Today’s data infrastructure was built for humans. In the future, we believe the primary consumer of data will become AI systems. This shift raises a fundamental question: what happens when you optimize data infrastructure not for dashboards and batch jobs, but for scale, throughput, cost, and composability?

Spiral answers this question by entirely reimagining the modern data stack from the ground up. That’s why we’re leading their Series A.

Data for AI is Different

Applications have always relied on three building blocks: data, algorithms, and user experience. Typically, major platform shifts have transformed one or two of these layers: cloud computing redefined algorithms, for example, and mobile reshaped user interaction. However, in this AI era, all three are transforming at once.

People often notice UX changes (e.g., chatbots, copilots) and algorithmic developments (e.g., LLMs), but what is often overlooked and long overdue for reinvention is the data layer itself. Today’s systems are still optimized for a world of reports and dashboards. They rely on relational databases, schema-bound warehouses, and batch ETL pipelines.

AI systems need something very different: the ability to work with messy, multimodal, high-volume data, such as embeddings, documents, images, audio, and knowledge graphs. They require real-time streams, low-latency retrieval, unprecedented scale, tolerance for ambiguity, and a cost structure that makes continuous adaptation viable. In short, they need a new kind of data system, built from the ground up for machines, not humans, as the primary consumer.

Resolving the Invisible Bottleneck

While there has been plenty of innovation across vector databases, feature stores, and the streaming engine layer, the bottom of the stack has seen little change. The file formats that quietly move and store the world’s data have remained virtually the same for decades.

Parquet and other legacy columnar formats were built for the MapReduce era, optimized for CPU-bound analytics on structured data. But today’s workloads look very different. That’s why Spiral started by creating Vortex, a next-generation open columnar format purpose-built for modern AI data systems: it delivers 200x faster random-access reads, 2–10x faster scans and writes, and GPU-native decompression that streams data directly from S3 into GPU memory.

“Modern GPUs can consume terabits per second, but legacy formats force CPUs to sit in the middle, decompressing everything first. That’s broken. Vortex fixes it.” - Will Manning, co-founder & CEO of Spiral

Beyond File Formats

While Vortex serves as the foundation, the company is now building on it with Spiral, a database purpose-built and reimagined from the ground up for the AI era. Spiral powers real-time, multimodal, AI-native data systems at unprecedented scale and sustainable cost. It delivers GPU-saturating throughput, unified access across all data types, and end-to-end permissioning, all running natively on object storage. 

Working with visionary founders often feels like catching a glimpse of the future. That’s exactly the case with Spiral’s founders, Will Manning, Rob Kruszewski, and Nick Gates. In our view, they are three of the sharpest technical minds in data infrastructure. With deep systems expertise and experience building critical data infrastructure at Palantir and Citadel, organizations that faced these challenges at massive scale, they bring a deep clarity on where data infrastructure is headed and what it will take to get there.

We look forward to partnering with the Spiral team as they build the data infrastructure for the AI era.