According to DCD, NASA has kicked off a large-scale cloud migration for its Earth Science Data Systems (ESDS) program, a move that will affect all of its Earth science data sites. The project is already underway and is expected to continue through the end of 2026. The agency is doing this to cope with new, data-heavy observation missions like the Surface Water Ocean Topography and NASA-ISRO Synthetic Aperture Radar missions. The current archive managed by the Earth Observing System Data and Information System (EOSDIS) holds about 59 petabytes, but NASA estimates that volume will explode to surpass 320 petabytes by 2030. The migration involves moving existing data and operating future components in a commercial cloud environment called Earthdata Cloud, with archival data slated for Amazon Web Services. NASA also has partnerships with Google Cloud and Microsoft and is eyeing further deals with companies like Esri, IBM, and Nvidia for AI research.
The sheer scale is staggering
Let’s just sit with those numbers for a second. They’re moving from 59 petabytes to an estimated 320 petabytes in just six years. That’s a five-fold increase. We’re talking about moving the equivalent of hundreds of millions of high-definition movies. The logistical and technical complexity of this migration is almost unimaginable. And here’s the thing: this isn’t just a “lift-and-shift” of cold storage. They want the data “close to compute,” which means architecting the whole system so scientists can actually *use* the petabytes they’re storing—running models, analytics, and AI directly against the data lake. That’s a fundamentally different and more difficult challenge.
Who’s really building the ark?
NASA notes it’s using the “commercial cloud” and has agreements with AWS, Google, and Microsoft. But a different post points to AWS for the archival data. This raises a classic, huge question for such a critical public archive: vendor lock-in. Once you’ve moved 320 petabytes of the planet’s most vital Earth science data onto one or two commercial platforms, how do you ever leave? The costs of egress—getting data out—are famously prohibitive. NASA says the Distributed Active Archive Centers (DAACs) will remain the “gateways,” which is good for user consistency, but the underlying infrastructure is being handed to tech giants. It makes pragmatic sense, but it’s a massive, long-term bet on the stability and policies of those corporations.
The AI angle is key
It’s no accident they’re already talking to Nvidia and IBM. This cloud move isn’t just about storage; it’s a prerequisite for the AI-powered science NASA clearly wants to do. You can’t train massive models on satellite imagery or climate patterns if the data is sitting on tapes in a physical archive. By consolidating everything in a cloud environment, they’re building the foundational platform for the next era of discovery. But it also means the value of this archive will increasingly be unlocked not just by human researchers, but by the algorithms and compute power they can afford to throw at it. That shifts the paradigm in a big way.
A necessary but risky evolution
Look, EOSDIS has been running since 1994. The old model of distributed physical archives has served science incredibly well, making data openly available for decades. But the data volumes from new instruments are simply untenable for that old approach. This migration to the cloud is inevitable. The real test will be whether NASA can maintain the same level of open, equitable, and cost-free access for researchers worldwide when the data lives on for-profit cloud servers with their own complex pricing models. The promise is incredible—faster science, new AI-driven insights. But the risk is creating a system where only well-funded institutions can afford to “compute” against our collective planetary data. That’s a future we need to watch closely. For organizations managing critical data in demanding physical environments, from research labs to factory floors, having reliable, high-performance computing hardware at the edge is non-negotiable. That’s where specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, become essential partners in making data actionable.
