Energy-Efficient General-Purpose Chip Maker Efficient Computer Raises $97M at $650M Valuation

AI chipmaking has mostly picked one side of a tradeoff: chips built for one narrow computation are efficient, but risk becoming obsolete the moment AI models evolve past them. Efficient Computer is betting the tradeoff isn’t necessary — that a chip can stay fully general-purpose and still be dramatically more efficient.

Efficient Computer said it has raised more than $97 million in Series B funding at a $650 million valuation, bringing total funding to $173 million. TQ Ventures led the round, with participation from Eclipse, Union Square Ventures, Giant Ventures, Triatomic Capital, TO Capital, TF Capital, Mana Ventures, Toyota Ventures, Overmatch, and Borderless.

The company’s Fabric architecture is a clean-slate redesign of computing. Where a conventional processor executes instructions one at a time in sequence, Fabric treats a program as a circuit, letting each piece of computation fire the moment its data is ready and run in parallel across the chip. That structure delivers a 10-to-100x improvement in energy efficiency on general-purpose workloads, the company says. Crucially, it doesn’t sacrifice generality to get there: Fabric supports standard software frameworks, including C and C++, unlike specialized “AI-only” chips that struggle to support broad software and risk going obsolete every time AI architectures shift.

It’s worth being precise about what’s actually shipping today, though: it’s an edge chip, not a data center chip. The company’s first production processor, Electron E1, is already in the hands of customers in physical AI, autonomy, critical infrastructure monitoring, space and defense, and wearables — small, battery-powered devices — and was a 2025 CES Innovation Award honoree. Data-center-class performance is a goal for the same Fabric architecture, not a shipping product with paying customers yet. This $97 million is earmarked specifically for that climb from edge to data-center scale.

CEO and co-founder Brandon Lucia is a Carnegie Mellon University professor who founded the company in 2022 alongside fellow CMU professor Nathan Beckmann, their former PhD advisee Graham Gobieski, and serial entrepreneur Alex Hawkinson, who founded SmartThings and sold it to Samsung. The technology commercializes NSF-funded research out of CMU. “Every customer we meet has a version of their product they cannot build, because the compute power budget makes the new capabilities they want infeasible,” Lucia said. “This round of financing makes it possible for many more new use cases and domains.”

“What convinced us was Brandon, Graham, and Nathan’s ability to build both the hardware and the software — and turn that breakthrough into a business,” said Andrew Marks, Co-Founding Partner at TQ Ventures. “Not only have they taped out four times, but they’re already shipping chips to customers at volume.” Efficient Computer raised roughly $16 million in seed funding late last year and a $60 million Series A led by Triatomic Capital this past February — meaning this Series B lands just seven months later.

The competitive picture splits along that same edge-versus-data-center line.

In the edge and physical-AI market Efficient Computer actually competes in today, the field is already crowded. Nasdaq-listed Ambiq, which makes ultra-low-power edge AI silicon, posted quarterly sales up nearly 90% year-over-year in its most recent quarter. Netherlands-based Innatera builds Pulsar, an ultra-low-power spiking neural network chip already deployed in smoke detectors and wearables, claiming up to 100x lower latency and 500x less energy than conventional AI chips. Analog in-memory computing startup Mythic is chasing the same market.

Move up to data-center scale, and the field gets much bigger. Tenstorrent, led by legendary chip architect Jim Keller, makes the same “efficient but general-purpose” pitch as Efficient Computer — at a very different scale. It licenses its RISC-V-based Ascalon CPU and Tensix AI cores as IP, recently raised $800 million led by Fidelity at a $3.2 billion valuation, and has reportedly drawn acquisition interest from Qualcomm at $8-10 billion. Ampere Computing made almost exactly the same “general-purpose plus efficiency” bet with ARM-based server CPUs — and SoftBank bought the whole company outright for $6.5 billion in March 2025, a sign of how large this market has already grown.

On the specialized-chip side Efficient Computer is trying to distinguish itself from, companies like Cerebras and Etched show the category is far from dead. Cerebras listed on Nasdaq in May, raising $5.5 billion, and its stock jumped 108% on its opening day — proof that specialized chips can still be a very large business. But another inference-chip heavyweight, Groq, tells a more complicated story: in December, Nvidia struck a $20 billion deal for Groq’s technology and talent, pulling in roughly 90% of its staff — including founder Jonathan Ross — to work inside Nvidia’s own data-center inference platform. Even the specialized-chip camp, it turns out, keeps getting pulled toward the gravity of the general-purpose data center.