Intel Unveils Chip That Processes Encrypted Data at Accelerated Speed

Intel has introduced a specialized processor designed to dramatically accelerate one of the most challenging tasks in cybersecurity: computing directly on encrypted information. The prototype chip, called Heracles, is engineered to perform fully homomorphic encryption (FHE) calculations thousands of times faster than conventional server processors.

FHE allows computers to process data without ever decrypting it, meaning sensitive information can remain protected even while it is being analyzed. In theory, this capability could enable privacy-preserving AI, or medical or financial analytics performed without exposing confidential data.

But in the real world, the technology has faced a major obstacle. Standard processors can handle encrypted workloads, but the computations are extraordinarily slow. Mathematical operations required for FHE often take orders of magnitude longer than similar operations on unencrypted data.

Intel’s Heracles chip is intended to close that performance gap. Demonstrated at the IEEE International Solid-State Circuits Conference in San Francisco, the processor is purpose-built for the unusual mathematical operations that encrypted computing requires. Intel claims it can accelerate key FHE calculations by more than a 1,000 times, and in some cases more than 5,000 times, compared with high-end server CPUs.

Parallel Compute Engines

The architecture of the Heracles chip is far different than the typical CPU. Instead of running operating systems or standard software, Heracles functions as an accelerator dedicated to encrypted workloads. It incorporates dozens of parallel compute engines arranged in a grid, enabling large volumes of encrypted data to be processed simultaneously.

The processor also relies heavily on memory bandwidth. FHE dramatically increases the size of data once it is encrypted, sometimes expanding information many times beyond its original form. To manage that data flow, the chip integrates high-bandwidth memory stacks capable of feeding large datasets to its compute units at high speed.

Intel says this combination of parallel processing and specialized arithmetic units enables the chip to handle the complex operations that dominate FHE workloads. These include large-integer calculations and other mathematical routines that are inefficient on traditional processors.

Milliseconds vs. Microseconds

Intel researchers simulated a privacy-protected query to a database. In the scenario, a voter sends an encrypted request to verify that a ballot was properly recorded in an encrypted election database. Because the data remains encrypted throughout the process, neither the database nor the server ever exposes the underlying information.

On a conventional server processor, the simulated verification took milliseconds. The Heracles chip completed the same operation in microseconds.

That time difference becomes highly significant at scale. Processing tens of millions of queries could shrink from weeks of compute time to less than an hour.

Still in Prototype

The research and design work began roughly five years ago as part of a research effort supported by the U.S. Defense Advanced Research Projects Agency (DARPA). The program aimed to explore hardware designs capable of making fully homomorphic encryption practical for real-world computing systems.

Along with Intel, several startups are developing competing accelerators, and researchers are exploring alternative approaches, including photonic processors. Some companies are focusing on software platforms that enable encrypted queries without requiring specialized hardware.

The emergence of dedicated silicon could be a turning point. If encrypted computation becomes fast enough, it could change how organizations handle sensitive information in cloud environments. Use cases could range from privacy-preserving AI models to secure medical research using encrypted patient records.

Intel has not announced commercial plans for the chip, but researchers say the prototype demonstrates that large-scale encrypted computing may finally be within reach. The company is continuing to refine both hardware and software designs as it develops future versions of Heracles.