Neurophos Raises $110 Million In Series A Funding Round

Listen to this article

Neurophos, an Austin-based startup specializing in photonic AI chips, recently completed a $110 million Series A funding round, which appears to be oversubscribed and brings its total funding to $118 million. The round was led by Gates Frontier, with notable participation from strategic investors like Microsoft’s M12, Aramco Ventures, and others, suggesting strong interest in energy efficient AI hardware.

Neurophos is a semiconductor company focused on developing photonic chips for AI inference, spun out from Duke University and the Metacept incubator. Founded in 2020 by Dr. Patrick Bowen and Dr. Andrew Traverso, the team includes veterans from major tech firms such as NVIDIA, Apple, Samsung, Intel, AMD, Meta, ARM, Micron, Mellanox, and Lightmatter. The company emphasizes optical computing to overcome limitations in traditional silicon-based GPUs, particularly in power consumption and scalability for data centers.

This Series A round, valued at $110 million, follows a $7.2 million seed round in late 2023 and an earlier $800,000 pre seed. The investment will fund the rollout of Neurophos’ first integrated photonic system, including optical processing unit (OPU) modules for data centers, software tools, and developer kits. It also supports operational growth, such as expanding the Austin headquarters, hiring up to 80 new employees locally, and opening a San Francisco engineering office to engage early customers. No public valuation was disclosed for this round, though the investor lineup, including climate focused and tech giants, indicates confidence in its long term potential.

Neurophos’ approach uses metamaterial-based optical modulators to enable dense, efficient photonic computing, potentially offering 100x improvements in performance and energy use compared to current GPUs. This could help mitigate AI’s escalating energy demands, making it more accessible across industries, though real world deployment will depend on manufacturing scalability and integration challenges. In a competitive AI chip market dominated by players like Nvidia, this funding positions Neurophos as a contender in next-generation hardware, with early customer interest from unnamed partners.

Neurophos, an emerging player in the photonic computing space, has marked a significant milestone with its $110 million Series A funding round. This oversubscribed raise, led by Gates Frontier, the venture arm associated with Bill Gates, and featuring a diverse coalition of investors, underscores the growing urgency to address the power and scalability bottlenecks in artificial intelligence infrastructure. With total funding now at $118 million, the Austin-based startup is poised to accelerate its transition from research to commercial deployment, focusing on optical processing units (OPUs) that promise to redefine energy efficient AI inference.

Founded in 2020 by Dr. Patrick Bowen and Dr. Andrew Traverso, Neurophos originated as a spinout from Duke University and the Metacept incubator, which was established by physicist David R. Smith. The company’s roots trace back to metamaterials research initially aimed at applications like invisibility cloaks, but pivoted to solve practical challenges in AI computing. By leveraging micron scale metamaterial optical modulators, a breakthrough that achieves a 10,000x miniaturization over traditional photonic elements, Neurophos integrates over one million optical processing elements onto a single chip. This design enables massive parallelism in optical computing, where light based operations replace electron based ones, reducing heat generation and energy loss inherent in silicon GPUs. The result is a chip that claims up to 100x gains in both performance and energy efficiency, positioning it as a drop-in replacement for data center accelerators while adapting to evolving AI workloads.

The technology’s core innovation lies in its metasurface modulators, which function as tensor core processors for matrix vector multiplications critical to AI inferencing. Unlike conventional photonic chips that struggle with size, manufacturability, and the need for power hungry digital to analog converters, Neurophos’ approach minimizes these hurdles by encoding phase profiles at subwavelength scales. This allows for sculpted surface acoustic waves (SAWs) or light beams that can focus, route, or block data with unprecedented precision, all while maintaining compatibility with existing fabrication processes like lithography on lithium niobate substrates. Early validations through simulations, vibrometry, and particle experiments demonstrate capabilities such as unidirectional wave routing, particle patterning in fluids, and the creation of narrow jets with full width half maximum (FWHM) smaller than the wavelength, features that could extend beyond AI to biomedicine, quantum acoustics, and lab on a chip systems.

In the broader market context, Neurophos enters a crowded yet rapidly expanding AI hardware landscape, where demand for inference capabilities is surging amid the proliferation of large language models and generative AI. Traditional players like Nvidia dominate with silicon-based GPUs, but face criticism for their energy intensity, data centers alone are projected to consume up to 8% of global electricity by 2030. Photonic alternatives, including competitors like Lightmatter and Lumotive, offer advantages in speed (light travels faster than electrons) and reduced heat, but have historically lagged in scalability. Neurophos differentiates itself by emphasizing manufacturability and a timeline for mid 2028 chip availability, with pilots potentially starting in 2027. The company has already secured multiple unnamed customers, and Microsoft’s involvement signals potential synergies with hyperscale AI deployments.

The investor syndicate reflects strategic alignment with these goals. Gates Frontier’s leadership aligns with Bill Gates’ focus on climate and technology intersections, while M12 brings Microsoft’s AI expertise. Aramco Ventures adds an energy perspective, potentially eyeing reductions in compute related emissions, and Bosch Ventures contributes industrial hardware insights. Other participants, such as Carbon Direct Capital, Tectonic Ventures, Space Capital, DNX Ventures, Geometry, Alumni Ventures, Wonderstone Ventures, MetaVC Partners, Morgan Creek Capital, Silicon Catalyst Ventures, Mana Ventures, and Gaingels, provide a mix of deep tech venture support and diversity focused backing. This broad coalition not only validates the technology but also mitigates risks in a capital intensive sector.

Quotes from key stakeholders highlight the raise’s implications. Dr. Marc Tremblay, Microsoft’s Corporate Vice President and Technical Fellow of Core AI Infrastructure, noted, “Modern AI inference demands monumental amounts of power and compute… We need a breakthrough in compute on par with the leaps we’ve seen in AI models themselves.” CEO Patrick Bowen echoed this, stating, “Moore’s Law is slowing, but AI can’t afford to wait. Our breakthrough in photonics unlocks an entirely new dimension of scaling.” M12’s Managing Partner Michael Stewart praised the team’s progress toward a “realistic plan to deliver products,” while Carbon Direct Capital’s Jonathan Goldberg emphasized emissions reductions as “essential” alongside compute delivery. MetaVC Partners’ Chris Alliegro, an early backer, framed the investment as tackling “the limits imposed by silicon.”

Recommended: WitnessAI Raises $58 Million In Funding

Looking ahead, the funding will prioritize product milestones: datacenter ready OPU modules, a comprehensive software stack, and early access hardware. Expansion plans include bolstering the Austin HQ and establishing a San Francisco site to facilitate customer collaborations. While no specific valuation was disclosed, the round’s size for a Series A, coupled with prior seed funding of $7.2 million in December 2023 and $800,000 pre seed, suggests a post money valuation potentially in the hundreds of millions, though this remains speculative based on comparable photonic startups. Challenges ahead include proving mass manufacturability, competing with established incumbents, and navigating regulatory hurdles in AI hardware. However, if successful, Neurophos could contribute to a paradigm shift, enabling exaflop scale AI without proportional energy spikes, and making advanced computing more sustainable and widespread.

Please email us your feedback and news tips at hello(at)techcompanynews.com