What a natural diamond actually costs — the wholesale benchmark the trade pays, the asking price the counter shows, and the spread between them, across ten shapes and eleven weight bands. Every figure is dated, every source is named, and every number we modelled rather than observed says so.
Refreshed 2026-10-26 (Monday dual-basis refresh, both legs moved together in one step, per the chief's 2026-09-09 per-section dating rule): the IDEX price-driver average for Round, 4/4 (1.00-1.49), D-K/IF-I1 read 3,018.98 today against 3,025.18 on 2026-10-19 (page's own table, no timestamp, hourly refresh; fetched 2026-10-26), read direct from https://www.idexonline.com/diamond_prices_index at 19,717 bytes (newsroom/sources/2026-10-26/wire/idex.html; every IDEX figure used on this page confirmed by an exact-digit grep of that saved file, hits: 0.50 ct 1359.28: 1, 0.70 ct 1499.48: 1, 0.90 ct 1877.59: 1, 1.00 ct 3018.98: 1, 1.50 ct 4536.48: 1, 2.00 ct 8752.92: 1, 3.00 ct 16127.08: 1, 4.00 ct 32277.43: 1, 5.00 ct 21448.93: 1). RAPI's newest published point read 3,876.42 dated 2026-10-25 against 3,878.11 on 2026-10-24 on the index's own daily series (https://rapaport.com/diamond-price-index/, door.py 200, 220,266 bytes, newsroom/sources/2026-10-26/wire/rapaport.html; exact-digit grep: 927.08 1, 1291.31 1, 3876.42 1, 18264.96 1); the page's own pct field was not used. tools/rapi_fetch.py could not open the page this run (Python SSL protocol error, not an HTTP refusal), so the tool's own parser was run on the curl-saved bytes. Both bases, every IDEX-sourced row in the weights table, the dual-basis table, the top-level spread table and the RAPI context rows were brought onto today's read in the same step (0.30 ct RAPI 927, 0.50 ct 1359, 1.00 ct 3019, 3.00 ct 16127; dual-basis RAPI legs 0.30 ct 927, 0.50 ct 1291, 1.00 ct 3876, 3.00 ct 18265; IDEX-against-RAPI gaps recomputed as 1,359.28/1,291.31 = +5.3%, 3,018.98/3,876.42 = 77.9% so -22.1%, 16,127.08/18,264.96 = 88.3% so -11.7%). Year-on-year legs recomputed against the same 1 January and 1 September 2025 baselines this page already carries (carried from earlier reads, not in the page's current 31-day window, so not re-read this run): 0.30 ct 927.08/845.11 = +9.7% and 12.4% under September 2025; 1.00 ct 3,876.42/4,125.20 = -6.0% on the year and 12.9% under September 2025; 3.00 ct 18,264.96/18,225.61 = +0.2% on the year. The two 'magic weights' ratios were recomputed on today's levels: 0.30 to 0.50 ct RAPI 1,291.31/927.08 = +39%, and 1.00 to 2.00 ct IDEX 8,752.92/3,018.98 = +190%. The 4.00 ct band stays withheld (IDEX 32,277 against 21,449 for 5.00 ct on a 0.32% sample). This is a refresh, not a correction. Retail ($5,454, Blue Nile, dated 2026-07-30), the Rapaport List figure ($8,600) and the shapes, colour and clarity tables keep their own older dates and were not touched this run; the one-carat counter-over-trade spread moved from 80% to 81% only because its trade denominator moved (5,454/3,019 = 1.807). The counter price is 81% above the trade benchmark and about 37% below Rapaport list. Earlier history of this note, newest first: Refreshed 2026-10-19 (Monday dual-basis refresh, both legs moved together in one step, per the chief's 2026-09-09 per-section dating rule): the IDEX price-driver average for Round, 4/4 (1.00-1.49), D-K/IF-I1 read 3,025.18 today against 3,029.75 on 2026-10-12 (page's own table, no timestamp, hourly refresh; fetched 2026-10-19), read direct from https://www.idexonline.com/diamond_prices_index at 19,728 bytes (newsroom/sources/2026-10-19/wire/idex_diamond_prices_index.html; every IDEX figure used on this page confirmed by an exact-digit grep of that saved file, hits: 0.50 ct 1356.92: 1, 0.70 ct 1494.66: 1, 0.90 ct 1879.31: 1, 1.00 ct 3025.18: 1, 1.50 ct 4533.58: 1, 2.00 ct 