A new wave of microelectronics startups is tackling AI's most pressing infrastructure challenge: how to deliver computing power without draining the grid. Rather than waiting for nuclear reactors or grid upgrades, these companies are pursuing alternative chip architectures, photonic systems, and quantum hardware to make AI data centers fundamentally more efficient. Two major funding announcements this week signal that investors and policymakers are betting on hardware innovation as the real solution to the energy crisis.

What Are These Startups Actually Building?

The Northeast Microelectronics Coalition (NEMC), a federally backed innovation network, awarded $1,013,287 in combined grants to eight startups through its PROPEL program . The cohort includes companies pursuing radically different approaches to the same problem: reducing the energy footprint of artificial intelligence workloads.

One standout is Ayo Electronics, a Boston-based company that received $100,000 to develop a customized photonic chip designed to improve AI inferencing speed and energy consumption by orders of magnitude compared to conventional graphics processing units (GPUs) . Rather than using traditional silicon transistors, photonic chips use light particles to process information, a fundamentally different approach that could slash power requirements for running AI models after they're trained.

Another recipient, VioNano Innovations, is developing polymer-enhanced patterning materials that improve precision in semiconductor manufacturing while extending the capabilities of existing fabrication tools . This approach enables more efficient chip designs without requiring costly new equipment, directly advancing energy-efficient computing across AI, data center, and communications industries.

The remaining six awardees are tackling complementary challenges: long-wave infrared detection on silicon platforms, protein sensing for biomanufacturing, AI-driven design automation for radio frequency components, chip-scale spectrometers, high-performance RF switches, and high-temperature digital circuits for extreme environments .

How Are Quantum Computers Entering the Data Center Race?

Meanwhile, a more ambitious bet is unfolding at Sygaldry, a quantum computing startup founded by Chad Rigetti, the entrepreneur behind the earlier quantum company Rigetti Computing. Sygaldry just revealed that it has raised $139 million in total funding, including a $105 million Series A led by Breakthrough Energy Ventures that closed in March . The company's earlier $34 million seed round was led by Initialized Capital and closed in August.

Sygaldry's pitch is straightforward but audacious: quantum hardware could be fundamentally more efficient at powering AI workloads than GPUs. The company is designing servers for AI data centers that combine both quantum hardware and classical chips, with the goal of delivering commercial systems that provide speed advantages for AI workloads by the end of the decade .

"Quantum is going to be a fundamentally more efficient way of translating power into intelligence," said Chad Rigetti, Sygaldry cofounder.

Chad Rigetti, Cofounder of Sygaldry

The investing thesis hinges on a critical observation: the energy intensity of large language models (LLMs) continues to grow at an unsustainable rate. Breakthrough Energy Ventures, the climate-focused fund backing Sygaldry, sees quantum as a potential path to breaking this paradigm . Rigetti and his cofounders, Idalia Friedson and Michael Keiser, are betting that quantum hardware will eventually remove computational limits that classical systems face.

Why Should You Care About These Funding Announcements?

These developments matter because they represent a shift in how the industry is approaching AI's energy problem. Rather than relying solely on grid infrastructure upgrades, nuclear power deals, or smaller AI models, companies are investing in fundamentally different computing architectures. The NEMC grants signal that the federal government and regional innovation networks view hardware innovation as critical to maintaining U.S. competitiveness in semiconductors and AI.

The PROPEL program has awarded 50 grants totaling $4.7 million since launching in 2024, building what NEMC Director Mark Halfman describes as a "pipeline of innovation acceleration across the microelectronics ecosystem" . These grants support both manufacturing needs, such as fabrication and packaging, and operational costs like electronic design automation software and patent strategy .

Mark Halfman

Steps to Understanding the Hardware Innovation Pipeline

• Photonic Chips: Use light instead of electrons to process information, potentially reducing energy consumption by orders of magnitude compared to traditional GPUs for specific AI tasks like inference.
• Quantum Hardware: Leverages quantum physics principles to solve certain computational problems exponentially faster than classical computers, with potential applications in AI workload acceleration by 2030.
• Manufacturing Efficiency: Improvements in chip design and fabrication processes enable more efficient transistor layouts and better power handling without requiring entirely new manufacturing equipment.
• Specialized Architectures: Custom chips designed for specific tasks, such as infrared sensing, RF components, or high-temperature operation, reduce wasted power by eliminating unnecessary general-purpose computing overhead.

Rigetti acknowledges that quantum computing has long been dismissed as perpetually 50 years away, a running joke in the industry. However, he believes the timeline is accelerating, particularly now as AI's energy demands create genuine commercial pressure . The question is whether quantum systems can move from laboratory demonstrations to production-ready hardware within the next four years.

The convergence of these two funding announcements suggests that the microelectronics industry is pursuing a diversified strategy: some companies are optimizing existing silicon-based approaches through better manufacturing and photonic integration, while others are placing longer-term bets on quantum systems. Neither approach is guaranteed to succeed, but together they represent the most concrete technical response yet to the data center power crisis that has become central to AI's future viability.