This distributed data storage startup wants to take on Big Cloud

The surge in artificial intelligence (AI) companies has created an unprecedented demand for computing resources, prompting startups like CoreWeave, Together AI, and Lambda Labs to thrive by offering distributed computing capabilities. However, many organizations still rely heavily on the leading cloud providers—AWS, Google Cloud, and Microsoft Azure—for data storage. These established services were primarily designed to keep data close to their own computing resources, rather than allowing it to be distributed across various clouds or regions. Ovais Tariq, co-founder and CEO of Tigris Data, shared insights with TechCrunch, stating, "Modern AI workloads and infrastructure are adopting distributed computing in lieu of traditional big cloud solutions. We aim to extend this approach to data storage, as compute power is meaningless without effective storage solutions." Founded by the team behind Uber's storage platform, Tigris Data is developing a network of localized data storage centers to meet the requirements of contemporary AI workloads. Tariq described their AI-native storage platform as one that automatically adapts to compute needs, allowing data to replicate to locations where GPUs are positioned, while also supporting billions of small files and ensuring low-latency access for tasks like training and inference. Recently, Tigris secured $25 million in Series A funding, led by Spark Capital and supported by existing investors such as Andreessen Horowitz. The startup is positioning itself against established giants, which Tariq refers to as "Big Cloud." He criticizes these incumbents for not only providing pricier data storage options but also for their inefficiencies. Historically, these major providers have implemented egress fees—often referred to as "cloud tax"—when customers wish to migrate to other cloud services or download data for use with cheaper GPUs or to train models in different geographical areas. Tariq likens this to having to pay extra at a gym if you choose to stop your membership. Batuhan Taskaya, head of engineering at Fal.ai, one of Tigris' clients, revealed that such costs once constituted a significant portion of their cloud expenses. In addition to egress fees, Tariq points out the latency issues often faced with larger cloud providers, stating, "Egress fees are merely a symptom of a larger issue: centralized storage that fails to keep pace with a decentralized and rapid AI ecosystem." Most of Tigris' over 4,000 clients are generative AI startups focused on developing image, video, and voice models, which typically require large datasets that demand low latency. Tariq emphasized the importance of local compute and storage, saying, "Imagine interacting with an AI agent that processes audio locally. You need the lowest latency possible." Large cloud services, he argues, are not tailored for AI workloads, as streaming extensive datasets for training or real-time inference across various regions can lead to latency bottlenecks that hinder model performance. In contrast, localized storage allows for quicker data retrieval, enabling developers to run AI workloads more reliably and cost-effectively through decentralized clouds. Taskaya from Fal.ai noted, "Tigris enables us to scale our workloads across any cloud while offering access to the same data filesystem without incurring egress fees." Additionally, organizations are increasingly motivated to keep their data close to distributed cloud options, especially in regulated industries like finance and healthcare, where data security is critical. Tariq highlighted a growing awareness among companies regarding the importance of data, citing how Salesforce recently prevented its AI competitors from accessing Slack data. "Organizations are realizing how vital their data is in driving large language models and fueling AI initiatives," Tariq stated. "They seek more control over their data rather than leaving it in the hands of others." With their recent funding, Tigris plans to expand its data storage centers to accommodate rising demand, with plans for further growth following an 8x increase in scale since the startup's inception in November 2021. Currently, Tigris operates three data centers in Virginia, Chicago, and San Jose, with aspirations to expand across the U.S., Europe, and Asia, targeting cities like London, Frankfurt, and Singapore.