NVIDIA’s Q2 revenue hits $96.2B, up +106% year-on-year; Data Center up 117%. Next-quarter guidance: $108B
Huang Renxun speaks directly
Chip sales will double next year—and the bottleneck isn’t demand, it’s capacity. Cloud providers have piled up orders worth over $2T. Even GPU rental prices have started to rise
Put it into plain language
AI demand isn’t “about to take off”—it already has taken off. Capacity can’t keep up.
My stance
The industry is bullish, and the market is moving in an “earnings validation” mode—not a broad, emotion-driven bull run.
Three pieces of evidence:
R&D chip-component stocks’ net profit in the mid-year report: +186%; semiconductor component lead times stretched from 4 months to 10 months; MetaMuse launches and gets 730,000 downloads in 5 days—demand is shifting from training to inference and the consumer side.
But here are two buckets of cold water that stay steady:
The probability of another rate hike by the Fed in October has risen to 53%, creating headwinds on the valuation side. A-share AI chip stocks have, on average, pulled back more than 20% since July—this rebound is an oversold rebound/repair, don’t treat it as a brand-new bull market.
I’m watching three lines for opportunities:
Domestic computing power substitution (policy + bid/one-click), the price-hike supply chain (power semiconductors, storage), and inference compute. After today, there are only two kinds of people—those who pick stocks by earnings, and those who chase headlines and buy at highs.
Position sizing matters more than direction
$NVDA $META
#AI股持续上涨还有哪些投资机会
Huang Renxun speaks directly
Chip sales will double next year—and the bottleneck isn’t demand, it’s capacity. Cloud providers have piled up orders worth over $2T. Even GPU rental prices have started to rise
Put it into plain language
AI demand isn’t “about to take off”—it already has taken off. Capacity can’t keep up.
My stance
The industry is bullish, and the market is moving in an “earnings validation” mode—not a broad, emotion-driven bull run.
Three pieces of evidence:
R&D chip-component stocks’ net profit in the mid-year report: +186%; semiconductor component lead times stretched from 4 months to 10 months; MetaMuse launches and gets 730,000 downloads in 5 days—demand is shifting from training to inference and the consumer side.
But here are two buckets of cold water that stay steady:
The probability of another rate hike by the Fed in October has risen to 53%, creating headwinds on the valuation side. A-share AI chip stocks have, on average, pulled back more than 20% since July—this rebound is an oversold rebound/repair, don’t treat it as a brand-new bull market.
I’m watching three lines for opportunities:
Domestic computing power substitution (policy + bid/one-click), the price-hike supply chain (power semiconductors, storage), and inference compute. After today, there are only two kinds of people—those who pick stocks by earnings, and those who chase headlines and buy at highs.
Position sizing matters more than direction
$NVDA $META
#AI股持续上涨还有哪些投资机会