8710.89: 1, 3.00 ct 16123.19: 1, 4.00 ct 33124.56: 1, 5.00 ct 21531.97: 1). RAPI's newest published point read 3,891.97 dated 2026-10-18 against 3,891.67 on 2026-10-17 on the index's own daily series (https://rapaport.com/diamond-price-index/, 200 on both identities, 228,431 bytes, newsroom/sources/2026-10-19/wire/rapi.html; exact-digit grep: 944.13 1, 1299.25 1, 3891.97 1, 18191.3048 1); the page's own changePct field was not used. Both bases, every IDEX-sourced row in the weights table, the dual-basis table, the top-level spread table and the RAPI context rows were brought onto today's read in the same step (0.30 ct RAPI 944, 0.50 ct 1357, 1.00 ct 3025, 3.00 ct 16123; dual-basis RAPI legs 0.30 ct 944, 0.50 ct 1299, 1.00 ct 3892, 3.00 ct 18191; IDEX-against-RAPI gaps recomputed as 1,356.92/1,299.25 = +4.4%, 3,025.18/3,891.97 = 77.7% so -22.3%, 16,123.19/18,191.30 = 88.6% so -11.4%). Year-on-year legs recomputed against the same 1 January and 1 September 2025 baselines this page already carries: 0.30 ct 944.13/845.11 = +11.7% and 10.7% under September 2025; 1.00 ct 3,891.97/4,125.20 = -5.7% on the year and 12.5% under September 2025; 3.00 ct 18,191.30/18,225.61 = -0.2% on the year. The two 'magic weights' ratios that stood from earlier refreshes were recomputed on today's levels: 0.30 to 0.50 ct RAPI 1,299.25/944.13 = +38%, and 1.00 to 2.00 ct IDEX 8,710.89/3,025.18 = +188% (the page had carried +37% and +193%, which no longer reconcile with the levels beside them). The 4.00 ct band stays withheld (IDEX 33,125 against 21,532 for 5.00 ct on a 0.32% sample). This is a refresh, not a correction. Retail ($5,454, Blue Nile, dated 2026-07-30), the Rapaport List figure ($8,600) and the shapes, colour and clarity tables keep their own older dates and were not touched this run. The counter price is 80% above the trade benchmark and about 37% below Rapaport list. Earlier history of this note, newest first: Refreshed 2026-10-12 (Monday dual-basis refresh, both legs moved together in one step, per the chief's 2026-09-09 per-section dating rule): the IDEX price-driver average for Round, 4/4 (1.00-1.49), D-K/IF-I1 read 3,029.75 today against 3,048 on 2026-10-05, read direct from IDEX's live driver table at 19,741 bytes on 2026-10-12 (newsroom/sources/2026-10-12/idex.html; every one of the nine driver figures used on this page confirmed by an exact-digit grep of that saved file, one hit each). RAPI's newest published point read 3,903.60 dated 2026-10-12 against 3,903.03 on 2026-10-09, read from the index's own daily series via tools/rapi_fetch.py. Both bases, every IDEX-sourced row in the weights table below, AND the dual-basis table were brought onto today's read in the same step (0.30 ct RAPI 950, 0.50 ct 1367, 1.00 ct 3030, 3.00 ct 16061; dual-basis RAPI legs 0.30 ct 950, 0.50 ct 1312, 1.00 ct 3904, 3.00 ct 18257). This is a refresh, not a correction: the dual-basis table had honestly labelled its own older basis, and every figure it carried was correct for the date beside it. Retail ($5,454, Blue Nile, dated 2026-07-30), the Rapaport List figure ($8,600) and the shapes table keep their own older dates and were not touched this run. The counter price is 80% above the trade benchmark and about 37% below Rapaport list. Refreshed 10 October 2026 (RAPI leg only): every RAPI figure on this page has been moved, in one step, onto the index's newest published daily point, 2026-10-09, read this morning via tools/rapi_fetch.py (200) and cross-checked against newsroom/tape/2026-10-10.json, which carries the same 3,903.03. Headline 3,903.03; dual-basis 0.30 ct 951, 0.50 ct 1,310, 1.00 ct 3,903, 3.00 ct 18,172; the 0.30 ct weights row 951 a carat and 285 a stone with its retail spread recomputed as 2,015/951 = 2.12 so +112%; the IDEX-against-RAPI gaps