AI reliability hinges on data access, not just models

Inc42 reported this week that reliable AI applications at scale depend on two things: up-to-date information and clear rules governing what they can access. The observation isn’t new—enterprises have been wrestling with this for over a year—but it’s becoming harder to ignore. As AI agents and automated systems interact more directly with enterprise data, the infrastructure supporting them is emerging as the real bottleneck, not the models themselves.

This isn’t just theoretical. Startups are raising money to address it. Zeit AI, which automates enterprise data engineering, secured €5 million in seed funding earlier this month. Delos Data, founded by former Intel engineers, raised $100 million to build chips that accelerate data movement in AI data centers. Even Oracle is adjusting its security controls to account for AI agents that bypass traditional application-level protections. The common thread: data access and control are now the critical factors determining whether AI initiatives succeed or stall.

The trend reflects a broader shift in enterprise AI strategy. When we covered the pivot from models to data infrastructure last month, it was still early. Now, the evidence is piling up. Together AI’s partnership with Saudi firm Humain to access non-U.S. data centers isn’t just about compute—it’s about where and how data can be processed. Delos Data’s focus on hardware for faster data movement speaks to the same challenge: AI applications can’t scale if they’re starved of the right information at the right time.

Investors are taking notice. The €5 million for Zeit AI and $100 million for Delos Data suggest confidence that data infrastructure is the next frontier. But the market is still figuring out what “clear access rules” actually look like. Incumbents like Oracle are retrofitting security for AI agents, but startups are building from scratch. The tension is whether enterprises will adopt new tools or try to adapt old ones—an open question that will shape funding and acquisitions in the coming quarters.

What’s next? Watch for more deals like Together AI’s, where startups seek alternative data centers to avoid regulatory or operational constraints. Expect Oracle and others to double down on security layers tailored for AI agents. And pay attention to whether Delos Data’s chip-focused approach gains traction—if data movement is the bottleneck, hardware could become the next battleground. The model wars are over. The data wars have just begun.

Sources: inc42.com