Engineers from Pennsylvania have developed a silicon-photon chip that utilizes light waves instead of electricity to perform complex computations, as reported by Nature.
The chips are characterized by a radical acceleration in data processing and reduced energy consumption. These features are particularly relevant for training artificial intelligence systems.
The interaction of light waves with silicon represents one potential path to creating computers that surpass modern chips, which are essentially based on principles established in the 1960s.
Instead of using silicon wafers of uniform height, engineers employed wafers with indentations approximately 150 nm deep in certain areas. These height variations allow for the control of light wave propagation across the microchip and enable mathematical calculations at maximum speed without the addition of other materials.
The new chips also offer enhanced data security: information is not stored in the computer’s memory, as computations are performed simultaneously.
According to the engineers, this chip model is ready for commercial production and can be adapted for graphics processors, which are in high demand due to the widespread interest in AI.
Earlier in February, OpenAI CEO Sam Altman began discussions with investors to raise $5-7 trillion for AI chip production.