Nvidia’s Rubin CPX Inference GPU Launch in Doubt After Report Finds No GDDR7 Memory or Substrate Orders
Nvidia’s planned launch of Rubin CPX, an inference GPU it had previously slated for the second half of 2026, is in doubt after a report citing multiple industry sources said the company is not placing orders for related memory or substrates. The report said no activity had been confirmed around the GDDR7 graphics DRAM Rubin CPX was expected to use, raising the possibility that the program has been withdrawn or materially revised.
Nvidia had said Rubin CPX would carry 128 gigabytes of GDDR7 across eight 16-gigabyte chips, and suppliers had expected the product to broaden demand for GDDR7 memory and related substrates in AI servers. The report said Nvidia removed Rubin CPX from its roadmap at GTC 2026 in March. The update centers on the CPX inference variant and contrasts with earlier supply-chain reports that Nvidia’s first big batch of Rubin chips should complete final packaging in July-August and begin shipping to system assembly partners, with Taiwan Semiconductor Manufacturing Co. targeting mass production in the third quarter.
From the sources (4 posts)
@dnystedtFoxconn has won orders across all three major Nvidia AI server platforms, which are expected to cause 2nd half-2026 revenue to surge, media report, particularly as Morgan Stanley estimates the build-of-materials for Vera Rubin servers (VR20
@dnystedtNvidia’s first big batch of Rubin chips will finish (final packaging) in July-August and start being delivered to system assembly partners, media report, citing unnamed supply chain sources. TSMC began manufacturing the initial batch in lat
@firstadopterOh. $NVDA
@jukan05[Exclusive] NVIDIA's "Rubin CPX" Launch in Doubt… No Memory or Substrate Orders The launch of NVIDIA's inference GPU "Rubin CPX" has become uncertain. Although the company had initially announced plans to release it in the second half of t