Industries / Food, farming and forestry / Food and beverage processing / Confectionery
Cocoa and chocolate making
Cocoa and chocolate making turns cleaned cocoa beans into chocolate: roasting and grinding them into a paste called cocoa mass, pressing out part of the fat (cocoa butter) to leave cocoa powder, adding sugar and milk, and grinding and stirring the mix until it is smooth.
7
Milestones achieved, from 1828 to 2025.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1828
Fat pressed out of cocoa
Before 1828 ground cocoa was sold in sweetened cakes so rich in fat that the butter floated on top of the drink, wrote Arthur W. Knapp in his 1920 history of the trade. In 1828 Van Houten of Amsterdam invented a way of pressing out part of the butter. The pressed cake made a lighter, cheaper cocoa powder, and the butter became the extra fat that eating chocolate needs. K. Kok, writing for the Zaan industrial heritage society, says the method was protected by a patent, so for years other Dutch makers kept selling unpressed cakes, and that Van Houten later also treated cocoa with alkali to soften its taste, the step now called Dutching. Cadbury first sold a pressed "cocoa essence" in 1866 and Fry a pressed cocoa in 1868, by Knapp's account.
Credit
the Van Houten family of Amsterdam (Kok names Coenraad Johannes van Houten; other accounts name his father Casparus)
1875
Milk added to chocolate
Daniel Peter, a candlemaker in Vevey, Switzerland, who had married into the Cailler chocolate family and opened his own chocolate business in 1867, tried adding his neighbor Henri Nestlé's powdered milk to chocolate. It went moldy, so he switched to Nestlé's condensed milk, and in 1875 he made milk chocolate that kept, by the account of Swissinfo, the Swiss public broadcaster's international service. Knapp gives 1876. Cailler later built a works near Gruyères, where milk was plentiful, and Cadbury built its own milk condensing factories in England, Knapp wrote. Henri Nestlé never made chocolate himself; his company merged with the firms of Peter, Cailler and Kohler in 1929.
Credit
Daniel Peter; Henri Nestlé
1879
The conche smooths chocolate
Rodolphe Lindt of Bern built a machine in 1879 that slowly stirs and airs warm liquid chocolate for a long time, by Lindt & Sprüngli's account; the company says the result, which it calls "chocolat fondant" (melting chocolate), was less acid and bitter. In the conche, named for its shell shape, a heavy roller is pushed back and forth through the chocolate. Knapp wrote in 1920 that conching ran from a few hours to a week, and that nobody agreed on how it worked. In 2019 the physicist Elena Blanco and colleagues showed in a study of a simplified recipe that the mixing work, together with adding emulsifiers in stages, lets the mass flow with more solid particles in it.
Credit
Rodolphe Lindt
2017
A fourth color of chocolate
Barry Callebaut unveiled ruby chocolate in Shanghai on 5 September 2017: pink chocolate made from what it calls ruby cocoa beans, to which, it said, "no berries or berry flavour, nor colour, is added", after many years of development, as Bakers Journal reported. The US Food and Drug Administration let the company market-test it from November 2019 as a product outside the US standards for chocolate. Its permit, extended in 2021 for about 60 million pounds (27,216 tonnes) a year, describes ruby chocolate as made with citric acid and without added coloring.
Credit
Barry Callebaut
2020
Hollow sugar dropped
In March 2018 Nestlé launched Milkybar Wowsomes in the UK and Ireland, the first product made with its sugar of a more porous structure, made by hollowing out sugar crystals; the bar had 30% less sugar than other chocolate in Nestlé's range. In February 2020 Nestlé told ConfectioneryNews that it had stopped making the bar because consumer demand was "underwhelming", and that it was working on new sugar reduction technology for 2021.
Lessona way of cutting sugar that works in the factory still has to sell; Nestlé gave weak demand as the reason, and a FoodNavigator writer asked whether marketing was to blame.
Credit
Nestlé
2024
Cocoa prices quadruple
In April 2024 London cocoa futures averaged US$11,042 a tonne, 301% above the US$2,752 of a year earlier, and on 19 April reached an all-time high of US$12,567 a tonne, by the International Cocoa Organization's (ICCO) monthly review. The ICCO put the cause in Côte d'Ivoire and Ghana, which grow about 54% of the world's cocoa: over three seasons their output fell by 467,000 and 448,000 tonnes, because of cocoa swollen shoot virus disease, illegal mining on cocoa land, aging trees and bad weather. It judged Ghana's decline structural, part of a trend that began in 2016/17.
Lessonwhen two countries grow more than half of the world's cocoa, disease and aging farms there reach every chocolate maker's costs.
Credit
figures from the International Cocoa Organization
2025
Chocolate made without cocoa
ChoViva, made by the German start-up Planet A Foods, is a chocolate substitute made by fermenting and roasting sunflower seeds (and, by 2026, grape seeds) and adding plant fats and sugar. By May 2025 Green Queen, a food industry news site, listed Lindt, Katjes and Lambertz among the brands using it, along with the store brands of Rewe, Aldi, Lidl and Kaufland. In 2026 Nestlé announced a ChoViva range under its Choco Crossies brand in Germany, and Barry Callebaut became its exclusive distributor to US food makers. In September 2026 Stiftung Warentest, the German consumer testing foundation, compared four ChoViva products bought in shops in June 2026 with the cocoa products they copy, from the same makers, and summed up its testers' verdict as "surprised, but not convinced" (our translation). Planet A Foods says ChoViva is in more than 130 products in 11 countries; that figure is its own.