recomputed on the new levels as 1,370/1,310.35 = 104.6% of RAPI so +4.6%, 3,048/3,903.03 = 78.1% so -21.9%, and 16,179/18,172.26 = 89.0% so -11.0%. The year-on-year legs were recomputed against the same 1 January and 1 September 2025 baselines this page already carries, not derived from the rounded percentages that stood beside them: 0.30 ct 951.27/845.11 = 112.6% so +12.6% on the year and 951.27/1,057.85 = 89.9% so 10.1% under September 2025; 1.00 ct 3,903.03/4,125.20 = 94.6% so -5.4% on the year and 3,903.03/4,448.65 = 87.7% so 12.3% under September 2025; 3.00 ct 18,172.26/18,225.61 = 99.7% so -0.3% on the year. The home-page natural-to-lab ratio card also stopped claiming, as it had on every build since at least 29 September, that both of its legs were "as carried on this morning's tape"; it now prints each leg's own date, read from this file and from lab-prices.json at build time. This is a refresh, not a correction: 3,906.01 was the right level for 2026-10-07 and was correctly dated beside its own date throughout, and the -5.3% year-to-date figure was correct against the 2026-10-07 point. The IDEX price-driver column and its 2026-10-05 date, the retail median and its 2026-07-30 date, the shapes, colour and clarity tables and the Rapaport list figure are untouched by this refresh. Corrected 7 October 2026: until this morning this page stated in the present tense that "RAPI's own newest published point is dated 2026-10-05" and carried the 1.00 ct index at 3,909.96. That sentence was false when read today: the index's own daily series had published points for 2026-10-06 (3,910.30) and 2026-10-07 (3,906.01), and this paper's own tape had already captured the 6 October point on 6 October. Every RAPI leg on this page has now been moved onto the 2026-10-07 point in a single step - headline 3,906.01, dual-basis 0.30 ct 956, 0.50 ct 1,314, 1.00 ct 3,906, 3.00 ct 18,208, the 0.30 ct weights row 956 a carat, the year-to-date figure -5.3%, and the two RAPI context rows - read from the index's own embedded daily series at 304,515 bytes on 7 October 2026 and confirmed by raw-byte grep (3906.01 one hit, 956.24 two, 1313.85 two, 18207.54 three). Where the bad claim came from: the page was refreshed on 5 October and the present-tense "newest" sentence was not re-checked on 6 or 7 October, so the claim aged into falsehood without any figure changing. Correct as published throughout: 3,909.96 was the right level for 2026-10-05 and was correctly dated beside its own date; the as_of field was honest; the IDEX price-driver column and its 2026-10-05 date, the retail median and its 2026-07-30 date, the shapes, colour and clarity tables and the Rapaport list figure were all right as printed and are untouched by this correction. Refreshed 5 October 2026 (Monday dual-basis refresh, per the chief's 2026-09-09 per-section dating rule): the IDEX price-driver average for Round, 4/4 (1.00-1.49), D-K/IF-I1 read 3,048.14 today against 3,037 on 29 September, read direct from IDEX's live driver table at 19,703 bytes on 5 October 2026; RAPI's newest published point read 3,909.96 dated 2026-10-05 against 3,911.50 on 2026-10-04, read from the index's own daily series via tools/rapi_fetch.py. Both bases, every IDEX-sourced row in the weights table below, AND the dual-basis table were brought onto today's read in the same step (0.30 ct RAPI 959, 0.50 ct 1,315, 1.00 ct 3,910, 3.00 ct 18,148, parsed out of Rapaport's own embedded daily series at 306,808 bytes on 5 October 2026; IDEX legs 1,370 / 3,048 / 16,179 from the same driver read, with the gaps recomputed as 1,370/1,315.27 = +4.2%, 3,048/3,909.96 = 77.95% of it so -22.0%, and 16,179/18,148.20 = 89.15% of it so -10.9%). This is a refresh, not a correction: the dual-basis table had honestly labelled its own older