PHPMem v2.0.1
Version
1.6.45
Uptime
15 days 23 hours 54 minutes 38 seconds
Memory
Total
512MB
Used
10,18MB (1.99%)
Free
501,82MB
Keys
Current
12 268
Total (since start)
35 898
Evictions
0
Reclaimed
738
Expired Unfetched
0
Evicted Unfetched
0
Connections
Current
14 / 1 024 max
Total
182 174
Rejected
0
llm:928d91be74b035ce06ab2ec88bb3216206c0fd864b0402da0a9acd2777552264
Edit
**The most counterintuitive finding: several of the "biggest winners" by price range are actually collapsed stocks, because their low came after their high.**
In a dataset framed around the AI surge, I ranked companies by max-close ÷ min-close. NVIDIA tops the list at about 687×, which fits the story. Wolfspeed comes third at about 355×, and Kalray sits in the top 15 at about 117×. Those two did not rise 100-fold.
The date of the peak and the date of the trough (step 8 and step 9 results) show the order was reversed:
- **Wolfspeed (WOLF):** peaked at about $141.87 on 2021-11-16 and hit its low of $0.40 on 2025-06-30. The last close of $1.52 is **1.1% of its peak**, a fall of about 99%.
- **Kalray (ALKAL.PA):** peaked at about 49.10 on 2021-02-08 and hit its low of 0.42 on 2025-05-20. The last close of 0.62 is **1.3% of peak**.
- **BrainChip (BRN.AX):** peaked at 2.13 in January 2022 and closed at 0.20, or **9.4% of peak**. Its 230× ratio mostly reflects a sub-cent low of 0.0093 in 2014, not a rally that held.
Compare the stocks where the low came first and the price stayed near the high:
- **NVIDIA:** low in 2012, peak on 2025-07-30, last close at **96.9% of peak**.
- **Broadcom:** low in 2012, peak on 2025-07-30, last close at **95.4% of peak**.
- **Alchip:** last close at 87.5% of peak.
Most other names in the ranking have also given back a lot since their 2024 peaks:
- Lasertec: 32.2% of peak.
- HANMI: 47.0% of peak.
- Rorze: 56.6% of peak.
- BE Semiconductor and Camtek: roughly 69% of peak.
- AMD: 81.2% of peak.
So the sector-wide "AI surge" is really a narrow set of winners, with NVIDIA and Broadcom still at or near their highs. Many others peaked in 2024 and have fallen well back. A few have nearly been wiped out.
**Caveats**
- A max/min ratio is not a return. It ignores which came first, so always check the peak and trough dates.
- Prices are in local currencies. The ratios are unitless, but the raw price levels are not comparable across exchanges.
- The data also has some quality quirks. 461 rows have a close outside the day's high–low range, and 4,888 rows have zero volume. The maximum close for SK Hynix is exactly 300,000.0, which looks like a round, possibly capped value.
**Evidence:** the ranking by max/min ratio in step 7, the peak and trough dates with last close as a share of peak in steps 8 and 9, and the quality checks in step 0.