Key Takeaways:
Nvidia's next-generation Kyber AI server rack faces a more than 12-month delay to 2028 due to printed circuit board manufacturing challenges, handing rivals a rare opening in the high-end AI computing market.
Nvidia Corp.'s Kyber rack-scale architecture — a server cabinet packing 144 of its most powerful graphics processing units into a single unit that functions as one giant computer — has been pushed back to 2028 from its planned 2027 debut, according to research firm SemiAnalysis. The setback stems from difficulties manufacturing a specialized multi-layer printed circuit board called a midplane that connects electronic modules within the system, SemiAnalysis said in a post Monday.
"Kyber NVL144 rack architecture has been delayed to 2028 as the PCB midplane remains challenging from a manufacturability standpoint," the firm said. The design mounts GPUs in compute trays that sit vertically rather than horizontally to boost density and reduce latency, a configuration that requires the complex midplane to route signals across all 144 chips simultaneously.
The delay extends to the larger NVL576 system — which links eight Kyber racks via optical connections — now also likely delayed or limited to small volumes, SemiAnalysis said. A backup plan to bolt two of Nvidia's current-generation racks together for comparable performance was scrapped after cloud service providers and hyperscalers rejected the design as awkward and costly to operate, according to the research firm. "It has since been cancelled due to heavy pushback from CSPs and hyperscalers over its odd design and heavy operational burden," SemiAnalysis said.
The product roadmap disruption leaves Nvidia with "no proven solution to expand the scale-up world size for Rubin Ultra," SemiAnalysis said, predicting the gap could give Advanced Micro Devices Inc. and Alphabet Inc.'s Google — whose in-house chips are already winning business from top AI labs — a rare technical opening at the high end of the market. Nvidia's current-generation Rubin systems remain in full production and begin shipping this fall to eight cloud partners, including Amazon Web Services, Microsoft Azure and Google Cloud.
PCB suppliers tied to Nvidia's supply chain bore the brunt of the market reaction. KB LAMINATES (01888.HK) plunged 13.68%, while KINGBOARD HLDG (00148.HK) fell 9.83%. Japan's Ibiden (4062.JP), for which Nvidia is the largest customer, closed down 8.37%, and Samsung Electro-Mechanics tumbled 8.09%. Nvidia shares fluctuated in premarket trading, last down less than 0.1% at $194.79.
SemiAnalysis projects Nvidia's data-center compute revenue will run 20% above Wall Street consensus in the second half of fiscal 2027, suggesting the current-generation Rubin cycle remains strong. But the Kyber delay raises questions about whether Nvidia's breakneck annual release cadence — a strategy that has kept it years ahead of competitors — is colliding with manufacturing limits. The midplane is produced on specialized PCB substrates, a supply chain node where capacity is constrained and lead times stretch beyond six months, according to industry data.
For investors, the key question is whether the delay creates a window for competitors. AMD's MI400 series, expected in 2027, and Google's sixth-generation TPU, already deployed internally, could gain traction if Nvidia cannot deliver a clear Rubin Ultra roadmap. Nvidia trades at roughly 35 times forward earnings, a premium that reflects its dominant position in AI training and inference. Any erosion of that lead — even a timing gap — could pressure the multiple.
This article is for informational purposes only and does not constitute investment advice.