basis, and every figure it carried was correct for the date beside it -- trade_ct against trade_date, rapi_ct against rapi_date, never one without the other. Retail ($5,454, Blue Nile, dated 2026-07-30), the Rapaport List figure ($8,600) and the shapes table keep their own older dates and were not touched this run. The counter price is 79% above the trade benchmark and still about 37% below Rapaport list. Both things are true at once: the retail margin is real, and the list itself has stopped being the price anyone pays. Trade benchmark dated 2026-10-05 (IDEX); RAPI's own newest published point is dated 2026-10-07, the index having moved from monthly to daily publication on 2026-09-07 - dated to the point, not to the day it was read. Until this morning this page read 2026-09-29 and 2026-10-04 respectively, and 80% rather than 79%; both were correct for the 29 September trade basis (5,454/3,037 = 79.59% above it) and the 79% is the same division against today's 3,048 (5,454/3,048 = 78.94% above it). Nothing on this page was wrong as published; the two sentences simply had to move with the basis in the same step. The 2026-10-04 point reads 3,911.50 and replaces the 3,912.48 at 2026-10-02 that this page carried before; both were correct for their own dates and neither is a correction of the other. Refreshed 4 October 2026: the 1.00 ct RAPI level and its date, the 3.00 ct dual-basis level (18,208 to 18,138, its IDEX gap moving from -10.9% to -10.5% on 16,230/18,138 = 89.48% of it) and the Rapaport source date were advanced to the newest published point, read direct from the index's own embedded daily series at 305,250 bytes on 4 October 2026. The 0.30 ct (960), 0.50 ct (1,315) and 1.00 ct (3,912) dual-basis levels are unchanged at this page's rounding and were re-read, not assumed. This is a refresh, not a correction: every 2 October level was correct for 2 October. Retail figure dated 2026-07-30, James Allen dropped from the retailer set - see the correction note. Corrected 23 September 2026: this page carried "RAPI's own newest published point is dated 2026-09-16" and a RAPI figure of $3,912.00; the index's own 1.00 ct series had published daily points past that date and read 3,914.90 on 22 September. The RAPI figure and its date were fixed on 23 September but no correction note was printed with the fix, which this note now supplies. Correct as published throughout: the trade benchmark, its 2026-09-21 IDEX date, the retail median and the 2026-07-30 retail date. Corrected 3 October 2026: earlier on 3 October this page carried two different claims about which RAPI point is newest. Its headline block had been moved to the 2026-10-02 point (3,912.48) while the weights table, the dual-basis table, the method note and the Rapaport source row were all still dated 2026-09-28, and the 0.30-carat trade figure (968 a carat) and the dual-basis RAPI levels (1,317 at half a carat, 3,915 at one, 18,092 at three) were still the 28 September ones. RAPI's newest published point was 2026-10-02 on that date, and every RAPI figure on this page was brought onto it: 0.30 ct 960, 0.50 ct 1,315, 1.00 ct 3,912, 3.00 ct 18,208. The page was then refreshed to the 2026-10-04 point and, this morning, to the 2026-10-05 point, both described above. The fault was a partial edit by this paper's own editing desk, not a wrong figure from the source. Correct as published throughout: every 28 September RAPI level was correct for 28 September, and the IDEX price-driver column, the retail median and all of their dates are untouched and unaffected.
Trade column is the IDEX price-driver average where a driver exists, RAPI where it does not. Retail is the observed five-retailer median at one carat and a declining multiple of the trade benchmark elsewhere — stated, not hidden.
