MIT’s Electronic-Photonic Chip Achieves Over 1 Pbps Data Transfer, Targets 10 Pbps by 2030 to Resolve AI Energy Crisis

July 19, 2026

MarketWise USA

Overview

A novel optical coupler developed through MIT’s FUTUR-IC program enables AI chips to transfer data at over 1 petabit per second (Pbps), with ambitions for 10 Pbps by 2030. This electronic-photonic integration dramatically reduces power consumption in AI data centers while simultaneously boosting speed and bandwidth, positioning it as a critical solution to the burgeoning AI energy crisis. Major chipmakers, including Nvidia, AMD, Broadcom, and Marvell, are already investing billions in CPO and photonic technologies, signaling strong industry validation.

In Depth

An innovative optical coupler emerging from MIT’s FUTUR-IC program promises to dramatically enhance AI chip data transfer capabilities and fundamentally alter the energy efficiency landscape of AI data centers. This technology will enable AI chips to transfer over 1 petabit per second (Pbps) of data, with a target of 10 Pbps by 2030.

Technical Details

This breakthrough is achieved through ‘electronic-photonic integration’ technology, which integrates both electronic and optical circuits onto a single chip. Traditional electrical signal data transmission has faced significant challenges with power consumption and heat generation due to the enormous data volumes and computational demands of AI data centers. The new optical coupler overcomes these physical limitations by enabling on-chip, light-based data transmission. Specifically, optical signals can transmit data far faster and with significantly lower energy consumption than electrical signals, thus dramatically improving the overall power efficiency of data centers. This technology is expected to eliminate data transfer bottlenecks, accelerating AI model training and inference.

Background & Context

The evolution of AI has heightened its dependence on high-performance hardware, making data center power consumption a global concern. Conventional copper interconnects generate heat due to electrical resistance, leading to increased power consumption. This issue is exacerbated as AI workloads scale, with projections suggesting that continued growth of AI data centers at the current pace could place significant strain on global power grids. Consequently, leading chip manufacturers like Nvidia, AMD, Broadcom, and Marvell are making multi-billion dollar investments in photonic technologies, including Co-Packaged Optics (CPO) and silicon photonics. They recognize optical technology as key to sustainable AI computing scaling and are already integrating it into their next-generation products.

Strategic Significance & Outlook

MIT’s technology offers a promising solution to the AI data center power consumption problem, establishing a foundation for further AI development. If 10 Pbps data transfer speeds are realized in the future, it would enable the efficient processing of massive models and complex tasks currently unimaginable in AI systems. This could revolutionize all fields where AI is applied, such as autonomous driving, new drug discovery, and climate change modeling. As commercialization progresses, it will also lead to reduced data center operational costs, making it an attractive sector for investors. International collaboration and standardization will be key to accelerating the widespread adoption of this technology.

Source: https://marketwise.com/investing/mit-electronic-photonic-integration-ai-data-center-energy-crisis/

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.