market adjustment reflects a shift in the AI server industry from frenzied growth to rational expansion, and the Company Valuation of supply chain companies has been adjusted accordingly. Goldman Sachs' analysis states that while growth is slowing down, AI infrastructure investment remains a major growth driver for the Technology industry, but due to various limiting factors, its growth rate will be more moderate than previously expected by the market.
Editor/jayden
Today, after hitting the high near 24,000 in the early session and entering bearish positions, the index fell sharply by nearly over 600 points, immediately recouping yesterday's losses significantly.
Moreover, today it broke the new high again, reaching a maximum of 24,076, but by the end of the market, it fell back by about 70 points, producing a bearish candle. The current trend has not yet been broken, but from the previous low until now, it has risen close to 6,000 points. It is believed that those with positions can continue to hold until there is a clear trend reversal for profit-taking. Those without positions can wait for a pullback to get in. Actually, it is hoped for a quick pullback, as it allows for entry and also provides a healthy breath.
Currently, the outlook remains the same as before. It is believed that even if there is a pullback, it shouldn't be too deep. However, if Futures fail to stabilize and close below 22,350, there may still be room for decline. The chance of Futures falling below 21,400 in the short term should be low, so it is considered that if a significant pullback occurs, it presents a good opportunity to incrementally go long. Recently, there has been a consistent approach to not hold positions overnight, only focusing on immediate trades, as there is no high chasing and no casual short selling.
Comment(5)
Reason For Report