| Weight | Trade / ct | Trade, stone | Retail / ct | Retail, stone | Spread | Basis |
|---|---|---|---|---|---|---|
| 0.30 ct | $927 | $278 | $2,015 | $605 | +113% | Modelled |
| 0.50 ct | $1,359 | $680 | $2,781 | $1,390 | +105% | Modelled |
| 0.70 ct | $1,499 | $1,049 | $2,941 | $2,059 | +97% | Modelled |
| 0.90 ct | $1,878 | $1,690 | $3,595 | $3,236 | +91% | Modelled |
| 1.00 ct | $3,019 | $3,019 | $5,454 | $5,454 | +81% | Observed |
| 1.50 ct | $4,536 | $6,804 | $7,866 | $11,800 | +73% | Modelled |
| 2.00 ct | $8,753 | $17,506 | $14,055 | $28,109 | +61% | Modelled |
| 3.00 ct | $16,127 | $48,381 | $23,504 | $70,513 | +46% | Modelled |
| 4.00 ct | — | — | — | — | — | Withheld |
| 5.00 ct | $21,449 | $107,245 | $29,035 | $145,175 | +35% | Modelled |
| 10.00 ct | — | — | $37,474 | $374,740 | — | Retail only |
The 4.00-4.99 ct band is the one figure on this page we decline to print. IDEX carries it at $32,277 a carat — above the five-carat band at $21,449 — on a thin slice of the market. A four-carat stone does not cost more per carat than a five-carat stone. The band is too thin to mean anything, so it stays blank rather than being smoothed into something that looks like a price.
No public source publishes a wholesale per-carat figure for ten carats. IDEX carries no driver at that weight, RAPI publishes no level, and the independent Antwerp grids stop at six. The ten-carat line is a live retail asking average and is marked as such.
Markup falls as the stone grows. A third of a carat carries roughly double the trade price; a five-carat carries about a third on top. The small stone pays for the shop.
| Evidence | Reading |
|---|---|
| Rapaport's own guidance | most diamonds trade 10–30% below list |
| Signet gross margin | 39.5% · FY2026 · implies 1.65× |
| Brilliant Earth gross margin | 57.5% · FY2025 · implies 2.35× |
| Blue Nile gross margin | 19.2% · last disclosed · implies 1.24× |
| Weight | RAPI · D–H / IF–VS2 | IDEX · market average | Gap |
|---|---|---|---|
| 0.30 ct | $927 | no driver | — |
| 0.50 ct | $1,291 | $1,359 | +5.3% |
| 1.00 ct | $3,876 | $3,019 | -22.1% |
| 3.00 ct | $18,265 | $16,127 | -11.7% |
RAPI is the average asking price of the ten per cent lowest-priced D–H, IF–VS2, GIA-graded stones on RapNet. It is a floor for a narrow top specification, not a market mean. IDEX price drivers average a far broader spread of colour and clarity. At one carat the broad market sits 22.1% below the narrow top spec (3,019 / 3,876 = 77.9% of it), which is what you would expect — and a useful reminder that a single 'diamond price' does not exist.
Ten cuts, one carat, against the round brilliant — live asking prices, 169,324 stones
| Shape | Asking / ct | Against round | Cross-source range | Family |
|---|---|---|---|---|
| Round brilliant | $4,581 | baseline | baseline | Round |
| Pear | $3,798 | -17.1% | conflicted | Elongated brilliant |
| Marquise | $3,637 | -20.6% | conflicted | Elongated brilliant |
| Oval | $3,546 | -22.6% | −20 to −24% | Elongated brilliant |
| Emerald | $2,822 | -38.4% | −35 to −38% | Step / square |
| Radiant | $2,785 | -39.2% | −33 to −39% | Step / square |
| Princess | $2,776 | -39.4% | −32 to −39% | Step / square |
| Cushion | $2,753 | -39.9% | −40% | Step / square |
| Asscher | $2,694 | -41.2% | −35 to −41% | Step / square |
| Heart | $2,616 | -42.9% | −30 to −43% | Step / square |
Pear and marquise are the two shapes we will not price to a single number. Live inventory at one house puts pear 44% below its own round; another retailer's published differential puts it 18.7% above. That is not rounding — it is two different books. Read the shape tier, not the decimal.
Roughly a third of rough is lost cutting one. You pay for the waste.
Oval, pear, marquise. Cut with the grain of the rough, so less is thrown away — and they look larger face-up than their weight.
Emerald, asscher, radiant, princess, cushion, heart. Cheapest per carat, and the least forgiving: an open table shows every inclusion the round would have hidden.
