Chamath Palihapitiya
· Chamath Palihapitiya
· 02 августа 2026, 18:28
· ⏱ 5 мин чтения
| Читать в Substack ↗
Резюме
The article argues that excessive leverage in both retail and institutional AI trades is exacerbating market volatility and triggering forced liquidations across global equities. Concurrently, a strategic fracture is emerging in AI governance, with hardware providers like Nvidia pushing for open-weight ecosystems while closed frontier labs lobby for regulatory moats to slow development.
•South Korea's KOSPI erased nearly 40% of its value from its peak, wiping out $2.18T in market value, exacerbated by retail crowding into 2x leveraged single-stock ETFs for Samsung and SK Hynix.
•China's largest DRAM producer, CXMT, surged 466% in its Shanghai debut, raising concerns that rising Chinese capacity will pressure conventional memory prices.
•South Korea announced a 20 trillion won (~$14 billion) sovereign wealth fund to support strategic domestic industries, including AI and chips.
•Nvidia and partners launched the Open Secure AI Alliance to defend open-weight models, contrasting with closed labs like OpenAI and Anthropic who are lobbying for international mechanisms to slow AI development.
•Leopold Aschenbrenner's AI hedge fund, Situational Awareness, faced severe margin calls due to carrying 4x gross exposure on $45B in assets, forcing the fire sale of its public-equity portfolio to Citadel.
SK Hynix is facing dual headwinds: cascading liquidation risk from retail-crowded 2x leveraged single-stock ETFs that 'add exposure after gains and cut it after losses,' and potential pricing pressure
SK Hynix is facing dual headwinds: cascading liquidation risk from retail-crowded 2x leveraged single-stock ETFs that 'add exposure after gains and cut it after losses,' and potential pricing pressure on conventional memory from China's CXMT.
Risk: The new 20 trillion won South Korean sovereign wealth fund could provide a structural bid for strategic domestic chip industries, offsetting retail ETF volatility.
The author notes that China's CXMT is rapidly scaling DRAM capacity, which 'raised concerns that rising Chinese capacity could pressure conventional memory prices.' This poses a direct margin headwind
The author notes that China's CXMT is rapidly scaling DRAM capacity, which 'raised concerns that rising Chinese capacity could pressure conventional memory prices.' This poses a direct margin headwind for global conventional memory suppliers like Micron.
Risk: Strong pricing power and supply tightness in advanced HBM (High Bandwidth Memory) for AI could completely offset weakness in conventional DRAM pricing.
Nvidia is aggressively defending open-weight AI models by launching the Open Secure AI Alliance and rallying 230+ signatories, strategically countering closed labs like Anthropic and OpenAI who are lo
Nvidia is aggressively defending open-weight AI models by launching the Open Secure AI Alliance and rallying 230+ signatories, strategically countering closed labs like Anthropic and OpenAI who are lobbying for development limits.
Risk: Regulatory capture by closed frontier labs (via the June 2 executive order and international mechanisms) could successfully slow frontier model training, reducing overall compute demand.
This newsletter, published August 02, 2026,
features Chamath Palihapitiya
discussing 000660.KS, MU, NVDA.
3 trade ideas extracted by AI with direction and confidence scoring.