YWO NEWS | EQUITIES | DEEP DIVE / ANALYSIS
Nvidia’s next-generation Kyber NVL144 rack architecture has been pushed back to 2028 after manufacturing difficulties with a key circuit board proved insurmountable on the original timeline, research firm SemiAnalysis reported on Monday, CNBC reported. The system had been slated to debut alongside Vera Rubin Ultra in 2027, making this a slip of more than 12 months for what Nvidia had positioned as the centrepiece of its next rack-scale generation.
NVDA opened Monday’s premarket session last down less than 0.1% at $194.79, per CNBC data. The stock showed little movement in premarket trading following the report. .
The PCB Midplane Problem
Kyber is not a chip — it is a server cabinet that packs 144 of Nvidia’s most powerful GPUs into a single unit, wiring them to function as one integrated compute block. The system mounts chips vertically rather than horizontally to increase density and cut latency. At its heart sits a specialised multi-layer printed circuit board — the PCB midplane — that connects the electronic modules within the cabinet.
That midplane is where the programme has stalled.
“Kyber NVL144 rack architecture has been delayed to 2028 as the PCB midplane remains challenging from a manufacturability standpoint,” SemiAnalysis said, as reported by CNBC.
The NVL576 — a larger configuration linking eight Kyber racks through optical connections — is also either delayed or constrained to small volumes, SemiAnalysis added, per the same CNBC report.
Nvidia did not respond to CNBC’s request for comment, CNBC noted.
The Backup Plan Was Already Dead
What makes the timeline slip more consequential is what happened to the contingency. Nvidia had explored bridging the gap by bolting two current-generation racks together to approximate Kyber’s compute footprint. Cloud service providers and hyperscalers rejected the workaround outright.
“It has since been cancelled due to heavy pushback from CSPs and hyperscalers over its odd design and heavy operational burden,” SemiAnalysis said, per CNBC.
According to SemiAnalysis, the alternative design was not pursued further. . SemiAnalysis concluded that Nvidia now has “no proven solution to expand the scale-up world size for Rubin Ultra,” as CNBC reported, citing the research firm directly. The double-rack design was reportedly presented at Nvidia’s GTC conference in San Jose on 16 March 2026, where Jensen Huang appeared onstage next to a Vera Rubin Ultra Kyber Compute Tray and a Vera Rubin Ultra Kyber NVLink MidPlane.
Where AMD and Google May Find a Gap
SemiAnalysis flagged that the void at the top of the rack-scale market could give Advanced Micro Devices and Google, whose in-house chips have already been winning business from top AI labs, a technical opening they have not had before, according to CNBC.
The framing matters. Nvidia has operated on a roughly annual cadence at the chip level, and competitors have struggled to match the release rhythm. SemiAnalysis noted that any delay to Nvidia’s rack-scale roadmap could be relevant to competitors operating in the same market segment.
That said, the current-generation picture remains intact. Nvidia’s existing Rubin systems are in full production and begin shipping this autumn to eight cloud partners, including Amazon Web Services, Microsoft Azure, and Google Cloud, CNBC reported. And SemiAnalysis projected Nvidia’s data-centre compute revenue will run 20% above Wall Street consensus in the second half of fiscal 2027 .
Cadence Under Strain
The Kyber slip is part of a pattern SemiAnalysis characterised as Nvidia’s annual release cadence colliding with manufacturing limits, per CNBC. Designing ever-denser rack architectures is one problem. Getting Taiwan’s manufacturing base to produce the specialised multi-layer PCBs at volume and yield is another.
Nvidia’s chip roadmap and its rack-scale architecture roadmap are no longer moving in lockstep. The chip — Rubin Ultra — is on schedule for 2027. The cabinet designed to house it at full scale is now a year behind. For hyperscalers who have already committed capex plans around the Kyber timeline, that divergence may affect product deployment timelines for some customers.
The commercial impact of the reported delay remains uncertain based on publicly available information. . What SemiAnalysis and Anniek Bao’s CNBC report do establish is that Nvidia has no announced bridge product for the highest-density configurations and no confirmed revised timeline beyond the 2028 guidance from SemiAnalysis.
The 20% data-centre revenue beat projected for the second half of fiscal 2027 may help offset some near-term operational concerns. However, the longer-term competitive impact of the reported delay remains uncertain. .
Risk Disclaimer: Trading CFDs involves substantial risk and may not be suitable for all investors. CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You may lose some or all of your invested capital. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money. Past performance is not a reliable indicator of future results. This article is provided for general informational and educational purposes only and does not constitute financial, investment, legal, tax, or trading advice, nor a recommendation, solicitation, or offer to buy or sell any financial instrument.