Diamond Foundry says it grew a 200 mm diamond crystal for future chip cooling

Diamond Foundry says it has grown one continuous diamond crystal measuring 200 millimeters across, claiming a world first. Its November 5 announcement brings a potential heat-spreading material closer to semiconductor scale. The immediate achievement is crystal growth, not a finished cooling product: the sample is unpolished, and its suitability for commercial chip manufacturing remains unestablished.

Moving heat away from tiny hotspots

AI processors concentrate substantial power into small areas of silicon. Local heat buildup can restrict performance, increase cooling needs and accelerate component wear. Diamond Foundry’s proposed role for diamond is to spread that heat at the material level, opening possibilities for different chip layouts and power densities rather than simply adding cooling outside the processor.

The attraction is thermal conductivity: how readily a material carries heat. Interesting Engineering puts high-purity single-crystal diamond’s room-temperature conductivity at roughly 2,000 watts per meter-kelvin, several times that of copper. That is a reference value for the material, not a measured result for Diamond Foundry’s new sample or a demonstration of improved AI-chip performance.

The single-crystal distinction also matters. This is one continuous structure, not smaller diamonds joined together or a disk containing many crystal grains. Boundaries between grains and other defects can affect how heat and electrical charge move through a material. Diamond also combines its heat-carrying ability with electrical insulation, a useful combination for semiconductor cooling.

Size has limited those possibilities. Diamond could serve as a heat spreader or as a substrate—the supporting material within a chip package—but conventional lab-grown single crystals have historically been small. Diamond Foundry presents the 200 mm result as a way past that constraint, toward material that can be manufactured at the scale used by semiconductor plants.

Keeping one crystal ordered as it grows

Growing a larger diamond requires preserving an ordered arrangement of carbon atoms across the material. Small inconsistencies can introduce defects that undermine the desired properties. One route is chemical vapor deposition, which grows diamond from a carbon-containing gas in a controlled environment. Microwave plasma versions of that process are important in research on large, high-quality diamond crystals.

Diamond Foundry previously announced a 100 mm single-crystal diamond wafer made through heteroepitaxy: growing a crystal on a different underlying material while encouraging an ordered diamond structure. The new crystal doubles that earlier diameter. The company has not disclosed whether the 200 mm sample uses precisely the same growth process, according to Interesting Engineering.

The company credits scientists Stefan Gsell and Martin Fischer with pursuing scalable single-crystal diamond for 25 years. They joined Diamond Foundry in 2022 through its acquisition of Augsburg Diamond Technology, or Audiatec. Their crystal-growth expertise was combined with Diamond Foundry’s reactor technology and manufacturing capabilities, linking the current milestone to work that began before the acquisition.

Diamond Foundry says reaching this size required advances in plasma physics, reactor engineering and process control. Its account also describes work on growth chemistry, crystal engineering, measurement and automation. CEO Martin Roscheisen credits hundreds of company engineers, scientists and technicians, describing repeated experiments, failures, inventions and rebuilt machines over more than a decade.

A crystal still needs to become a usable wafer

The physical sample is not yet a production-ready semiconductor wafer. Tom’s Hardware notes that it needs further processing before chipmakers can use it. Beyond polishing, manufacturers need consistent thickness, a sufficiently flat surface, low defect density and dependable thermal and electrical properties across the usable area. Those requirements determine whether a large crystal can become a useful manufacturing input.

Repeatability matters too: manufacturers must show that they can produce suitable material across multiple runs and integrate it into device fabrication. Diamond Foundry has not disclosed manufacturing yield or production cost. Those missing figures leave the economics unresolved even if the crystal-size milestone holds.

Nor does using diamond to carry heat mean replacing silicon with diamond electronics. Manufacturing electronic devices from diamond remains difficult, including reliably controlling its electrical properties through doping—the addition of impurities. Integrating diamond with existing semiconductor materials and processes also requires further development. A thermal application and a diamond-based electronic device therefore face different technical demands.

Roscheisen’s confidence about a larger crystal remains a company forecast. Tom’s Hardware reports that a company graph projects a 300 mm single-crystal diamond could be possible before 2030, matching the wafer diameter used in leading semiconductor processes. That is a further growth target, not evidence that either size is ready for commercial chip production.