The rapid expansion of artificial intelligence (AI) enterprises has significantly increased the demand for computing power. Companies such as CoreWeave, Together AI, and Lambda Labs have capitalized on this surge by offering distributed computing capabilities. However, data storage remains predominantly centralized, with most organizations relying on major cloud providers like Amazon Web Services (AWS), Google Cloud, and Microsoft Azure. These traditional storage systems are designed to keep data close to their own computing resources, limiting flexibility across multiple clouds or regions.
Ovais Tariq, co-founder and CEO of Tigris Data, aims to address this limitation. Modern AI workloads and infrastructure are opting for distributed computing over traditional big cloud solutions, Tariq explained. We intend to offer a similar approach for storage because, without efficient storage, computing capabilities are compromised.
Tigris Data, established by the team behind Uber’s storage platform, is developing a network of localized data storage centers tailored to meet the needs of contemporary AI workloads. The company’s AI-native storage platform is designed to:
– Synchronize with computing resources, allowing data to automatically replicate to locations where GPUs are situated.
– Support vast quantities of small files.
– Provide low-latency access essential for training, inference, and agentic workloads.
To advance these objectives, Tigris recently secured a $25 million Series A funding round led by Spark Capital, with participation from existing investors, including Andreessen Horowitz. This strategic move positions Tigris in direct competition with established cloud giants, which Tariq refers to as Big Cloud.
Tariq critiques these incumbents for offering data storage services that are not only more costly but also less efficient. Historically, AWS, Google Cloud, and Microsoft Azure have imposed egress fees—often termed the cloud tax—on customers wishing to migrate data to another provider or download data for purposes such as utilizing more affordable GPUs or training models across different global regions. This practice can be likened to a gym charging extra fees for members who decide to terminate their membership.
Batuhan Taskaya, head of engineering at Fal.ai and a Tigris customer, highlighted the financial impact of these fees, noting that they previously constituted the majority of Fal’s cloud expenditures.
Beyond the financial implications of egress fees, Tariq points out the latency issues associated with larger cloud providers. Egress fees are merely a symptom of a deeper problem: centralized storage systems that cannot keep pace with a decentralized, high-speed AI ecosystem, he stated.
Tigris serves over 4,000 customers, many of whom are generative AI startups developing image, video, and voice models that require handling large, latency-sensitive datasets. For instance, in scenarios involving AI agents processing local audio, low latency is crucial. Having compute and storage resources in close proximity ensures rapid data access, enabling developers to run AI workloads more reliably and cost-effectively using decentralized cloud solutions.
Taskaya emphasized the benefits of Tigris’s approach: Tigris enables us to scale our workloads across any cloud by providing consistent access to the same data filesystem from all locations without incurring egress charges.
Several factors drive the need for data proximity to distributed cloud options. In highly regulated industries like finance and healthcare, ensuring data security is a significant barrier to adopting AI tools. Additionally, companies are increasingly seeking greater control over their data. Tariq referenced Salesforce’s recent decision to prevent AI competitors from accessing Slack data as an example. Organizations are becoming more aware of the critical role data plays in fueling large language models and AI, Tariq observed. They desire more control and are reluctant to entrust it to external entities.
With the new funding, Tigris plans to expand its network of data storage centers to meet growing demand. The company has experienced an eightfold annual growth since its inception in November 2021. Currently, Tigris operates data centers in Virginia, Chicago, and San Jose, with plans to extend its presence in the U.S. and expand into Europe and Asia, targeting cities such as London, Frankfurt, and Singapore.
