{"id":163995,"date":"2026-03-12T11:12:33","date_gmt":"2026-03-12T09:12:35","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/intel-predstavila-mechtu-anonimov-chip-heracles-dlya-raboty-s-zashifrovannymi-dannymi-bez-deshifrovki"},"modified":"2026-03-13T11:32:17","modified_gmt":"2026-03-13T09:32:17","slug":"intel-predstavila-mechtu-anonimov-chip-heracles-dlya-raboty-s-zashifrovannymi-dannymi-bez-deshifrovki","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/intel-predstavila-mechtu-anonimov-chip-heracles-dlya-raboty-s-zashifrovannymi-dannymi-bez-deshifrovki","title":{"rendered":"Intel has presented the dream of the anonymous \u2014 the Heracles chip for working with encrypted data without decryption.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Intel first revealed the chip<\/a> for Fully Homomorphic Encryption (FHE), which performs tasks 5000 times faster than traditional server processors. This is the Heracles accelerator, designed for the U.S. military to counteract attacks from quantum computers. Heracles speeds up the processing of fully encrypted information without any decoding, maintaining privacy at all stages.<\/p>\n

\"Intel<\/p>\n

Fully Homomorphic Encryption (FHE) is a promising cryptographic technology that allows arbitrary computations on encrypted data without needing to decrypt it. This addresses a fundamental trust issue with cloud services and AI systems: data remains completely confidential to platforms even during processing. However, the main obstacle to the practical application of FHE remains the extremely low performance of standard processors and graphics cards.<\/p>\n

You can read more about the 'puzzling' mathematics of Fully Homomorphic Encryption in the article \u2018Full Homomorphism \u2013 and No Trust!\u2019<\/a> from our archive. Without going into detail, it should be noted that such operations on standard CPUs and GPUs can be thousands to tens of thousands of times slower than working with open data. That\u2019s why leading companies and startups are actively developing specialized accelerators for FHE. After all, anonymity will always be in demand.<\/p>\n

The chip named 'Heracles' was started by Intel five years ago. Before the ISSCC conference last month, little was known about it. Now, Intel is ready to share information about it cautiously, as competitors are not resting and it\u2019s important to stay relevant. According to company representatives, Heracles accelerates FHE computations by up to 5000 times compared to the best Intel Xeon server processors (such as the 24-core Sapphire Rapids).<\/p>\n

The chip is manufactured using the most advanced 3nm FinFET process and has an area approximately 20 times larger than competing FHE solutions (over 200 mm\u00b2 compared to about 10 mm\u00b2 for competitors) and works closely with two 24 GB HBM modules. The chip and memory are housed under a common water block and are cooled by a liquid coolant. Overall, the design resembles a high-end GPU but addresses different tasks.<\/p>\n

The architecture of 'Heracles' features a cell structure that connects paired compute tiles via a bus with a bandwidth of 9.6 TB\/s. The external HBM memory has a total capacity of 48 GB buffered by a shared cache memory of 64 MB for all paired tiles. The access channel to external memory provides 819 GB\/s. Such specifications are far beyond what even the most powerful graphic accelerators can offer today.<\/p>\n

The company claims that Heracles is the first FHE chip operating 'at scale'. In other words, this is not a laboratory demonstration, but a solution capable of executing real large-scale tasks. In a live demonstration at ISSCC, the chip processed an anonymous request to an encrypted voter database (voter registration verification function) in 14 microseconds instead of 15 milliseconds on a Xeon \u2014 a difference that reduces the verification time from 17 days to 23 minutes when checking 100 million ballots.<\/p>\n

The public presentation of 'Heracles' can be seen as an important step in the race for the commercialization of hardware FHE platforms. Startups like Duality Technologies, Optalysys, Niobium Microsystems, and others are also actively promoting their solutions, but Intel is aiming to maintain its leading position in the industry which will support this promising development. If the technology becomes widespread, it could radically change the approach to privacy in cloud computing, data analysis, machine learning, and many other areas where confidentiality is often sacrificed for processing speed.<\/p>\n

Source:<\/strong><\/p>\n