Amazon Buys 2 Million Chips... Is It Really Good News for Semiconductors? "Why Is Wall Street Worried About NVIDIA?" | Industry Analyst Dr. Lee Ju-wan

Watch on YouTube ↗  |  August 30, 2026 at 03:00  |  20:11  |  815 Money Talk (815머니톡)
Speakers
Lee Ju-wan — Industry Analyst

Summary

Industry analyst Lee Ju-wan discusses Nvidia's results and AWS's reported 2 million GPU purchase with host Kim Tae-sung. He separates semiconductor supply and memory cycles from AI-bubble sentiment, flagging caution on Nvidia's credit-financed sales, Amazon's custom-chip weakness, and Big Tech AI capex sustainability. He also notes Samsung Electronics and SK Hynix near-term HBM orders are already reflected, sees NAND oversupply risks, and questions SK Hynix's ADR/buyback logic.

  • Nvidia's earnings surprise and $290 billion purchase commitments are discussed; immediate HBM order visibility is seen as already reflected in Samsung Electronics and SK Hynix.
  • Lee argues Samsung Electronics and SK Hynix risk is memory cycle supply, not whether AI is a bubble.
  • AWS's 2 million GPU purchase may become credit sales for Nvidia, creating collection and receivables cycle risk.
  • Amazon's large Nvidia purchase is read as a sign of weak in-house AI silicon competitiveness.
  • Big Tech and cloud AI capex is called bubble-like because final paying customers are insufficient.
  • Nvidia's credit and equity financial platform is viewed as a high-probability future earnings cushion, with valuation caveats.
  • NAND and enterprise SSD demand is seen rising, but NAND oversupply may make price declines faster than DRAM.
  • SK Hynix's ADR and share cancellation strategy is criticized as inconsistent.
Ideas
Lee Ju-wan Industry Analyst 0:00
Nvidia's credit sales create receivables cycle risk.
Lee says Nvidia has already secured a hidden next income stream: selling chips on credit to companies that cannot repay, then converting receivables into equity stakes. This financial-platform model can generate dividend and equity income, has a high success probability, and is already visible. Even if GPU sales slow, Nvidia has a future earnings cushion. However, investors may view rising financial income as negative if core growth slows.
Lee Ju-wan Industry Analyst 0:50
Watch memory cycle, not AI bubble.
Lee argues that the largest risk for Samsung Electronics and SK Hynix is memory price decline and a peaking memory cycle, not whether AI is a bubble. Memory prices rose because suppliers cut supply, not because demand was strong. Since memory earnings are supply-driven, an AI bubble verdict would not directly change their results, though investor sentiment can hit share prices. Investors should separate the memory cycle and supply from AI-demand sentiment.
Lee Ju-wan Industry Analyst 7:38
Amazon lacks custom AI chip competitiveness.
Lee says Amazon choosing to buy Nvidia GPUs while rivals develop their own chips indicates Amazon lacks in-house silicon capability. When the custom-silicon era takes off, Amazon's competitive position weakens, which raises the question of whether Amazon stock should be sold.
Lee Ju-wan Industry Analyst 8:21
Big Tech AI capex lacks paying customers.
Lee says cloud and Big Tech AI capex is not supported by final paying customers: roughly 90% of AI-related revenue is data-center rental, while only 10% is actual AI services. Companies are investing before cash generation, cash is depleting, and the bubble is growing. In one or two years, paying customers are unlikely to be sufficient to recover the investment. If Big Tech capex stops, Nvidia's growth stops too.
Lee Ju-wan Industry Analyst 18:09
NAND oversupply may hit SSD profitability.
Lee says enterprise SSD demand should rise alongside AI data-center storage, but NAND is likely to shift into oversupply faster than DRAM. China's NAND market share has already moved above 12%, and NAND has no HBM-like buffer; mobile and desktop NAND are even more vulnerable. NAND price declines could be much faster than DRAM, so SSD demand growth and memory makers' profitability can diverge.
Up Next

This 815 Money Talk (815머니톡) video, published August 30, 2026, features Lee Ju-wan discussing NVDA, 000660.KS, 005930.KS, AMZN, XLK, NAND Flash. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Lee Ju-wan  · Tickers: NVDA, 000660.KS, 005930.KS, AMZN, XLK, NAND Flash