China is sharply increasing exports of synthetic diamonds needed for AI.

The West has seen a sharp rise in the popularity of Chinese lab-grown diamonds. These synthetic diamonds have proven useful for cooling microchips used in AI technologies. China produces over 70% of the world's synthetic diamonds.

As Bloomberg reports, synthetic gemstones, traditionally associated with jewelry, are now widely used as cooling materials for microchips thanks to advances in AI, enabling the creation of denser and more powerful semiconductors. This industry's momentum has skyrocketed after several Chinese manufacturers reported that their customers have confirmed the effectiveness of using synthetic diamonds as heat sinks. Shares of Chinese synthetic diamond companies have soared, with Zhecheng Huifeng Diamond Technology Co. and SF Diamond Co. alone posting gains of 51% and 40%, respectively, in a single week.

Beijing previously announced restrictions on the export of certain types of synthetic diamonds and strengthened its control over high-tech supply chains. Last November, the Ministry of Commerce and Customs announced that a special license would be required for the export of certain types of synthetic diamond micropowders, single crystals, wire saws, and grinding wheels. In addition to cooling chips used in AI technologies, synthetic diamonds are also used in the manufacture of certain types of munitions and radar components.

Comments

[Justme]

If China is focused on industrial results, then our response can often consist of issuing various types of regulations for all of this: OSTs, GOSTs, STPs...
You can also write a complex report (by the way, for a pretty penny..)

[Orange Bigg]

Quote: JustMe

If China is focused on industrial results, then our response can often consist of issuing various types of regulations for all of this: OSTs, GOSTs, STPs...
You can also write a complex report (by the way, for a pretty penny..)

Incidentally, another revolution in microelectronics is currently underway. The era of transistor shrinking is over: a revolutionary method for creating 3D chips of the future has been unveiled—with a fantastic yield of 98-100%. Silicon "skyscrapers" instead of "suburbs." A group of researchers from the University of Illinois at Urbana-Champaign has developed a revolutionary technology for vertically stacking high-performance silicon circuits directly on top of each other. This breakthrough could help the semiconductor industry continue to increase the computing power of processors without further physical shrinkage of transistors. This development offers a solution to one of the most serious problems facing chipmakers amid the slowing of Moore's Law. For decades, performance has been increasing by shrinking components, but now the industry has reached the fundamental physical limits of silicon, where further miniaturization becomes economically and technically impractical. Instead of shrinking components, American engineers have proposed building chips "upward." Layering multiple silicon circuits allows for a radical increase in transistor density, shorter signal transmission distances within the chip, and significantly improved energy efficiency.

Monolithic 3D integration, developed in Illinois, takes a different approach: each subsequent circuit layer is grown and formed directly on top of the previous one. This allows for incredibly dense vertical contacts and micron-level alignment accuracy.
Temperature has always been the main obstacle for monolithic 3D chips. Manufacturing standard high-performance silicon devices requires heating to temperatures close to 1000 degrees Celsius. However, once the first layer of circuitry and metal wiring is completed, any subsequent thermal processes must not exceed 400 degrees Celsius, otherwise the underlying structure will simply melt or degrade. To overcome this thermal barrier, researchers have developed a method for transferring ultrathin monocrystalline silicon nanomembranes onto existing chip layers. The chemical bonding process requires temperatures no higher than 200 degrees Celsius, well within the safe thermal limit.

Using a new technology, scientists successfully created a three-layer stack, each layer containing 625 silicon transistors. The yield of usable components was astonishing, ranging from 98% to 100%. Moreover, the transistors produced at low temperatures demonstrated qualities completely comparable to standard counterparts manufactured at 1000°C.

https://www.ixbt.com/news/2026/05/31/3d-98-100.html

Recently, researchers from Huawei, together with a team from Nanjing University, published a paper in the journal Nature Electronics reporting the creation of a RISC processor using 2D materials. This could potentially help achieve record transistor density without the use of ASML's advanced sanctioned lithographs. Using only local equipment, China will be able to produce chips on par with those from TSMC, Intel, and other industry leaders.

https://3dnews.ru/1142550/huawei-pokazala-chip-s-2dtranzistorami-ih-plotnost-sravnima-s-2nm-tehprotsessom-tsmc

[Peter1First]

As far as I remember, a 40nm processor already had 30 different layers on the die. And that's on a relatively flat chip. If you were to assemble everything from a flat surface and pile it up high, how would it be cooled? Where would the heat be dissipated? But the Chinese apparently have a slightly different approach: they simply make several low-layer dies, where it's easier to create a fine structure, and then "glue" them together into a relatively multilayer one, circumventing the problem of doubling the precision from layer to layer...

[Orange Bigg]

Russian scientists have created something incredible – a crystal that thinks like the human brain. This breakthrough could forever change the world of computing, marking the end of the era of traditional silicon processors. Traditional computer chips are being replaced by systems inspired by nature's most perfection – our brain. The results of text recognition experiments are particularly impressive. The crystal demonstrates nearly 100% accuracy in this task, performing it over fifty times in a row without any loss of quality. By comparison, modern artificial intelligence systems begin to tire and make mistakes much earlier.

