Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Dark pool data becomes more useful when it is treated as a record of transactions rather than a shortcut to a market call.
A 90-day review of publicly reported tape data for NVDA, SPY, and AAPL shows how different securities distribute large off-exchange prints across the trading day. The sample covers June 9 through September 7, 2026, although the latest trading records in the pull run through September 4. The review includes 1,895 prints worth about $181.1 billion after filters were applied.
The result is a map of where size traded, when it traded, and how the pattern differed between two individual stocks and a broad-market ETF. It does not identify a buyer or seller. The feed does not provide bid/ask side or directional intent, so the analysis does not assign either to a print.
The data set
The pull covered three symbols:
- NVDA as the semiconductor and artificial-intelligence-related stock in the sample.
- SPY as the broad-market ETF comparison.
- AAPL as a widely held individual stock with a different trading profile from NVDA.
The source was MobyTick’s historical prints endpoint. Each record included the ticker, timestamp, execution price, share count, notional value, and available trade-condition information. The pull used a minimum size of 50,000 shares and retained closing-auction records for review.
For the published intraday comparisons, the analysis used regular-session records from 9:30 a.m. through 4:00 p.m. Eastern Time. It removed prints within 0.3% of that day’s regular-session close. That step matters because closing activity can overwhelm an intraday view without telling the reader much about when size traded during the session. Prints above 15 million shares were also excluded from the cited intraday set. None of the three filtered samples required that second exclusion.
The final counts were:
| Symbol | Prints reviewed | Notional value | Trading days represented | Median print value |
|---|---|---|---|---|
| NVDA | 691 | $38.8 billion | 60 | $24.5 million |
| SPY | 844 | $109.7 billion | 54 | $94.5 million |
| AAPL | 360 | $32.6 billion | 59 | $40.9 million |
These are aggregated observations from the selected sample. They are not estimates of total market-wide dark pool activity.
SPY carried the largest notional footprint
SPY accounted for about $109.7 billion of the filtered notional value, more than NVDA and AAPL combined. That is not surprising for an ETF with broad market exposure, but the size of the difference is useful context: the largest print count does not always belong to the most familiar individual stock.
SPY also had the highest median print value at roughly $94.5 million. NVDA’s median was about $24.5 million, while AAPL’s was about $40.9 million. The comparison shows why raw print counts and notional value should be reported together. A ticker can produce many prints without producing the largest dollar footprint, and an ETF can show a much higher dollar median even when its share counts do not look extraordinary beside an individual stock.
SPY’s retained sample carried the largest aggregate footprint at $109.7 billion, with a median print value of approximately $94.5 million. Its activity was concentrated in the afternoon and final-30-minute buckets, which contributed $41.5 billion and $31.2 billion respectively. Those are sample-level distributions, not evidence of who traded or what price will do next.
NVDA’s activity was concentrated in the middle of the day
NVDA produced 691 retained prints totaling $38.8 billion. The middle portion of the session, defined here as 10:30 a.m. through 12:59 p.m. Eastern, accounted for $16.7 billion, or about 43% of the retained notional value. The afternoon window from 1:00 p.m. through 3:29 p.m. contributed another $12.0 billion.
NVDA’s retained sample totaled $38.8 billion across 691 prints and 60 trading days. The middle-session bucket contributed $16.7 billion, or about 43% of retained notional value, while the afternoon bucket contributed $12.0 billion. These aggregates describe the sample’s timing and size distribution without exposing an individual print or assigning intent.
NVDA’s median print size was 115,000 shares, and the 90th-percentile print size was 700,000 shares. That gives readers a better sense of the distribution than the largest transaction alone. A single large record can attract attention, but the median and percentile figures describe what the sample looked like across the full window.
AAPL sat between the two profiles
AAPL produced 360 retained prints totaling $32.6 billion across 59 trading days. Its median print value was approximately $40.9 million, higher than NVDA’s but well below SPY’s. The 90th-percentile AAPL print size was 817,000 shares.
AAPL’s retained sample totaled $32.6 billion across 360 prints and 59 trading days. Its middle-session bucket accounted for $15.4 billion, or about 47% of retained notional value, and its afternoon bucket contributed $11.6 billion. The figures describe historical concentration in the selected sample; they do not identify a participant or predict a future move.
AAPL’s middle-session window accounted for $15.4 billion, or roughly 47% of retained notional value. Its afternoon window contributed $11.6 billion. The distribution was more concentrated in the middle of the day than in the first and last 30 minutes, even after the close-proximity filter was applied.
Time of day changes the picture
The time buckets make the cross-symbol differences easier to see:
| Symbol | 9:30–10:29 | 10:30–12:59 | 1:00–3:29 | 3:30–4:00 |
|---|---|---|---|---|
| NVDA | $6.2B | $16.7B | $12.0B | $3.9B |
| SPY | $7.9B | $29.1B | $41.5B | $31.2B |
| AAPL | $3.5B | $15.4B | $11.6B | $2.1B |
SPY’s last-30-minute bucket remained large even after records close to the regular-session close were removed. That reflects the ETF’s broad notional scale and the way large transactions can cluster late in the session. NVDA and AAPL showed smaller late-session totals relative to their full samples.
The table is a descriptive view, not a trading rule. It can help a reader ask better questions about a symbol’s historical activity: Does size appear throughout the session? Is the sample dominated by one time window? Are the biggest records isolated events or part of a wider distribution? Those are answerable questions. Predicting the next move from the existence of a print is a different claim, and this data set cannot support it by itself.
How to read the result
Historical dark pool prints can add context to a market review when the analysis keeps three boundaries in place.
First, report the tape fields that are actually present: timestamp, price, shares, notional value, ticker, and trade conditions. Do not add buyer, seller, side, or intent when the source does not provide them.
Second, separate raw activity from the filtered view. Closing records, extended-hours records, and very large transactions can change the shape of an intraday chart. State the filters so another reader can reproduce the comparison.
Third, describe where size traded without turning that description into a promise about what price will do next. A cluster is a historical observation. It is not a recommendation, a guarantee, or proof of a forecast.
This sample gives NVDA, SPY, and AAPL three distinct historical profiles: NVDA and AAPL concentrated much of their retained notional value in the middle of the session, while SPY carried a much larger dollar footprint and more late-session activity. That is the practical value of a data-native review. It shows the structure in the records and leaves the unsupported certainty out.
Method note: Source data came from MobyTick’s historical prints endpoint. The published comparison used regular-session records from June 9 through September 7, 2026, with the latest available records through September 4. Prints within 0.3% of the same-date regular-session close and prints above 15 million shares were excluded from the cited intraday set. Figures are rounded; the aggregates were recomputed from the source data set for verification.
