Mojo Vision's Micro-LEDs Aim to Break AI's Data Bottleneck

CUPERTINO, CA – March 25, 2026 – Mojo Vision, a company specializing in high-performance micro-LEDs, has secured a $17.5 million strategic investment from Future Ventures, the deep-tech venture firm led by Steve Jurvetson. The funding is earmarked to accelerate the commercialization of a novel technology aimed at the heart of a growing crisis in the tech world: the data bottleneck throttling the growth of artificial intelligence.

As AI models become exponentially more powerful, the data centers that house them are hitting fundamental physical limits. This new investment, which follows a $75 million financing round last year, positions Mojo Vision’s micro-LED platform as a potential solution to the crippling constraints of bandwidth and power consumption that threaten to stall AI’s progress. The capital infusion highlights a growing consensus that incremental improvements are no longer sufficient.

“AI infrastructure is reaching fundamental limits in bandwidth density and power efficiency, and incremental improvements are no longer enough,” said Nikhil Balram, CEO of Mojo Vision, in a statement. “Our micro-LED platform was purpose-built to overcome this tradeoff.”

The AI Infrastructure Crisis

The explosive growth of AI has created an insatiable demand for computational power, but this demand has exposed a critical weakness in modern data center architecture: the interconnects. These are the communication pathways that allow processors and memory units to talk to each other. As AI workloads involve shuffling petabytes of data between thousands of chips, these pathways have become congested highways during rush hour.

Industry analyses confirm the severity of this bottleneck. A traditional data center rack might consume 5-10 kilowatts of power. In contrast, a rack dedicated to AI workloads can demand over 30 kW, with some high-density systems pushing 100 kW. This immense power draw not only leads to staggering operational costs and environmental concerns but also generates enormous amounts of heat, forcing a costly shift from air cooling to complex liquid cooling solutions. Power availability itself is now a primary constraint on where new AI data centers can even be built.

This challenge is twofold. First is bandwidth density—the amount of data that can be moved through a given physical space. Second is power efficiency, measured in energy per bit of data transferred. Existing copper and even some optical technologies struggle to scale, consuming too much power and taking up too much space. The result is a system where incredibly fast processors often wait idly for data to arrive, an inefficiency that AI operators cannot afford.

A New Class of Optical Interconnect

Mojo Vision is proposing a radical departure from conventional solutions. Instead of relying on traditional lasers, the company has developed a platform that uses dense arrays of microscopic LEDs, each acting as a tiny, highly efficient data transmitter. This approach enables what the company calls massively parallel optical I/O.

The core of the technology is an integrated system that combines tiny Gallium Nitride (GaN) micro-LED emitters and silicon photodetectors on advanced 300mm CMOS wafers. These components are arranged to communicate through highly multicore fiber bundles, creating thousands of parallel optical lanes in a footprint roughly the size of a grain of sand. This parallelism is the key to achieving a massive leap in bandwidth density, enabling the transfer of terabits of data per second across very short distances with exceptionally low latency.

“By replacing lasers with dense arrays of micro-LEDs, the company’s approach has the potential to deliver dramatic gains in bandwidth density while reducing energy per bit,” noted Steve Jurvetson, Founder and Managing Director of Future Ventures. “With thousands of optical channels operating in parallel, this architecture offers a fundamentally more scalable path than traditional approaches. This is the kind of step-change AI infrastructure needs.”

Originally founded in 2015 to develop augmented reality contact lenses, Mojo Vision spent years and over $200 million perfecting its proprietary micro-LED display technology. In a strategic pivot, the company is now leveraging that deep expertise to solve this even larger problem in AI infrastructure, where its unique capabilities in creating ultra-small, power-efficient, and high-density light emitters have found a critical application.

Strategic Backing and Industry Validation

The investment from Future Ventures is more than just capital; it's a significant vote of confidence from one of Silicon Valley's most prominent deep-tech investors. Jurvetson's track record includes early, transformative investments in companies like SpaceX and Tesla, indicating a focus on foundational technologies with the potential to reinvent entire industries. This backing suggests that Mojo Vision’s approach is seen not as an incremental improvement but as a fundamental shift.

Further validating this direction is Mojo Vision’s recently announced strategic collaboration with Marvell Technology, a giant in data infrastructure semiconductors. Marvell not only became the largest investor in Mojo Vision’s recent financing but is also actively co-developing solutions with the company. This partnership aims to integrate Mojo’s micro-LED platform with Marvell's high-speed connectivity portfolio to create a new class of short-reach interconnects specifically for AI clusters.

This collaboration directly targets the communication between accelerators, memory pools, and other components within a data center, where speed and efficiency are paramount. The partnership provides Mojo Vision with a clear path to market and access to Marvell's extensive customer base in the hyperscale and cloud data center sectors. It also pits the technology against competing solutions, including the silicon photonics and co-packaged optics being heavily developed by giants like NVIDIA, setting the stage for a new era of innovation in data center hardware.

Beyond the Terrestrial Data Center

While the immediate focus is on solving the crisis within terrestrial AI data centers, the long-term implications of Mojo Vision’s technology extend far beyond. The press release hints at applications in “distributed and orbital computing infrastructure,” where the technology’s core attributes—power efficiency, light weight, and scalability—are not just beneficial but essential.

As AI workloads are increasingly distributed to the network edge, the need for low-power, high-bandwidth connections becomes critical. For applications in autonomous vehicles, remote sensing, and smart cities, processing must happen close to the data source, in environments where power and space are strictly limited.

Even more compelling is the potential in orbital computing. A new generation of companies and research projects, from Kepler Communications to Google's Project Suncatcher, are exploring the concept of placing data centers in space. Powered by uninterrupted sunlight and cooled by the vacuum of space, these orbital platforms could process vast amounts of data, particularly from Earth observation satellites, without straining terrestrial power grids. In this extreme environment, a technology that delivers terabits of data in a compact, low-mass, and power-sipping package could be the enabling technology that makes large-scale orbital AI a reality. Mojo Vision's platform, born from a vision of placing a display in a human eye, may ultimately find its most transformative role in forming the high-speed nervous system for computing on Earth and in the stars.