Jetson TX2 and Xavier Hit End of Life as Memory Supply Tightens

Nvidia has moved its entire TX2 and Xavier Jetson module families to Non-Cancelable, Non-Returnable (NCNR) status. Final purchase orders must land by July 1, existing orders convert to NCNR on July 15, and last shipment closes July 15 of next year.

The trigger is manufacturing capacity. Memory vendors have progressively reallocated production toward higher-margin HBM and DDR5 in response to AI accelerator demand, leaving LPDDR4 — the memory type underpinning TX2 and Xavier — undersupplied and difficult to source at scale. Connect Tech, a Canadian system integrator and Nvidia channel partner, confirmed NCNR status for its own inventory and stated the deadlines are "based on Nvidia timelines."

Confirmed NCNR modules include the Jetson TX2 NX, Jetson TX2i (all SKUs), Jetson AGX Xavier 32GB Industrial, and Jetson Xavier NX in both 8GB and 16GB configurations. Connect Tech indicates Nvidia has flagged all TX2 and Xavier variants on its side, suggesting the scope is broader than the SKU list alone. The TX2 family dates to 2017; Xavier parts launched in 2018, with some industrial variants shipping as recently as 2021.

For engineering and procurement teams running robotics, autonomous vehicle, or factory automation stacks on these modules, the impact is immediate and unbudgeted. NCNR status removes the option to cancel or return orders, converting any in-flight procurement into a firm liability. Organizations that have not yet locked in multi-year supply agreements now compete in a closing window against every other integrator in the same position. The July 1 purchase-order deadline is weeks away, not quarters.

Nvidia points to its Orin family as the migration path. Orin runs on LPDDR5 and remains available, though at elevated prices. For Xavier NX deployments, the Orin NX is characterized as close to a drop-in replacement provided designs do not rely on specialized I/O configurations; form factor and power envelope are comparable. AGX Xavier-to-AGX Orin transitions are more straightforward, as both use the same 699-pin connector family, though power delivery and thermal validation are still required. For production hardware certified in regulated industries — medical robotics, industrial safety systems — requalification cycles can stretch months.

The broader signal is structural, not cyclical. This is not a temporary shortage that resolves when a fab cycle completes. Memory manufacturers have made a deliberate, durable choice to exit high-volume LPDDR4 production in favor of architectures with better margin and AI-workload alignment. Any embedded or edge platform still dependent on legacy DRAM nodes faces the same pressure: constrained supply, NCNR status, and accelerated end-of-life, regardless of deployment timeline.

Nvidia has not issued a public statement directly confirming the timelines; all deadlines currently derive from Connect Tech's supplier communication. Teams should verify directly with their Nvidia distribution channel before treating the July dates as authoritative. What is not in question is direction: LPDDR4-based Jetson modules are exiting the supply chain, and Orin migration is now a procurement imperative, not a roadmap aspiration.