Why I'm Not Selling Semiconductors Amid Plunge: AI Productivity

An analysis suggests that a strong won could reduce Samsung Electronics' and SK Hynix's operating profits by approximately 12%. This scenario assumes no change in semiconductor prices and volumes, meaning the actual impact could be greater. Analysts typically calculate profits assuming an exchange rate around 1,500 won.

Current analyst consensus (market expectations) may be 10-15% higher than actual profits due to exchange rate assumptions, so it needs to be adjusted downwards. If profitability is adjusted, the P/E (price-to-earnings) ratio will increase proportionally.

Looking at past semiconductor cycles, profits have plunged by more than 50% in 9 out of 10 cases after a boom. The current P/E ratio of 4.8 times reflects market anxiety, driven by these past cycles of profit halving. This suggests that investors are not fully trusting the AI boom due to past trauma.

Micron's current P/E is around 6.8x, a slight increase from its bottom. This shows that the market's view on the memory industry is still trapped by past trauma, not fully trusting the AI boom. Current stock prices are interpreted as not fully reflecting expectations for AI.

SK Hynix and Samsung Electronics maintain about a 30% discount compared to Micron. This shows that the historical practice of discounting SK Hynix, from its past as a second or third-tier company, still persists despite its current HBM (High Bandwidth Memory) growth potential. Despite positive news from partners like Nvidia and expectations of rising HBM prices next year, the stock prices of Korean semiconductor companies remain undervalued.

Despite ongoing macroeconomic noise such as oil price and interest rate instability, semiconductor stocks remain undervalued. Applying past cycle valuations of 6-7 times, Samsung Electronics could be expected to reach the low to mid 2 million won range.

Individual investor deposits decreased by 40% compared to May, weakening the market's buying power. As individual investors' influence wanes, market momentum is slowing. Foreign investors tend to decide on domestic stock trading by observing trends in overseas competitors like Micron and SanDisk.

High Bandwidth Memory (HBM) is bringing structural changes to the semiconductor market. Producing one HBM replaces the production of three standard DRAM units. The increasing proportion of HBM from Samsung and Micron, confirmed quarterly, is central to this change.

The decline in P/E due to interest rate hikes and increased discount rates is an acceptable factor. However, smart money will likely continue to explore fundamental opportunities until critical points are reached, such as oil prices at $120 or 30-year Treasury yields at 6%. Even amidst current wars and high interest rates, there will be continued focus on companies' intrinsic value.

Instead of fearing temporary market plunges, it's important to execute staggered purchases. Department Head Kim Jang-yeol emphasized 30 staggered purchases, explaining that this means consistently approaching the market over approximately 10 trading days. He advised investors to adopt a long-term, planned buying strategy rather than reacting impulsively to market volatility. He added that staggered purchases could begin even at today's closing price.

In the past, high inflation and interest rate hikes led to a vicious cycle, but the current situation is different. The U.S. economy maintains robust employment, and efficiency gains from AI adoption are creating a virtuous cycle of employment and investment. The analysis suggests that appropriate interest rate hikes are slowing inflation without collapsing the economy.

The most critical indicator for judging the sustainability of AI investment is 'productivity.' General businesses, as customers of big tech, must see tangible profits from using AI for the next wave of AI investment to occur. Increased corporate profits must lead to national GDP productivity improvements for a virtuous cycle to be possible, which will be the key basis for AI adoption to be recognized as an investment rather than just an expense.

Multifactor Productivity (MFP) is an important indicator for measuring the effect of AI. MFP represents the increase in production solely from improved efficiency, without additional labor or capital input. Organizational efficiency, technological innovation, and AI utilization are included in this metric, signifying increased productivity through efficiency improvements. Improvements in corporate work methods can all lead to an increase in MFP.

During the IT revolution from mid-1995 to 2000, the U.S. productivity growth rate significantly jumped from 1.5% to 2.5-2.7%. MFP growth also rose from 0.4% to 1.3%. This is a historical example of how past IT investment boosted overall economic efficiency, serving as a significant comparison point for assessing AI's current potential.

During the past IT bubble, labor productivity growth was evident, with MFP (total factor productivity) contributing almost 0.8 percentage points to the overall 1 percentage point increase in production. Investments in PCs, communication equipment (Cisco), and software led to corporate work innovation, improving productivity. This case demonstrates that the synergy of IT hardware and software is essential for productivity improvement.

When companies use AI to boost productivity and cut labor costs, a virtuous cycle of reinvestment in AI is formed. This reinvested capital flows into cloud service providers (CSPs), and 50% of CSP expenditure creates a trickle-down effect to memory, foundry, and network equipment companies. Ultimately, AI adoption drives growth across industries.

There's a potential 2-3 year lag between AI investment and observed productivity gains. During this period, highly indebted companies may face crises due to cash flow mismatches. Companies with high debt ratios, like Oracle, could be particularly vulnerable, requiring caution in investment.

Assume 30% of total corporate payroll ($30 trillion) is exposed to AI. If this exposed portion is reduced by 3% annually for 10 years, 40% of the savings, or $300 billion, could be reinvested in AI. Of this, $100 billion is expected to trickle down into the semiconductor market, including memory.