Credit
Planet A Foods (Maximilian Marquart, chief executive)
Watch for
Cocoa grown from cells
California Cultured, a start-up in West Sacramento, grows cocoa from plant cells in tanks instead of on trees. It has told the US Food and Drug Administration (FDA) that it considers its cocoa powder generally recognized as safe (GRAS), the status a new food ingredient needs in the US, and plans to launch it later in 2026; it says the powder holds about 20 times the flavanols of ordinary cocoa. The ingredients group Puratos says its US arm will sell a chocolate containing this cocoa to professional customers by the end of 2026.
Achieved whenthe FDA answers the GRAS notice with a "no questions" letter, and a product made with cell-grown cocoa is on routine sale with its flavanol content measured by an outside laboratory.
Claimed by
Puratos, February 2026; California Cultured, March 2026
Less energy to roast beans
Bühler, a maker of food processing machines, says its new Lucent cocoa roaster uses at least 20% less energy per batch than its earlier Tornado roaster and gives up to 20% more throughput.
Achieved whena cocoa processor running a Lucent roaster publishes its measured energy use per tonne of beans beside that of the roaster it replaced.
Claimed by
Bühler, May 2026, at the interpack trade fair in Düsseldorf
Deforestation checks on cocoa
The EU Deforestation Regulation requires companies that sell cocoa and chocolate, and six other commodities, in the EU to check that they were not grown on deforested land. Its start was postponed in December 2024 and again on 19 December 2025. The European Commission now says it applies to large and medium-sized companies from 30 December 2026 and to micro and small ones from 30 June 2027, and it published a simplification review on 4 May 2026.
Achieved whenthe rule applies on 30 December 2026 without a further delay, and an EU enforcement authority or an outside researcher reports that cocoa entering the EU is being checked against the land where it grew.
Claimed by
European Commission, May 2026
Open
01
Cocoa-free bars measured openly
A public side-by-side measurement of chocolate made without cocoa against ordinary chocolate of the same kind (dark, milk, coating) from more than one cocoa-free maker: how each melts, how hard it is and how it snaps, how fine its particles are, and whether trained tasters can tell them apart in blind tests, with all raw data published.
The briefCredit · unclaimed02
Conching end point from sensors
Logs of many conching runs in small conches or stone grinders (the melangeurs used by bean-to-bar makers), with motor power, temperature, readings from cheap gas sensors over the chocolate, and hand-taken viscosity and fineness samples, plus a model that predicts when the chocolate has reached its target, so the run can stop then instead of after a fixed time.
The briefCredit · unclaimed
Open 01 · the brief
Cocoa-free bars measured openly
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public side-by-side measurement of chocolate made without cocoa against ordinary chocolate of the same kind (dark, milk, coating) from more than one cocoa-free maker: how each melts, how hard it is and how it snaps, how fine its particles are, and whether trained tasters can tell them apart in blind tests, with all raw data published.
Why it is still open
The performance figures for cocoa-free chocolate come from its makers; Planet A Foods says ChoViva replaces chocolate one for one and lowers emissions by 82 to 91%. Stiftung Warentest's 2026 test of four ChoViva products used tasting panels of three and ten trained tasters and contaminant tests on one bar, and reports no melting or particle-size measurements. A study for Fairtrade Germany by the rheingold institute interviewed 29 consumers. A Crossref search on 18 November 2026 found no published instrumental comparison of commercial cocoa-free chocolate with cocoa chocolate.
Built with what exists
a differential scanning calorimeter (DSC), which records the heat a sample takes up as it melts; a texture analyzer for snap; a micrometer or a laser diffraction instrument for particle size; triangle tests under ISO 4120, in which each taster picks the odd one out of three samples, with adult volunteers screened for allergies; products bought in shops only.
First step
Buy three cocoa-free products and their closest cocoa counterparts, and book an hour on a university food laboratory's DSC.
Done looks like
at least six matched pairs from at least two cocoa-free makers, measured three times each, triangle tests with at least 24 tasters per pair, and the raw data and method published under an open license.
Open 02 · the brief
Conching end point from sensors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Logs of many conching runs in small conches or stone grinders (the melangeurs used by bean-to-bar makers), with motor power, temperature, readings from cheap gas sensors over the chocolate, and hand-taken viscosity and fineness samples, plus a model that predicts when the chocolate has reached its target, so the run can stop then instead of after a fixed time.
Why it is still open
Run times are set by recipe and habit: Knapp gave "a few hours to a week" in 1920. In 2022 Maria Valverde-Ayllon and colleagues sampled a dark chocolate every 2 hours through a 20-hour refining and conching run at a Peruvian company and found no measurable or tasted difference after 12 hours, which cut 8 hours from each run; they sampled by hand, for one recipe. Sensors exist for large lines: Rheonics sells inline viscometers for a conche's recirculation loop, and T. A. Kuchmenko and R. P. Lisitskaya (2024) used chemical sensors to follow aroma development in a chocolate glaze during production. No open set of run logs for small conches or stone grinders, with a model of when to stop, was found on GitHub or in Crossref.
Built with what exists
a tabletop stone grinder; a plug-in power meter; a thermocouple; metal-oxide gas sensors on a microcontroller; a rotational viscometer and a micrometer for reference samples; scikit-learn or a similar library for the model.
First step
Fit a power meter and a thermocouple to one stone grinder and log five runs of the same recipe, taking a viscosity and fineness sample every hour.
Done looks like
at least 50 runs logged across three or more recipes, and on 10 held-out runs the model predicts the time at which viscosity and fineness stop changing to within 1 hour, with the logs published.