From "Invisibility Cloaks" to AI: Neurophos Builds Ultra-Efficient Optical Chips

Headline: From "Invisibility Cloaks" to AI: Neurophos Builds Ultra-Efficient Optical Chips

As AI advances rapidly and data center power consumption becomes one of the industry's biggest challenges, startup Neurophos has announced a hefty $110 million funding round. The money will go toward developing and commercializing optical processors that the company says will dramatically improve the efficiency of neural network inference. Neurophos builds its technology on advanced composite metamaterials and promises a breakthrough: computing with light instead of conventional electricity. The approach could open the door to more powerful and energy-efficient AI systems.

The metamaterials behind Neurophos's work have an interesting history. Researchers originally studied them as a way to build so-called "invisibility cloaks" — devices that steer light around an object so that an observer can't see it. Those early experiments showed how well metamaterials can manipulate electromagnetic waves. Neurophos has adapted these fundamental findings to an entirely different but equally pressing problem: speeding up AI computation. Instead of hiding objects, the light will now perform the complex math at the heart of neural networks — specifically during inference, when models are already trained and put to work on concrete tasks.

At the core of Neurophos's technology are purpose-designed composite metamaterials. These materials have unique optical properties that let them interact with light at the nanoscale. Traditional silicon chips process information by moving electrons; Neurophos's optical processors use photons, the particles of light, instead.

Light travels at enormous speed with almost no energy loss, which makes it an ideal carrier of information for high-performance computing. Neurophos designed its chip architecture to get the most out of these advantages. The architecture can run complex matrix multiplications and other operations critical to neural network inference with unprecedented speed and energy efficiency.

In essence, the computation happens as light passes through a specially structured metamaterial, which eliminates the need for large numbers of electronic switches and minimizes latency.

The impact of optical chips like these could be genuinely transformative. First, consider data center power consumption, which today accounts for a significant share of global electricity use. Neurophos promises its optical processors will cut that consumption by an order of magnitude, delivering substantial resource savings and a smaller carbon footprint.

Second, Neurophos's technology lets chips sidestep fundamental physical limits that conventional silicon runs into, such as Moore's law. Shrinking transistors and packing them more densely keeps getting harder and more expensive. Optical chips offer an alternative path to exponential performance gains that doesn't depend on miniaturizing silicon components.

That could enable more powerful and affordable AI systems capable of processing huge volumes of data in real time — critical for fields such as autonomous driving, medical diagnostics, and scientific research.

Neurophos's $110 million raise is a strong signal that investors believe optical computing can solve the AI industry's pressing problems. The jump from fundamental metamaterials research — once tied to the far-fetched idea of "invisibility cloaks" — to ultra-efficient optical AI chips is an impressive leap. The funding will let Neurophos accelerate commercialization of its unique architecture, which promises not only to speed up the most demanding neural networks but also to make them significantly more energy-efficient. If the technology lives up to expectations, it could become one of the key drivers of the next wave of AI, making it more accessible, powerful, and sustainable.