PHPMem v2.0.1
Version
1.6.45
Uptime
17 days 10 hours 39 minutes 25 seconds
Memory
Total
512MB
Used
12,72MB (2.48%)
Free
499,28MB
Keys
Current
14 060
Total (since start)
40 994
Evictions
0
Reclaimed
760
Expired Unfetched
0
Evicted Unfetched
0
Connections
Current
15 / 1 024 max
Total
237 918
Rejected
0
llm:9d43ed0d14a12d1a355c8a5cfb8a15ae29249cab0f792b0b5212d609ef6dd430
Edit
**The most surprising finding is that the biggest price multiples did not come from the household-name AI chip designers. They came from obscure equipment makers in Japan and Korea, while Samsung, the region's giant, barely moved.**
I compared each stock's highest close with its lowest close since 2012. This ratio is currency-neutral, so the mix of USD, JPY, KRW and TWD prices doesn't distort it (steps 1–3).
- **NVIDIA leads, as expected.** It went from about 0.26 to 179, roughly 687×.
- **Lasertec is second at about 443×.** This Tokyo-listed chip-inspection equipment maker beat AMD (about 130×) and Broadcom (about 149×). Its close ran from about 100 to about 44,520 yen.
- **HANMI Semiconductor and Rorze follow.** HANMI (Korea) is at about 284× and Rorze (Japan) at about 283×, both ahead of Broadcom and AMD. Wolfspeed (about 355×) also sits in the top five.
- **Japanese equipment names show huge multiples.** Advantest is about 85×, Disco about 75× and Tokyo Electron about 50×.
- **Samsung is the laggard.** Samsung is the largest company in the set, yet its max/min close ratio is only about 5.6×. SK Hynix is about 17×, and Kokusai Electric about 3.3× (only 434 trading days of data).
Anyone expecting the AI surge to show up mainly in the famous chipmakers would be wrong. The widest multiples come from the "picks and shovels" suppliers of inspection, test and handling equipment.
**Caveats**
- A max/min close ratio depends on the start date and on where each stock began. Many of these stocks began 2012 at very low prices, so the multiples say little about recent performance.
- SK Hynix's maximum close is exactly 300,000.0, a suspiciously round number.
- Prices are in local currencies, so raw price levels can't be compared across companies.
- The data has quality issues. In 461 rows the close falls outside the day's high–low range, 4,888 rows have zero volume, and 6,840 rows are flat days (step 0).