Weight tells you how big. Shape tells you what it looks like. Colour and clarity tell you what it costs — and between them they move the price of a one-carat round by more than a factor of two in each direction.
| Colour | Index | Applied / ct | |
|---|---|---|---|
| D | 176 | $8,135 | Colourless |
| E | 154 | $7,118 | Colourless |
| F | 139 | $6,425 | Colourless |
| G | 118 | $5,454 | Near colourless |
| H | 100 | $4,622 | Near colourless · index base |
| I | 87 | $4,021 | Near colourless |
| J | 74 | $3,420 | Near colourless |
| K | 66 | $3,051 | Faint |
A K costs 62.5% less per carat than a D at the same weight, shape and clarity. An independent Antwerp grid puts the same swing at 61.6% — close enough, from a different desk, to trust the shape of the curve.
| Clarity | Index | Applied / ct | |
|---|---|---|---|
| IF | 171 | $8,181 | Internally flawless |
| VVS1 | 150 | $7,176 | Very very slightly included |
| VVS2 | 129 | $6,172 | Very very slightly included |
| VS1 | 114 | $5,454 | Very slightly included |
| VS2 | 100 | $4,784 | Very slightly included · index base |
| SI1 | 83 | $3,971 | Slightly included · often eye-clean |
| SI2 | 69 | $3,301 | Slightly included |
SI2 costs 59.6% less than IF. The two grades that matter most to a buyer's wallet are the ones nobody can see: the step from VS2 to SI1 costs 17% and is invisible to the eye in most stones.
We publish colour and clarity as two separate ladders, never crossed into a grid. Multiplying one by the other would reconstruct a proprietary price list cell by cell, and that list is somebody else's copyrighted product. The ladders are drawn from published index aggregates; the grid is not ours to print.
| Counter | Asking / ct | Tier |
|---|---|---|
| Ritani | $3,472 | value cut tier |
| Rare Carat | $4,420 | marketplace aggregate |
| Blue Nile | $5,454 | mainstream |
| PriceScope | $6,884 | published index |
| Whiteflash | $7,600 | super-ideal cut |
A 2.19× spread on an identical grading report, five counters now that James Allen is off the list (it was folded into Blue Nile, confirmed 2026-09-09). This is not error and it is not all margin — most of it is cut. The bottom of the range is a stone graded G VS1 and cut to a commercial standard; the top is the same paper cut to a super-ideal tolerance. The certificate does not price the diamond. The cut does.
| Step | Per carat | Reading |
|---|---|---|
| 0.90 → 1.00 ct | +23 to +24% | Like for like, same colour and clarity. Four independent methods converge inside a single point. |
| 0.30 → 0.50 ct | +39% | Per carat, RAPI against RAPI. |
| 1.00 → 2.00 ct | +190% | Per carat, on the IDEX drivers. Double the weight, nearly triple the rate. |
Divide the 1.00 ct row of the main table by the 0.90 ct row and you get +60%, not +24%. Both are right and they measure different things. The main table averages every stone inside a weight band, and the specification mix changes as the bands rise — more of the better colours sit above a carat. The +23 to +24% here is the same colour and the same clarity on both sides of the line. That figure is the one to plan a purchase around.
Buy 0.90 and nobody at the table can tell. That is the single cheapest piece of advice on this page.
| Measure | Figure |
|---|---|
| 1 ct round, natural, live retail average | $4,581 / ct |
| 1 ct round, lab-grown, live retail average | $550 / ct |
| Natural to lab multiple | 8.3× |
| Lab-grown wholesale since July 2018 | −96% |
| Lab-grown Q2 2026, overall | −13% year on year |
| Lab-grown Q2 2026, 1 ct rounds | +1% — the only riser |
The fancy-shape discount has inverted in lab-grown. Where a natural oval costs 23% less than a natural round, a lab oval costs 11% more than a lab round — and a lab princess costs 23% more. In a market with no scarcity, price follows growing time and cutting loss rather than rarity, and the ranking flips. It is the cleanest single illustration of what the two markets actually are.
De Beers realised $110 a carat in the second quarter of 2026, down 37% year on year; $105 across the half, down 32%. Its half-year price index reads 69, off 16%. The full-year 2025 average was $142.
