Over the past 7 days, I have seen a flood of 'analysis' on Samsung's Q2 earnings. 89.4 trillion won operating profit. 171 trillion won revenue. A 10x explosion.
The numbers look delicious. The story writes itself: AI super-cycle, memory shortage, Samsung reclaiming its throne.
Here is the problem: those numbers are not real.
The actual Q2 numbers? Operating profit of 10.4 trillion won. Revenue of 74 trillion won. A difference of nearly 8x.
I spent 26 years in this industry. I have audited ICOs that promised the moon and delivered a rug. I have watched traders build entire thesis on a single bad data point. This smells exactly like that.
This is not a hit piece on Samsung. This is a warning about how narratives form in markets. And how smart money positions itself before the hype machine goes to maximum volume.
Let me break this down the way I break down a liquidity book. No fluff. Just the order flow.
The market doesn't lie. The P&L does.
Context: The Structure of the Story
Samsung Electronics is an IDM. It designs, manufactures, and sells its own chips. It dominates memory: ~42% of DRAM, ~33% of NAND. In a rising market, that scale is a super-weapon. In a falling market, it is a concrete block.
The narrative being sold to you is this: AI demand for HBM (High Bandwidth Memory) is so strong that it is pulling the entire memory industry into a 'super-cycle' that will last until 2027. Samsung, as the largest memory maker, is the prime beneficiary.
This is a half-truth. Which makes it more dangerous than a lie.
Let me give you the other half, based on what I actually see in the on-chain data and the order books I track for my own portfolio.
Core: The Order Flow You Are Not Seeing
Here is the most important chart in the semiconductor industry right now, and almost no one is looking at it: the HBM market share split.
In 2024, SK Hynix has ~50% of the HBM3E market. Samsung has ~30%. Micron has ~20%.
Samsung is not winning. Samsung is losing.
This is not a technical debate. It is a liquidity debate. The 'growth' in Samsung's memory profit is coming from a price recovery in standard DRAM and NAND. Not from HBM volume.
The real money in AI memory is flowing to SK Hynix. The narrative says Samsung is the AI memory play. The order flow says the smart money is buying SK Hynix.
I executed a trade on this in early 2020. I deployed capital into a yield farming strategy on Compound and Uniswap. I learned that on-chain mechanics behave differently than paper models. The paper model said Samsung was the safe bet. The real data said the fast, decisive mover was SK Hynix.
Let me show you the specific data points that support this:
- HBM3E Certification Timeline: Samsung's HBM3E was certified by NVIDIA in mid-2024. SK Hynix was shipping volume in Q1 2024. A 6-month lag in a market growing 100% YoY is a lifetime. In crypto terms, this is the difference between being first to farm a liquidity mining pool and being the last bag holder.
- Packaging Technology Defect: SK Hynix uses MR-MUF (Mass Reflow Molded Underfill). Samsung uses TC-NCF (Thermal Compression Non-Conductive Film). The industry consensus is that MR-MUF is superior for HBM3E. Samsung is now racing to develop Hybrid Bonding for HBM4, which is a major technology pivot. This is not a sign of strength. It is a sign of a forced move.
- Capacity Allocation: Samsung is spending ~50 trillion won on CapEx in 2024. SK Hynix is spending a smaller absolute number but a higher percentage of revenue. Efficiency matters. I run a trading desk. I know that the guy who spends the most capital is not always the one who makes the best returns. It is the guy who executes with the highest hit rate.
Based on my audit experience in 2017, I know that a team that is behind on technology and trying to force a catch-up is the most dangerous kind of investment. The code had vulnerabilities. The roadmap had wishful thinking. The market did not care.
Contrarian: The Blind Spot in the Room
Everyone sees the AI demand. No one is looking at the supply side.
Samsung, SK Hynix, and Micron are all expanding capacity aggressively. New HBM production lines take 6-9 months to build and 12-18 months to ramp.
Here is the math: - 2023 HBM industry supply: ~50 billion dollars - 2024 HBM industry supply: ~200 billion dollars - 2025 HBM industry supply: ~400 billion dollars (projected)
If demand grows at 100% YoY, and supply grows at 100% YoY, then the market does not get tighter. It stays exactly where it is.
The 'tightness' is not a function of demand. It is a function of the ratio of demand to supply. If the ratio is stable, the price does not spike. The margin does not explode.
I survived the 2022 Terra collapse because I refused to hold stablecoins in a single protocol. The same principle applies here: do not buy a narrative that assumes infinite demand and zero supply response. That is not a market. That is a cult.
The second blind spot is customer concentration. Samsung's HBM business is ~80% dependent on NVIDIA. That is a single point of failure. If NVIDIA decides to dual-source more aggressively (which it will, to reduce risk), or if SK Hynix undercuts on price (which it can, with better margins), Samsung's HBM revenue is directly at risk.
In crypto, we call this the 'whale risk.' One big player controls your liquidity. You are not in control. The whale is.
The market doesn't let you relax. It punishes the lazy.
Takeaway: The Only Signal That Matters
I do not trade on predictions. I trade on triggers.
Here are the price levels and data points you need to watch. If these triggers fire, you move. If they don't, you wait.
Bullish Trigger (Long Samsung): Samsung's Q3 2024 official earnings report shows HBM3E revenue exceeding 30% of total memory revenue, and gross margin for HBM exceeds SK Hynix's disclosed margin. This would indicate that the TC-NCF to Hybrid Bonding transition is working faster than expected.
Bearish Trigger (Short Samsung): Samsung's Q3 2024 official earnings report shows HBM3E revenue below 20% of total memory revenue, and operating profit growth is entirely driven by standard DRAM/NAND price hikes. This confirms that the HBM catch-up is failing.
Neutral/Volatility Trigger: A new US export control rule that restricts Samsung's China NAND fab (25% of its NAND capacity) or an announcement from NVIDIA that it has qualified Micron's HBM3E as a primary supplier. This would reset the competitive landscape.
My Position: I am not long Samsung. I am not short Samsung. I am short the narrative. The current price-to-earnings ratio of ~15x is not expensive, but it is not a discount. The cycle is middle-aged. The easy money in memory was made in late 2023 and early 2024.
I have one rule: do not be the last one to buy the thesis. Be the first one to see its expiration date.
Final Signal: If you see a headline that says 'Samsung secures exclusive multi-year HBM deal with NVIDIA,' that is the top signal. That is when the liquidity provider steps back, and the bag holder steps in.
I don't wait for confirmation. By the time the news is public, the trade is over.