Why You Care
Are you tired of hidden fees and data bottlenecks when training your AI models? For many AI companies, storing and moving data efficiently is a constant headache. This issue often adds significant costs and slows down creation. A new player, Tigris Data, just secured a substantial investment to tackle this very problem. They aim to free your data from the grip of “Big Cloud” providers.
What Actually Happened
Tigris Data recently announced a significant financial milestone. The company raised a $25 million Series A funding round, as mentioned in the release. This investment was led by Spark Capital, with existing investors like Andreessen Horowitz also participating. Tigris Data is building a distributed data storage network. This network is specifically designed for the demanding needs of modern artificial intelligence workloads. The team behind Tigris previously developed Uber’s storage system, according to the announcement. They are now applying that expertise to create an AI-native storage approach.
Their system aims to address a growing challenge. While many AI companies use distributed computing services, most still rely on major cloud providers for data storage. These traditional systems were built to keep data close to their own computing resources. This setup often creates inefficiencies for AI applications that require data spread across various locations.
Why This Matters to You
If you’re involved in AI creation, you know data movement can be costly and slow. Tigris Data’s approach could significantly change how you manage your AI data. Their system promises to move data with your compute resources. It also automatically replicates data to where GPUs are located, as detailed in the blog post. This ensures low-latency access for essential tasks like training, inference, and agentic workloads. Think of it as having your data always ready and waiting exactly where your processing power needs it, without extra steps or delays.
For example, imagine you are training a large language model. This model requires massive datasets across different geographic regions. With traditional cloud storage, moving data between these regions incurs significant egress fees – often called a “cloud tax.” Tigris aims to eliminate or drastically reduce these costs. This allows you to improve your spending and accelerate your creation cycles.
“Modern AI workloads and AI infrastructure are choosing distributed computing instead of big cloud,” Ovais Tariq, co-founder and CEO of Tigris Data, told TechCrunch. “We want to provide the same option for storage, because without storage, compute is nothing.”
How much could you save by avoiding these data transfer fees?
Key Benefits of Tigris Data’s Approach:
- Cost Efficiency: Reduces or eliminates egress fees.
- Performance: Low-latency access for AI workloads.
- Flexibility: Data replicates to GPU locations automatically.
- Scalability: Supports billions of small files efficiently.
The Surprising Finding
Here’s an interesting twist: the high cost of moving data, not just storing it, is a major pain point. While many assume storage costs are the primary concern, the research shows egress fees can be crippling. Batuhan Taskaya, head of engineering at Fal.ai, a Tigris customer, highlighted this. He stated that these costs once accounted for the majority of Fal’s cloud spending. This finding challenges the common assumption that simply finding cheaper storage is the main goal. Instead, the ability to move data freely and affordably is paramount for AI companies. The incumbents, referred to as “Big Cloud” by Tariq, often charge customers to migrate data. This practice can lock companies into their ecosystems, making it expensive to switch providers or utilize diverse computing resources.
What Happens Next
With this new funding, Tigris Data is set to expand its localized data storage network. We can expect to see more data centers coming online in the next 12-18 months. This expansion will likely target key regions where AI creation is booming. For example, a company could soon seamlessly train models using GPUs in one cloud provider. Simultaneously, they could run inference on another, all while their data is managed efficiently by Tigris. This offers a level of flexibility previously hindered by cloud provider lock-in.
Companies should start evaluating their data storage strategies. Consider how much you currently spend on egress fees. Explore how a distributed, AI-native storage approach could benefit your operations. The industry implication is clear: the dominance of “Big Cloud” in data storage for AI workloads may face a significant challenge. As Tariq stated, “We want to provide the same option for storage, because without storage, compute is nothing.”