RAPI at 0.30 ct reads $927.08 at its 2026-10-25 point against $845.11 on 1 January, up 9.7% on the year, and sits 12.4% under its $1,057.85 level of 1 September 2025. The small-stone recovery is the only weight band on this page that is higher than it started the year.
One carat reads $3,876.42 at its 2026-10-25 point against $4,125.20 on 1 January, down 6.0% on the year, and 12.9% under its $4,448.65 level of 1 September 2025. Three carat reads $18,264.96 against $18,225.61, up 0.2% on the year.
US duty effective 24 July 2026: India 10%, EU zero, Hong Kong and China 12.5% plus Section 301. Canada goes to 50% from 19 August.
The G7 threshold on Russian-origin stones has stood at 0.50 ct since 1 September 2024, and a due-diligence statement has been required on polished imports since 1 January 2026.
The Fancy Colour Research Foundation index was −0.2% on the quarter and −0.9% on the year in Q1 2026. Blue is best read as unchanged: two houses reported the same release as +0.3% and −0.3%.
Gem Diamonds averaged $1,395 a carat in the first half, up 38%, with a 52.24 ct white at $32,908 a carat. At auction the Ocean Dream, 5.5 ct vivid blue-green, made $17.3m — about $3.15m a carat.
The Gareth Penny consortium is reported at about $1bn for Anglo American's 85%, against a book value near $2.3bn earlier this year.
The trade per-carat column at 0.50, 0.70, 0.90, 1.00, 1.50, 2.00, 3.00 and 5.00 carats is read directly from IDEX price drivers on 2026-10-26. The 0.30 ct figure is RAPI at its own newest published point, 2026-10-25. The one-carat retail figure is the median of five retailers priced on 30 July 2026 (James Allen was dropped 2026-09-15, folded into Blue Nile). The shape table is a live sample of 169,324 asking prices. The colour and clarity indices are published aggregates at 31 December 2025.
Retail per carat at every weight other than one carat. It is the trade benchmark multiplied by a factor that falls from 2.10 at 0.30 ct to 1.35 at 5.00 ct, calibrated on the observed 1.84 at one carat and cross-checked against the disclosed gross margins of three listed retailers. Rows carrying modelled figures say so in the table.
The 4.00–4.99 ct band, because its published per-carat average exceeds the band above it on a sample of 0.32%. And the colour × clarity grid, because building one from published premiums would reconstruct a copyrighted price list.
All figures are US dollars per carat unless a whole-stone column says otherwise. Rapaport's index documentation describes its levels in hundreds of dollars per carat; at that reading a one-carat stone would price near $388,000. The levels are only coherent as plain dollars per carat and are printed here as such.
This is a market reference, not a valuation and not an offer. Every stone prices on its own certificate, its own cut and its own day.
| Rapaport — RAPI daily index levels | 2026-10-25 | rapaport.com |
| Rapaport — annual statistics, colour and clarity aggregates | 2025-12-31 | rapaport.com |
| Rapaport — price list guidance on trade discounts | 2026-07-29 | rapaport.com/diamond-price-list |
| IDEX Online — polished price drivers by weight band | 2026-10-26 | idexonline.com |
| IDEX Online — composite polished index | 2026-07-12 | idexonline.com |
| StoneAlgo — live shape and weight asking-price sample, n=169,324 | 2026-07-30 | stonealgo.com |
| Blue Nile, Ritani, Whiteflash, Rare Carat — live inventory (James Allen dropped 2026-09-15, folded into Blue Nile) | 2026-07-30 | retailer sites |
| PriceScope — published diamond price index | 2026-07 | pricescope.com |
| Ajediam — independent Antwerp grid, used for ratios only | 2025-04 | ajediam.com |
| Anglo American — De Beers production and realised price report | 2026-07 | angloamerican.com |
| Gem Diamonds — half-year trading statement | 2026-07 | gemdiamonds.com |
| Fancy Colour Research Foundation — FCRF index Q1 2026 | 2026-04 | fcresearch.org |
| Edahn Golan — lab-grown wholesale series | 2026-07 | edahngolan.com |
| Signet Jewelers, Brilliant Earth — reported gross margin | FY2026 / FY2025 | SEC filings |
| US Federal Register — reciprocal tariff schedule effective 24 July 2026 | 2026-07-24 | federalregister.gov |
