Skip to content
QuantGPT
← All datasets

Fundamentals and estimates

Point-in-time fundamentals

QuantGPT

Quarterly income, balance sheet and cash flow, each dated by when it became public.

Use in my models: choose in My dataAll my data choices

What it is

Quarterly income, balance sheet and cash flow, each dated by when it became public.

  • fundamentals point in time: Quarterly statements point in time (income, balance sheet, cash flow, consensus and surprise), joined by filing date. Three stray rows (1899 for ASND, 1969, 1972) make the empty years; real coverage starts in 1992 (124 rows), 1998 (7,414) and is fullest 2020 to 2025 (23,000 to 27,000 a year).
  • fundamentals point in time versions: Earlier versions of rows that were restated, with when each was seen.

Source and licence

Source: EODHD statements, SEC filing dates.

Licence: licensed vendor (SEC is the public twin). This is our reading of the table; counsel decides.

Coverage and gap years

9,969 names, 490,245 rows in fundamentals point in time, from Sep 30, 1899 to Aug 31, 2026.

Years with no rows: 1900 to 1968, 1970 to 1971, 1973 to 1989.

Thin years: 1969, 1972, 1990 to 1994.

TableRowsFirstLast
fundamentals point in time490,245Sep 30, 1899Aug 31, 2026
fundamentals point in time versions23,682Jan 31, 1993Jul 31, 2026

Freshness

Last row Aug 31, 2026: 39 days behind on the day measured. Updates: as companies report.

Point in time

Keyed on the filing date (the earnings release day), not the period end, so a backtest sees a quarter only once it was public.

Kept: when a figure is restated, the earlier version stays in the restatement store (fundamentals point in time versions) with the day it was seen. It holds 23,682 earlier versions for 940 names, for periods from Jan 31, 1993 to Jul 31, 2026. The main table carries the newest.

Delisted names

Companies that later delisted are in from 1998; before that the history is survivors only, so early years lean toward the winners.

Known issues

  • fundamentals point in time is quarterly only, with one stray 1899 row (ASND). Depth: about 120 firms in 1992 and 2,100+ from 1998 (the engine's own capabilities map). (WAREHOUSE, Gaps and freshness that matter)
  • Net income: 44 of 44 exact. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • Revenue: AAPL 9 of 9, MSFT 9 of 9, NVDA 9 of 9 exact. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • XOM revenue is 2.3% to 3.3% under SEC in all 8 quarters. The engine carries sales and other operating revenue; SEC's Revenues tag is total revenues and other income (it adds equity-affiliate and other income). Example: 2026-03-31 engine 83,161 against SEC 85,138. The engine shows 81,058 for both 2024-12-31 and 2025-03-31: the same figure two quarters running. SEC's totals for those quarters are 83,426 and 83,130, so one of the two engine rows is likely carried forward. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • JPM revenue is gross, not net. SEC has no quarterly Revenues tag for JPM. The engine's 2025-03-31 figure, 68,907, is noninterest revenue (22,037) plus gross interest income. JPM's net revenue (RevenuesNetOfInterestExpense) that quarter is 45,310, so the engine is +52% against the figure JPM reports. Any margin, P/S or growth figure for a bank inherits this. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • EPS: EPS diluted is the vendor's reported EPS, not GAAP. It equals earnings calendar EPS actual row for row. Against SEC GAAP diluted EPS: The column EPS GAAP diluted (net income over diluted shares) tracks GAAP: NVDA 2.391 against 2.39, XOM 1.000 against 1.00. JPM reads 7.572 against 7.70, off only because its share count is stale (below). Use EPS GAAP diluted wherever a page says EPS without saying adjusted. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • Shares (shares diluted) carry forward and mix fields. JPM reads 2,793,700,000 in five straight quarters (2025-06 to 2026-06). SEC's 10-Q figures: 2,793.7M (2025-06), 2,767.6M (2025-09), 2,720.2M (2026-03) and 2,694.2M (2026-06); the 10-K gives no fourth-quarter figure. The engine is +3.7% in the latest quarter. XOM 2025-03-31 reads 4,462M (SEC 4,372M, +2.1%; the same value as 2024-09-30). XOM 2025-09-30 reads 4,331M (SEC 4,285M, +1.1%; the same value as the quarter before). XOM 2026-03-31 reads 4,183M, which is that quarter's net income (SEC shares 4,202M). Across the warehouse, 63 rows have shares diluted exactly equal to net income. The newest are DCI 2026-04-30; XOM, HMN and JCI 2026-03-31; and HWM 2025-06-30. AAPL's and MSFT's fourth-quarter rows carry the full-year weighted average (15,004,697,000 and 7,453,000,000), not the quarter's. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • Share errors reach market cap. prices daily shares out is the latest shares diluted, and market cap is raw close times that figure. JPM on 2026-10-08: 2,793.7M x 331.42 = 925.9B. With SEC's latest 2,694.2M it would be 892.9B (+3.7%). (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • Dates and labels. period end is moved to the calendar month end: AAPL 2025-09-30 for 2025-09-27, NVDA 2026-01-31 for 2026-01-25. fiscal year and fiscal quarter are calendar labels, not the company's fiscal ones: AAPL's 2025-09-30 row reads 2025 Q3; Apple calls it FY2025 Q4. NVDA's 2026-01-31 row reads 2026 Q1; NVIDIA calls it FY2026 Q4. MSFT's 2026-06-30 row reads 2026 Q2; Microsoft calls it FY2026 Q4. filing date is the earnings release date, not the SEC filing date. MSFT's and NVDA's match the SEC date. AAPL's is 1 day before the 10-Q or 10-K. XOM's is 3 days before the 10-Q (19 days before the 10-K), and JPM's is 17 to 31 days before. The release is public, so this is not look-ahead, but the column is named for the filing. 44,737 rows (9.1%) carry an imputed filing date (filing date imputed). The engine's own capabilities map says about 9%. Three stray rows predate 1990: ASND with period end 1899-09-30 and no values, and one row each in 1969 and 1972. Real coverage starts in 1992 (124 rows) and passes 7,000 rows a year from 1998. 7,179 rows are flagged currency suspect; 30,588 were converted from another currency (FX converted). (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))
  • SEC companyfacts had no 2026-06-30 quarter for XOM on the day of the check, so that engine row (revenue 114,529, net income 14,525) is unchecked. (ACCURACY, Quarter by quarter (2024-06 to 2026-07, 44 company quarters with an SEC figure))

Where it is used

Engine tools that reach it (your AI, over MCP):

  • financial_model A company's financial model from the point-in-time statements: TTM figures, complete years, growth rates, margins, a DCF (growth for five years fading to the terminal rate by year ten, a Gordon...
  • get_data Standard warehouse series without SQL.
  • query_warehouse Run read-only SQL (DuckDB dialect) directly against any research database.

Ask AI

Every count and date here was measured on Oct 9, 2026 from the QuantGPT engine's warehouse and is read from the catalogue file when the site is built. Metadata only: no vendor rows are shown.