Scientists from St. Petersburg have created something astonishing: a crystal made of organic molecules and copper clusters that can process information like neurons in the brain. At first glance, this may seem like science fiction—how could an ordinary crystal mimic the operation of a complex biological computer? However, the nature of this material is truly unique. When exposed to a laser beam, the crystal's electronic structure is transformed. It begins to behave like a living neural network, capable of not only storing information but also processing it, similar to how our brains do. The laser pulse acts as a nerve signal, and the crystal itself acts as an artificial neuron, responding to these signals and changing its properties.

The crystal created by Russian scientists demonstrates astonishing properties that could revolutionize our understanding of computer technology. It consumes ten times less energy than existing neuromorphic devices, paving the way for more compact and energy-efficient systems. But the real breakthrough is the crystal's ability to store information for up to two hundred days—a thousand times longer than current analogs.

https://dzen.ru/a/Z6iiGIfL5SXavKa6

The creation of smart crystals by Russian scientists is more than just a scientific achievement; it's a window into the future. We are witnessing the birth of a technology that could completely change our understanding of computing and artificial intelligence. Perhaps we are now witnessing the first steps toward creating true artificial intelligence—not just a fast computing machine, but a system capable of human-like thinking, but with far greater capabilities.

[knn54]

NTK "IMK" (Kharkiv), PJSC "Poltava Diamond Instrument" have apparently been "decommunized".

[Smoked]

Well, as the world's leading manufacturers say, the traditional approach—smaller size with increased power—has reached its limits, but our Chinese comrades have proposed going the larger route, the very same "denser and more powerful" mentioned in the article. Smart people working for the future.

[archiroll]

What can you expect from them? They even lower the base rate – they don’t understand anything about economics...

[topol717]

Quote: arhitroll

What can you expect from them? They even lower the base rate.

Inflation in China is 0%, producer price index is 2.4%. When inflation in Russia drops to at least 4%, the rate will be much lower.

[esl462]

This is from the series: until you learn to swim, the pool won’t fill up.

[topol717]

Quote from esl462

This is from the series: until you learn to swim, the pool won’t fill up.

No, this is from the series: there is no need to give low-quality examples for comparison.

[Idle_piston]

[Quote]China is sharply increasing exports of artificial diamonds needed for AI[/ Quote]
The key word is "China is building up", and we have another meeting.

[topol717]

Quote: Idle_piston

and we have another meeting.

We mostly have natural diamonds, which we sell to India because no one else wants to buy them.

[Orange Bigg]

Quote from: topol717

Quote: Idle_piston

and we have another meeting.

We mostly have natural diamonds, which we sell to India because no one else wants to buy them.

Or more promises. Baikal Electronics announced the development of new Russian solutions for artificial intelligence systems, unveiling two future accelerators and a server kit with next-generation processors on May 18, 2026.

The Baikal BE-AI-D1000 accelerator (specialized processor) is designed for use in servers and data centers, performing in the same class as the NVIDIA L40S. The Russian AI chip delivers 1000 teraflops of FP8 operations and 500 teraflops of FP16 operations. The accelerator's memory capacity ranges from 48 to 64 GB, using GDDR instead of the more efficient HBM. The accelerator itself will cost approximately 720000 rubles (US$10000). The company is also developing a comprehensive server solution combining a GPU and a CPU, the latter of which is the new Baikal S2 processor based on the Arm Neoverse N2 architecture, first introduced in 2020.

A key aspect of the developments by the Russian company Baikal Electronics is the promised compatibility with the CUDA architecture in its new Russian AI chips. This is a crucial feature, as NVIDIA's CUDA ecosystem has created a high barrier to entry for competitors. This compatibility will allow Russian developers experienced with NVIDIA chips to write software for the new accelerators without retraining. The second of the Russian special-purpose processors announced at CIPR-2026 is the Baikal-AI-E1000. Designed for peripheral devices with a power consumption of no more than 30 W and a clock frequency of up to 2 GHz, it already features a working GPGPU core on an FPGA.

Both promising Russian products are at different stages of development in 2026, and the release of the entire line is planned for 2029-2030.

https://tehnoomsk.ru/archives/25499

[SmolH2]

It's good that the country has its own diamond-growing industries. In some respects, it's ahead of the rest. But not in terms of volume.

[Svetlana]

Diamonds are no longer grown in laboratories. They are produced industrially. Industry produces them.
...
Sorry, it's jarring. Labs, with their scientists and lab technicians, only develop technological processes and then transfer them to industry.

[gridasov]

Clearly, there is a shortage of modern physicists as specialists in this field of knowledge, capable of combining their specialized knowledge with analytical work. This is clearly reflected in periodic scientific and technical developments. Therefore, it is not the materials that determine the key efficiency of a development, especially in the field of energy production, but the process-based organization of stages as algorithms. It's like how mathematicians, in the process of substantiating infinity, fail to see infinity as a closed sequence, and they fail to perceive chaos as a local space within a scalable space of a different level, with a precisely defined polarization and consistent algorithms for process change.

[Melior]

From the point of view of banal erudition, each individual, critically
motivating abstraction, cannot ignore the criteria of utopian
subjectivism, conceptually interpreting generally accepted defanizing
polarizers, so the consensus reached by dialectic
material classification of general motivations in paradogmatic
predicate connections, solves the problem of improving the forming
geotransplantation quasi-puzzlelistates of all kinetically correlated
aspektov.