"US Can't Control AI" Semiconductor Market Expands

Along with anxieties that AI could rapidly replace human office jobs, this can be seen as a movement to cautiously explore development directions rather than an overall slowdown.

This is coupled with the phenomenon of increased semiconductor performance requirements due to the sophistication of AI models, which, while creating entry barriers and cost increases for latecomers, is expected to be a qualitative growth engine for the semiconductor industry.

These companies are implementing institutionalization for intellectual property protection to prevent unauthorized appropriation by Chinese companies, and they are making strategic moves to show a strict stance ahead of the US-China summit on September 24.

ASI (Artificial Super Intelligence), emphasized by SoftBank's Masayoshi Son, refers to superintelligence that surpasses human intelligence by more than 1000 times.

Unlike AGI (Artificial General Intelligence), which aims for human-level general intelligence, ASI is believed to possess far superior abilities that can overwhelm humans.

The advent of such ASI requires the advancement and performance improvement of semiconductor technology, forming a close complementary relationship between AI technology development and the semiconductor industry.

Production plans were also scheduled to be reduced by 5-10% year-on-year, but due to expectations of price increases next year, the downward adjustment was slightly eased.

This is a key approach for stable business operations in the volatile semiconductor market.

The spot market, based on PC DRAM, accounts for only 6-7% of transactions, and it is very negligible at around 1-2% of the total DRAM market.

Therefore, the spot market has insufficient influence to serve as a price barometer for the overall DRAM market.

These changes are expected to drive up companies' blended ASP (Average Selling Price).

There is an expectation that overall revenue and profitability will improve as the proportion of new products expands.

Dollar-based earnings estimates remain, and operating profit is expected to increase compared to the second quarter.

With increased predictability of demand due to diversification of data center customers, PBR alone is insufficient to evaluate valuation.

Therefore, a comprehensive valuation approach that considers both PER (Price-to-Earnings Ratio) and PBR simultaneously is needed.

The AI cycle is expected to continue growing until around 2030, with overall industry expansion anticipated due to surging data center demand.

As an era has arrived where it is difficult to judge using only existing indicators such as PBR, flexible investment responses according to changes in supply and demand are necessary.

The proportion of these sectors in foreign and institutional portfolios is expected to gradually expand, and maintaining this weighting is considered important for outperforming market returns.

Hyperscalers maintain operating profit margins in the mid-to-high 30s, and data center investment costs are expected to be recouped within 2-3 years.

Google is pursuing custom chip projects such as Blackstone and Apollo in a joint venture with Broadcom.

It aims for custom chip sales of $50-60 billion by 2027, and expects to achieve $250 billion in sales and a 60% operating profit margin by 2028.

This in-house chip strategy, coupled with the realization of long-term contracted GPU prices, contributes to margin improvement, and Google is highly likely to turn positive on an EBITDA basis within the next 2-3 quarters.

This shows a tendency to react more sensitively to negative news rather than potential market threats.

As the proportion of global equipment manufacturers' exports to China plummeted from 40% to the early 10s, Chinese companies are finding it difficult to secure necessary equipment in a timely manner.

Changxin Memory (CXMT) is on the US Department of Defense blacklist, and YMTC is on the US Department of Commerce blacklist, severely restricting their access to semiconductor equipment.

With a surge in orders for major companies, including TSMC, over the past five years, equipment supply bottlenecks are intensifying, and Chinese companies on the blacklist are inevitably pushed down the priority list for critical equipment supply.

Based on PBR valuation, SK Hynix is suggested as a buying opportunity around KRW 1.5 million, and Samsung Electronics in the mid-KRW 200,000 range.

The key to a semiconductor market rebound lies in changes in demand, not supply.

Improving ROI (Return on Investment) figures for hyperscalers are important, and it is necessary to prove market anxieties with actual performance numbers.