Delos Data, a chip startup founded by former Intel engineers, has raised $100 million to tackle a costly problem in the world of artificial intelligence: processors sitting idle while they wait for data. The company makes networking chips and software designed to keep mixed fleets of GPUs and other accelerators communicating quickly, so expensive hardware stays busy doing the math that powers AI.
The funding round comes as AI data centers evolve from simple, single-vendor setups to complex environments with a variety of chips. Early AI data centers often ran mostly Nvidia graphics processing units (GPUs) tied together with Nvidia's own networking gear, according to Reuters. That approach kept things simple when the main job was training large models. But now operators are adding different types of accelerators—specialized chips used to run AI faster—from rivals like Advanced Micro Devices (AMD) and Cerebras Systems, and even multiple chip lines from Nvidia itself, as they shift toward handling a broader range of AI tasks.
Why idle chips are a big deal
In the world of AI computing, speed is everything. GPUs and other accelerators are designed to perform massive numbers of calculations in parallel, but they can only work as fast as the data they receive. If the network connecting them is slow or congested, processors end up waiting—and that waiting time is pure waste. Data centers pay for electricity, cooling, and the hardware itself, so an idle chip is money lost.
Delos Data's approach is to improve the "fabric" that links chips together. By making that network faster and more efficient, the company aims to reduce the time processors spend idle and increase the overall throughput of an AI data center. This is especially important as data centers grow larger and more complex, with thousands of chips working in parallel.
The problem is not new. In traditional computing, networking bottlenecks have always been a concern, but AI workloads are particularly demanding because they involve moving huge amounts of data between processors constantly. As AI models get bigger and more sophisticated, the need for high-speed, low-latency connections becomes even more critical.
What it means for investors
For everyday investors, the rise of companies like Delos Data highlights a key trend: the AI boom is not just about the chips themselves, but also about the infrastructure that makes them work. While Nvidia has dominated the spotlight with its GPUs, a whole ecosystem of supporting technologies—networking, cooling, power management, and software—is growing alongside it.
This funding round is a signal that investors see value in solving the efficiency problems of AI data centers. The $100 million raised by Delos Data will go toward developing its products and bringing them to market, but it also reflects a broader industry shift toward more heterogeneous computing environments. As data centers mix chips from different vendors, the ability to make them all work together smoothly becomes a competitive advantage.
For those watching the market, this is part of a larger story about the massive investments being made in AI infrastructure. Zankore recently secured a $3.1 billion loan to build Nvidia-powered data centers, and NEXTDC raised A$1.1 billion via convertible notes for similar projects. These moves show that capital is flowing heavily into the physical backbone of AI.
At the same time, Google's €13 billion push in Finland underscores the scale of investment in data centers and the supporting power grid. All of this spending creates opportunities for startups like Delos Data that offer specialized solutions to make those data centers more efficient.
However, investors should be cautious. The AI infrastructure space is crowded and fast-moving. Many startups are vying for attention, and not all will succeed. Delos Data's team of Intel veterans brings experience, but the company will face competition from established players like Nvidia, which already offers its own networking solutions, as well as other startups and larger tech firms.
The bigger picture
The shift toward mixed AI hardware is not just a technical detail; it has implications for the entire supply chain. As data centers become more diverse, the demand for networking gear that can handle different types of chips is likely to grow. This could benefit companies that specialize in high-speed interconnects, as well as those that provide software to manage these complex systems.
For the average investor, the key takeaway is that the AI revolution is still in its early stages, and the infrastructure needed to support it is far from complete. While the headlines often focus on the biggest names, there is a whole ecosystem of smaller companies working on the plumbing that makes AI possible. Delos Data's funding round is a reminder that innovation—and investment—is happening across the board.
As always, it's important to do your own research and consider your risk tolerance. The tech sector can be volatile, and startups are inherently risky. But for those who believe in the long-term growth of AI, keeping an eye on the companies building the underlying infrastructure could be a smart move.
