Mythic's $125M Bet on Energy-Sipping AI Processors

In Silicon Valley's relentless race for AI dominance, an underdog is making waves with a radically different approach. Mythic, a Palo Alto chipmaker founded in 2011, just secured $125 million to challenge NVIDIA's GPU stronghold with analog processors that promise staggering energy savings.

The Efficiency Edge

The Series D round, spearheaded by deep-tech investor DCVC, will fuel development of Mythic's analog processing units (APUs) - chips that could deliver 100 times the efficiency of conventional graphics processors. What makes them special? They combine memory and processing on a single plane, slashing the energy typically lost shuttling data between components.

"Our current architecture squeezes out 120 trillion operations per watt," explains CEO Taner Ozcelik, an NVIDIA veteran who restructured the company last year. "We're not looking to replace GPUs - think of us as the next layer of acceleration."

Strategic Backing

The funding consortium reads like a who's who of tech and defense:

• Venture heavyweights NEA and Softbank KR
• Automotive giant Honda
• Defense titan Lockheed Martin

The broad interest underscores Mythic's cross-industry potential. Their technology could revolutionize everything from cloud computing to battlefield drones.

Beyond Chips: The Starlight Advantage

Mythic isn't just selling silicon. Their newly launched Starlight platform embeds APUs directly into image sensors, dramatically improving low-light vision for:

• Autonomous vehicles navigating dark roads
• Security cameras identifying threats at night
• Military drones operating in near-total darkness

The system extracts clearer signals from minimal light - crucial for applications where every photon counts.

Manufacturing Muscle

Unlike some competitors relying on overseas fabs, Mythic produces its chips using standard semiconductor processes within U.S. allied territories. This strategic choice ensures supply chain security for defense clients while maintaining cost competitiveness.

The fresh capital will expand production capacity and refine Mythic's software toolkit - critical steps as they prepare for commercial deployment in AI inference markets.

Benchmark Breakthroughs

Internal tests suggest Mythic's chips can handle massive AI models (up to 1 trillion parameters) without requiring expensive high-speed connections between processors. Early data shows they outperform top-tier GPUs in operations per watt when running large language models.

"Their approach mimics biological systems," notes Steve Jurvetson of Future Ventures. "By unifying computation and storage like neurons do, they've cracked fundamental efficiency barriers."

As data centers grapple with skyrocketing power demands from AI workloads, Mythic's technology couldn't be timelier. Their success may hinge on convincing developers that analog computing deserves a seat at the table alongside digital workhorses like NVIDIA's GPUs.

Key Points

🔋 Energy Revolution: Mythic's APUs promise 100x greater efficiency than conventional GPUs by combining processing and memory 🌌 Low-Light Leader: Starlight platform enhances imaging capabilities crucial for defense and autonomous systems 🤝 Strategic Partners: Backing from Honda and Lockheed Martin validates cross-industry potential 🇺🇸 Domestic Production: Chips manufactured using secure U.S.-allied supply chains