AI Semiconductor Market, Led by Astra

Concurrently, Jensen Huang, CEO of NVIDIA, emphasized the progress of AI, mentioning the advent of the AGI era. Currently, over 100,000 GPUs have been deployed for training the Astra model, which clearly demonstrates the impact of AI technology development on the semiconductor market.

Hwang San-hae, researcher at LS Securities, explained that competition among major artificial intelligence (AI) model companies like OpenAI and Anthropic is intensifying. This competition is acting as a factor that increases the cost of model development and deepens price competition. Furthermore, the macro environment and interest rate pressures remain significant factors, leading to deeper consideration of cost efficiency across the AI industry.

Kim Ji-hyeok, PB, predicted that 100,000 GPUs have already been deployed for Astra model training, and an additional 400,000 GPUs are expected to be deployed in the future. Accordingly, a total of approximately 500,000 GPUs are anticipated to be used, which will further increase demand for memory semiconductors. Specifically, memory demand is expanding not only to computation but also to inference areas, leading to the expansion of the entire AI industry, including physical AI and robotics, thereby driving the growth of the semiconductor market.

Indeed, Nomura Securities and Citi warned of potential profit adjustments in the third quarter. However, AI-led stocks tend to be more significantly influenced by the AI cycle itself rather than macro logic, so the impact may be limited compared to other export sectors such as automobiles, shipbuilding, and cosmetics.

He projected that a significant portion of the won's appreciation has already been reflected in the stock prices of export-oriented companies, and the won's strength will find support at its lower bound. That is, the AI sector will be more sensitive to the overall growth trend of the AI industry rather than direct exchange rate impacts.

Hwang San-hae, researcher, expressed concerns about oversupply, suggesting that aggressive expansion by competitors like Micron could trigger a downturn. Compounding this, tariff pressures from the U.S. government are adding to the oversupply fears. He warned that if oversupply actually occurs, it could become a negative factor leading to a peak-out in the semiconductor market.

Kim Ji-hyeok, PB, argued that concerns about oversupply in the semiconductor cycle might be premature for now. This is because HBM demand is expected to continue increasing, driven by GPU demand, and major companies like Micron are also pushing to expand HBM production. Furthermore, with the spread of AI PCs and AI devices, demand for general-purpose DRAM is also projected to increase significantly, suggesting that overall demand in the semiconductor market remains robust.

In terms of long-term supply and demand, the likelihood of increased capital inflow into the AI industry has grown.

Hwang San-hae, researcher, expressed concern that excessive reliance on the AI cycle could increase index volatility. He pointed out that AI investments by many general companies are currently in the stage of verifying efficiency, and it is a phase of cost input where return on investment (ROI) has not yet materialized. Therefore, he advised caution in investing, as excessive exposure to a single AI cycle theme could increase overall stock market risk.

He emphasized that rumors without official disclosures are usually unfounded, making it crucial to verify the facts.

LG Electronics is undergoing a transformation from an appliance company to a robotics company and is set to discuss robot actuator supply with big tech companies in October. He rated robots as a core industry for physical AI and positively predicted the future of the robotics industry.

Hwang San-hae, researcher, analyzed that the strength of Hyundai Motor's supply chain establishment in the U.S. is crucial for its second-half strategy. While concerns about profit reduction due to the strong won have eased, doubts about U.S. consumer spending power still linger. He advised that in the short term, robot component stocks like actuators might be more promising than large-cap stocks, explaining that the recovery of consumer discretionary spending will significantly influence Hyundai Motor's stock price.

Kim Ji-hyeok, PB, warned of profit-taking risks ahead of the September Federal Open Market Committee (FOMC) meeting and the U.S.-China summit. He specifically stated that the CPI inflation data will be a turning point for the market, and if core CPI slows, the likelihood of an interest rate freeze increases, potentially leading to a market rally. Therefore, he suggested a strategy of looking for buying opportunities after confirming inflation data.

Hwang San-hae, researcher, pointed out that while there is great uncertainty regarding the direction of interest rates, inflation, and the AI cycle, the slowdown in liquidity growth is a definite fact. He analyzed that this decrease in liquidity could be a factor pressuring the market, and advised that it is not too late to enter the market after a favorable scenario is confirmed. Therefore, he recommended a conservative approach by holding cash.