Eridu Emerges from Stealth with $200 Million to Tackle AI Networking Bottlenecks
Saratoga, California – Eridu, a startup focused on resolving the growing challenges of AI networking, has exited stealth mode with $200 million in funding. The company, founded in 2023, is tackling a critical infrastructure problem: as artificial intelligence models become increasingly complex and data-intensive, existing network infrastructure struggles to keep pace. This funding round, led by Socratic Partners, John Doerr, Hudson River Trading, Capricorn Investment Group, and Matter Venture Partners, signals a significant vote of confidence in Eridu’s approach to building a new generation of networking solutions specifically designed for the demands of modern AI.
The surge in popularity of large language models (LLMs) and other advanced AI applications is placing unprecedented strain on data centers and network connectivity. Techniques like mixture-of-experts models, which dynamically activate different parts of a neural network based on the input data, and the separation of AI inference into prefill and decode stages, are exacerbating the problem by requiring significantly more data movement. According to Promode Nedungadi, Chief Technology Officer at Eridu, the architectural and algorithmic trends driving AI are making the network problem “harder, not easier.” The company aims to address this challenge by fundamentally rethinking the architecture of AI networking, moving beyond incremental improvements to existing technologies.
Eridu’s strategy centers on a vertically integrated approach, encompassing silicon development, advanced packaging, and software optimization. This allows them to control the entire stack, enabling a level of performance and efficiency that they believe is unattainable using off-the-shelf components. The company is partnering with Taiwan Semiconductor Manufacturing Company (TSMC) for process technology and system integration, recognizing the crucial role of advanced manufacturing in achieving their ambitious goals. This collaboration highlights the growing recognition within the semiconductor industry that networking bottlenecks are directly impacting the progress of AI.
The Growing Demand for AI-Optimized Networking
The limitations of current networking infrastructure are becoming increasingly apparent as AI models scale. Traditional networking solutions were not designed to handle the massive data flows generated by LLMs and other AI workloads. The disaggregation of inference, where tasks are broken down into smaller steps, further compounds the problem, requiring more frequent and larger data transfers. As Nedungadi explained, “The amount of data being moved per token is growing.” This increased data movement translates to higher latency, reduced throughput, and slower AI performance.
Drew Perkins, CEO of Eridu, identifies three key dimensions to the scaling challenge: scale-up (interconnecting GPUs within a single training domain), scale-out (expanding the cluster fabric), and scale-across, an emerging requirement that is gaining attention from industry standards bodies. The company believes that addressing all three dimensions is essential to unlocking the full potential of AI. This holistic approach is reflected in their decision to develop custom silicon, as they believe that off-the-shelf solutions are insufficient to meet the demands of next-generation AI applications.
Silicon Innovation as a Differentiator
A core element of Eridu’s strategy is the development of proprietary silicon. Perkins asserts that “You don’t get to an order-of-magnitude higher scale using off-the-shelf silicon.” The company is investing heavily in advanced silicon design and manufacturing, leveraging chiplet-based architectures and advanced packaging techniques to achieve unprecedented levels of performance and efficiency. This approach allows them to optimize the hardware specifically for the unique requirements of AI networking, overcoming the limitations of general-purpose networking solutions.
Omar Hassen, Chief Product Officer at Eridu, emphasizes the need for a “different technology arc” than what is currently available in the mainstream market. He believes that advanced packaging, clean-sheet design, and a focus on innovation are crucial to surpassing the capabilities of existing incumbents. “We believe we’re on the right technology arc that can take us beyond what the existing incumbents are doing,” Hassen stated. This commitment to innovation is driving Eridu’s efforts to create a fundamentally new approach to AI networking.
Mixture-of-Experts and the Networking Challenge
The rise of mixture-of-experts (MoE) models is a key driver behind the need for more sophisticated AI networking solutions. As detailed in a comprehensive survey published in March 2025, MoE models dynamically select and activate the most relevant sub-models to process input data, significantly improving performance and efficiency, particularly with large-scale, multimodal data. The survey highlights that MoEs excel in handling complex datasets but also increase the demand for data movement within the network. Eridu’s technology is specifically designed to address this challenge, enabling efficient communication between the different experts within an MoE model.
The increasing complexity of AI models, coupled with the growing adoption of techniques like MoE, is creating a significant bottleneck in networking infrastructure. Eridu’s approach, which combines custom silicon, advanced packaging, and software optimization, aims to break through this bottleneck and unlock the full potential of AI. By focusing on the unique requirements of AI workloads, the company is positioning itself as a key enabler of the next generation of AI innovation.
Eridu at a Glance
- Founded: 2023
- Funding: $200 million
- Investors: Socratic Partners, John Doerr, Hudson River Trading, Capricorn Investment Group, and Matter Venture Partners
- Headquarters: Saratoga, California
- CEO: Drew Perkins
- What they do: AI networking
Eridu’s emergence is a clear indication of the growing importance of AI-optimized networking infrastructure. As AI models continue to evolve and become more demanding, companies like Eridu will play a critical role in enabling the next wave of innovation. The company’s focus on vertically integrated solutions and custom silicon development sets it apart from traditional networking vendors and positions it to address the unique challenges of the AI era.
Looking ahead, Eridu will be focused on bringing its technology to market and scaling its operations. The company has not yet announced specific product launch dates, but This proves actively working with customers to validate its technology and refine its offerings. The next key milestone for Eridu will be demonstrating the performance and efficiency gains of its solution in real-world AI deployments. Further updates on Eridu’s progress are expected in the coming months as the company continues to build out its team and refine its technology.
Key Takeaways:
- Eridu has secured $200 million in funding to build next-generation AI networking infrastructure.
- The company is focused on addressing the growing bottleneck in data movement caused by complex AI models, including those utilizing mixture-of-experts architectures.
- Eridu’s strategy centers on a vertically integrated approach, encompassing custom silicon development, advanced packaging, and software optimization.
- The company’s partnership with TSMC highlights the importance of advanced manufacturing in enabling AI innovation.
Stay tuned to World Today Journal for further updates on Eridu’s progress and the evolving landscape of AI networking. We encourage you to share your thoughts and insights in the comments below.