Quantum security system promises to shield AI data centers

In this scenario, researchers have sought to develop alternatives for the future, with projects whose objective is to transmit data efficiently and safely in the long term, so that energy and water consumption is also reduced with the development of the sector.

And, according to a study published in the scientific journal Advanced Photonicsa promising path may have emerged. Chinese researchers demonstrated an optical transmission architecture, which combines very high-speed classical data traffic with quantum key distribution (QKD) — all on the same multicore fiber, but on separate cores, avoiding interference.

The demonstrated system achieved aggregate transmission of up to 2 terabits per second, while maintaining reduced consumption and complexity.

In simpler terms, the system handles classical (data) and quantum (QKD) signals in parallel within multicore fibers, whose composition includes several separate channels in a single filament. This system allows different types of signals to travel in parallel with minimal interference.

According to the authors, the discovery “paves the way for the next generation of secure, scalable and cost-effective optical interconnections”, especially at a time when scientists fear that quantum computers of the future could break current encryptions and decipher data stored today.

“Our project aims to protect AI-driven data centers against quantum security threats, while meeting the high demands of modern data-driven applications,” say the researchers in the report exposed at Advanced Photonics.

What were the results of the experiment for safety?

Tests to implement the system were carried out for 24 hours, under laboratory conditions. During the period, the approximately 3.5 kilometers of fiber operated at an aggregate rate of 2 terabits per second of classical data, with an average encrypted secret key (QKD) rate of 205 kilobits per second.

According to the scientists, the test generated 1,440 secure keys and could be capable of protecting around 21.6 petabits per second: to give you an idea, with such speed it would be possible to download the entire Netflix catalog twenty times in just one second.

The authors highlight that the architecture does not encrypt petabits of data per second, as it may seem — the number presented in the study corresponds to the total amount of data that could be protected by the generated keys, not the volume of traffic literally encrypted by QKD.

In practice, keys are used to power high-speed symmetric systems (such as AES), not to directly encrypt large volumes of data.

Still, the results demonstrate that the technology can provide stable quantum keys even in the presence of terabit/s traffic, something that until recently was considered unfeasible due to the intense noise caused by high-power lasers.

The Chinese research not only solves a bottleneck in the speed requirements of data centers, as well as increases the level of security in transmitting this data using quantum cryptography.

“By harmonizing cutting-edge photonics with quantum cryptography, we pave the way for secure, energy-efficient, ultra-high-capacity networks capable of sustaining the exponential growth of data-driven technologies,” the scientists conclude.

New challenges in the quantum area: what to expect in the future?

The discovery, which still requires more in-depth results, comes at a good time. Experts estimate that, by 2033, global data transmission is expected to increase by up to 25% per year. Thus, the current infrastructure and expansion projects of the data centers and data servers will not be ready for an ever-increasing demand to support the countless applications that make use of this technology.

In Brazil alone, the third largest user of ChatGPT in the world, the AI ​​chatbot receives around 140 million messages per day.

It is now up to scholars, together with civil society, to promote solutions to the environmental cost of Artificial Intelligence. In addition to consuming monumental amounts of energy, experts estimate that, over the next two years, AI-powered data centers could consume between 4.2 billion and 6.6 billion cubic meters of water in cooling the machines responsible for data processing.

Likewise, construction and excavations associated with large data center complexes have generated environmental concerns and documented episodes of sediment buildup in river basins in regions of the United States, Germany and the United Kingdom.