Who really shrank? Buybacks measured in gold, 2015–2026

This is not investment advice and nothing here is a buy or sell idea.

Share buybacks are usually reported in dollars and share counts. Our Gold-Oz Screener (GAVF v2.2) asks something different: what happens if you measure a company in troy ounces of gold instead of in currency? We ran 36 large US and Nordic companies through the screener. Who actually got smaller, and did it help shareholders in gold terms?

1. Who shrank, who issued

Horizontal bar chart of the change in total shares outstanding for 36 US and Nordic companies, from early 2015 to October 2026. Newmont (+111.2%) and SSAB (+81.5%, from a 2016 rights issue) issued the most shares. Apple (−37.8%), JPMorgan Chase (−28.9%), McDonald's (−27.3%), Home Depot (−24.3%) and Caterpillar (−24.2%) reduced their share counts the most. Eighteen companies cut shares by more than 5%, and 16 stayed within ±5%. Six Nordic companies (Sandvik, Investor, Atlas Copco, Swedbank, Assa Abloy, Volvo) show exactly 0.0% because their share counts stayed constant on today's basis; a dagger footnote explains Atlas Copco's buybacks were redemptions. Source line: GAVF Gold-Oz Screener, BusinessCraft analysis; data as of 2026-10-19. Not a trading signal.

From early 2015 to October 2026, 18 of the 36 companies cut their share count by more than 5%. Apple cut its count the most, by 37.8%, followed by JPMorgan Chase (−28.9%), McDonald’s (−27.3%), Home Depot (−24.3%) and Caterpillar (−24.2%). Two companies went the other way: Newmont’s share count rose 111.2%, and SSAB’s rose 81.5%, which came from its 2016 rights issue. The other 16 stayed within ±5%. Six of them show exactly 0.0%: Sandvik, Investor, Atlas Copco, Swedbank, Assa Abloy and Volvo. That is correct: on today’s share basis their counts never changed. Atlas Copco returned cash through redemption programmes rather than by cancelling bought-back shares, and the other five had no net issuance or cancellation. Share counts are split-adjusted and include all share classes. Essity starts in June 2017, when it listed, and Novo Nordisk starts in January 2019.

2. Per share versus the whole company

Six small line charts, each indexed to 100 at the start of 2015. They compare per-share value in gold (share count held constant) with whole-company value in gold (actual shares). Final per-share vs whole-company values: Apple 313 vs 195, JPMorgan Chase 186 vs 132, Caterpillar 272 vs 207, McDonald's 78 vs 57, Newmont 141 vs 297, SSAB 78 vs 142 (rights issue 2016). The values are price-only, excluding dividends. Buybacks lift the per-share line above the company line, and issuance pulls it below.

Each panel shows two lines, both indexed to 100 at the start. The gold line is the per-share value in gold: market value with the share count held at its 2015 level. The grey line is the whole company in gold, using the actual share count. Buybacks push the gold line above the grey one, and issuance pushes it below.

Apple’s per-share value in gold ended at 313, while the whole company ended at 195. JPMorgan ended at 186 versus 132, and Caterpillar at 272 versus 207. McDonald’s is the reminder that buybacks don’t guarantee anything. Its whole company shrank to 57 in gold terms and its per-share value to 78, so the buybacks softened the fall without preventing it. Newmont is the mirror image. The company grew to 297 in gold, mostly by issuing shares, but per-share value only reached 141. SSAB shows the same pattern. After its 2016 rights issue, the company ended at 142 and per-share value at 78.

All returns here are price-only, excluding dividends. Total returns are therefore understated, most of all for companies that pay out a lot.

3. How big is each company in gold?

Dot plot on a logarithmic scale of the market value of 36 companies in million troy ounces of gold in October 2026. Values range from SSAB at 2.5 million ounces to Apple at about 1,195 million. The five smallest (SSAB, Boliden, Telia, Essity, H&M) are highlighted. A note says this is a size measure, not a valuation.

At a provisional gold price of USD 4,196/oz, Apple is worth about 1,195 million ounces, Alphabet 993 million and Microsoft 936 million. The smallest companies in our sample are SSAB (2.5 million oz), Boliden (3.6), Telia (4.3), Essity (4.6) and H&M (6.6).

This is a measure of size, not of value. A low ounce figure just means a smaller company. It doesn’t make the company cheap, and nothing in the next section suggests it predicts anything.

4. Did low ounce levels predict anything?

Two-panel bar chart of median forward per-share returns in gold after months when a stock was in its lowest or highest quartile of gold-ounce value. With look-ahead (left), low months appear to beat high months: +11.2% vs −13.9% at 12 months and +42.3% vs −26.4% at 36 months. Without look-ahead (right), the gap disappears: −7.4% vs −7.8% at 12 months and −12.4% vs −11.1% at 36 months. Returns are price-only, excluding dividends. Footer: not a trading signal.

As before, returns are price-only, excluding dividends. We checked whether months when a stock sat in its own lowest quartile of gold-ounce value were followed by better per-share gold returns than months in its highest quartile.

Run naively, with quartiles taken from the full 2015–2026 history, the answer looks like a clear yes. At 36 months the median was +42.3% after low months and −26.4% after high months. But that version uses future data: whether a month counts as “low” depends on prices that came later.

Once the quartiles are built only from data available at the time, the gap disappears. At 12 months the medians were −7.4% (low) and −7.8% (high). At 36 months they were −12.4% and −11.1%. The averages lean slightly towards the low bucket (−1.8% vs −5.9% at 12 months, and −5.1% vs −6.8% at 36 months), but with overlapping windows and only about 20 independent observations per bucket at 36 months, that is well within noise. We found no edge.

Limits and caveats

36 companies, one period (2015–2026), heavily shaped by gold’s rise from a USD 1,251 monthly average in January 2015.
All returns are price-only, excluding dividends.
October 2026 is month-to-date, and its gold price is a provisional futures close. Earlier months divide a month-end share price by the LBMA monthly average gold price.
Nordic share counts are mostly issued shares, so treasury shares (about 1–3%) are not deducted. Some yfinance price histories adjust for spin-offs (for example Atlas Copco’s Epiroc in 2018) without recording them, which can distort early ounce levels.
No statistical significance test was run in v2.2.

Source: GAVF Gold-Oz Screener, BusinessCraft analysis; data as of 2026-10-19. Not investment advice.

Disclaimer: This blog reflects my personal views and not those of any employer, client, or entity. The information shared is based on my research and is not financial or investment advice. Use this content at your own risk; I am not liable for any decisions or outcomes.