Correction, November 4, 2026: an earlier version of this edition named the wrong top lot and said the sale's lots mostly stayed near their estimates. The top lot was a Mauboussin necklace at HK$16 million, and several lots sold well above their estimates. The text below is corrected.

A correction first, because this newsletter's entire premise is that the paperwork matters more than the story. Two editions ago, this space previewed a specific lot heading into Sotheby's Hong Kong: a pair of unheated Burmese spinel earrings, 20.18 and 18.63 carats. Having now checked the confirmed results against multiple independent sources, that exact pair does not appear in the sale. The likeliest explanation is a mix-up with the Harry Winston spinel earrings this newsletter covered in July, a genuinely different lot from a different house's December 2024 sale. That was a research error on this newsletter's part, and it is corrected here rather than quietly dropped.

What actually happened on September 17 is, if anything, a better story. Sotheby's Hong Kong closed its autumn-opening High Jewelry sale at HK$267.5 million, up from HK$223.75 million at the same sale last September. A real spinel result did land, just not the one previewed, and it answers this year's re-rating question more honestly than a single headline lot could have.

Key Takeaways

→ Sotheby's Hong Kong beat last year's bar. The September 17 sale totaled HK$267.5 million ($34.1 million) across more than 180 lots, up from HK$223.75 million in September 2025. That is a genuine year-on-year gain at the first major sale of the autumn season, the one this newsletter flagged two weeks ago as the bar to watch (Rapaport; SCMP).

→ Rubies led, and some lots ran far above estimate. The top lot was a Mauboussin ruby and diamond necklace at HK$16 million, above its HK$11 million high estimate. A pair of Burmese ruby earrings and a 5.06 carat Burmese ruby ring each made HK$12.8 million, inside their estimates, and a 7.52 carat Paraiba tourmaline ring made HK$10.5 million, also inside its estimate. A pair of Cartier tutti frutti earrings made HK$7.9 million against an estimate of HK$650,000 to 850,000 (Rapaport).

→ The real spinel result: within estimate, not a re-rating. A cushion-cut, 21.36 carat Tanzanian spinel and diamond ring sold for HK$5.2 million ($660,859), within its HK$4.5 to 6 million estimate (Rapaport).

The index tracks 8 reference stones, weekly. Each row holds until a new dated transaction moves it in the same category. The spinel row updates this week on the Sotheby's result above.

Auction prices realized, not retail. Sources: Christie's, Sotheby's, Gemfields, Bonhams, GemGuide, Pala International.

Notable this week: the spinel reference changes to this ring. $30,940 per carat covers the whole ring, diamonds included, and it is close to the prior $35,000 reference, a different stone sold in a different year. It is not the re-rating multiple this newsletter went looking for. The lot sold within estimate, as did several other colored-stone lots in the sale, while others ran well above. One ring does not settle where spinel is priced.

Why this result matters more than a bigger one would have. A lot that sells at four times estimate, like the Harry Winston spinel earrings in December 2024, tells you demand from at least two bidders wildly exceeded the house's own pricing model. That is dramatic and easy to write about, but it is also a single data point that can reflect one determined buyer as much as a market shift. A lot that sells within estimate tells you the house's pricing was close to what the room would pay. That is less exciting, and it is still useful information about where a stone's level sits.

Reading this September against last September. The total rose, and the lots behind it were mixed. The spinel, the Paraiba ring and two Burmese ruby lots sold within estimate, while the top necklace and the Cartier tutti frutti earrings went well above. A total can rise for several reasons, and these results alone do not tell us which one drove it. They are not, on their own, evidence that any specific stone's per-carat level moved.

What this means for the spinel thesis specifically. The July case for spinel rested on two claims: that the material is genuinely underpriced given its rarity and treatment profile, and that the trade was starting to notice, evidenced by the December 2024 result. This September's result neither confirms nor kills that thesis. A single ring selling within estimate is consistent with a market that still prices spinel conservatively by default, with occasional outlier results, exactly the Harry Winston lot, when a signature house and a large matched pair create genuine competition. The pattern to watch is not any one sale. It is whether outlier results like December 2024 become more frequent relative to within-estimate results like this one.

The correction, and why it matters beyond this edition. Getting a lot description wrong is a real failure for a newsletter built on the discipline of only moving numbers on verified, dated transactions. The fix is not to quietly swap in whatever real result exists and hope no one notices the mismatch. It is to say plainly that the previewed lot could not be confirmed, explain the likely source of the error, and report the real data that did land. That is the same standard this newsletter has asked of every lab report and every producer's earnings release since June. It applies here too.

1. A rising sale total and a flat per-carat price can coexist. Hong Kong's autumn opener beat last year in total value while several colored-stone lots, the spinel included, cleared within estimate. Do not read total sale growth as proof that any specific stone's price level moved. They are different measurements.

2. Within-estimate results are data too, not a null result. A clean sale at the expected price is as real a data point as an above-estimate result. The spinel reference changed to a different stone, so the gap with the old $35,000 figure does not show that the market weakened.

3. When a call turns out wrong, or unconfirmed, say so before moving on. The previewed lot in this newsletter's own past edition could not be verified in the confirmed results. Correcting that openly is the same discipline that makes every other number in this index worth trusting.

→ November, Christie's Hong Kong and Geneva. The larger autumn test. Last November's Christie's Hong Kong sale hit HK$538.13 million, up 15 percent year on year. This September's gain sets an encouraging tone heading into it.

→ October's Gemfields Mixed-Quality ruby auction. Still the next like-for-like read on the ruby row against February's $279, and still outstanding.

→ Whether outlier spinel results recur. One within-estimate ring does not settle the July thesis. The next major spinel lot, wherever it surfaces, is the more useful data point than this one alone.

Next time an auction result crosses your feed, check one thing before the headline number: did it sell within estimate, or dramatically above it. Within estimate means the market and the house agreed on the price, which is the closest thing this asset class has to a confirmed level. Dramatically above means something more interesting happened, and it is worth asking what. Both are useful. Only one of them is a story.

— Carat Brief

Carat Brief is editorial. Nothing here is investment, legal, or tax advice. Gemstones are illiquid and require certified verification before purchase. Work with credentialed labs: GIA, AGL, SSEF, Gübelin, Lotus Gemology.

Sources: Rapaport; South China Morning Post; Sotheby's. Per-carat figures are our arithmetic, from reported sale prices and stated carat weights. They cover the whole jewel, not the stone alone.