Nvidia Kyber AI rack delayed to 2028 as circuit board manufacturing hits roadblock

Nvidia's next-gen AI rack system delayed to 2028 on manufacturing snags, SemiAnalysis says

NVIDIA's next marquee product -- the Kyber rack-scale architecture designed to house its 2027 Rubin Ultra chips -- has been delayed by more than 12 months to 2028, according to research firm SemiAnalysis, the latest in a string of reported setbacks raising questions about the AI giant's product roadmap. Kyber is a server cabinet that packs 144 of Nvidia's most powerful chips into a single unit so they can work together as one giant computer, providing the horsepower AI companies need to train and run their most advanced models. The design mounts graphics processing units in compute trays that sit vertically instead of horizontally to boost density and reduce latency, and had been slated to debut with Vera Rubin Ultra, Nvidia's next-generation rack-scale system, in 2027. The setback stems from difficulties manufacturing a key circuit board at the heart of the system, SemiAnalysis said in a post on Monday. "Kyber NVL144 rack architecture has been delayed to 2028 as the PCB midplane remains challenging from a manufacturability standpoint," the firm said, referring to a specialized, multi-layer printed circuit board that connects electronic modules within a system. NVL576 -- a larger system linking eight racks via optical connections -- is also likely delayed or limited to small volumes, the research firm said. Nvidia did not respond to CNBC's request for comment. The reported delay adds to mounting strains across Nvidia's product lines, underscoring concerns that Nvidia's breakneck annual release cadence is colliding with manufacturing limits. A backup plan -- bolting two of Nvidia's current-generation racks together for similar power -- has also been scrapped after cloud customers rejected the design as awkward and costly to operate. "It has since been cancelled due to heavy pushback from CSPs [cloud service providers] and hyperscalers over its odd design and heavy operational burden," SemiAnalysis said. That leaves Nvidia with "no proven solution to expand the scale-up world size for Rubin Ultra," SemiAnalysis said, predicting that could give rivals Advanced Micro Devices and 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 are in full production and begin shipping this fall to eight cloud partners, including Amazon Web Services, Microsoft Azure and Google Cloud. SemiAnalysis also projects Nvidia's data-center compute revenue will run 20% above Wall Street consensus in the second half of fiscal 2027. Shares of Nvidia fluctuated in premarket trading, last down less than 0.1% at $194.79.

Nvidia's Kyber AI rack is delayed to 2028

The Kyber rack that was meant to house Nvidia's 2027 Rubin Ultra chips has slipped to 2028, tripped up by one hard-to-build circuit board, research firm SemiAnalysis says. The delay leaves Nvidia without a proven way to scale up its most powerful systems, and hands AMD and Google a rare opening. Nvidia's next flagship AI machine has hit a wall, and the culprit is a single circuit board. The Kyber rack meant to house its 2027 Rubin Ultra chips has slipped to 2028. Research firm SemiAnalysis flagged the delay, and CNBC reported it. Kyber is not a chip. It is a server cabinet. It packs 144 of Nvidia's most powerful GPUs into one unit, so they behave like a single giant computer. That density is what trains and runs the largest AI models. One board, one big problem The hold-up sits in an unglamorous part. Kyber mounts its chips in vertical trays to cram more in and cut latency. Tying it all together is a multi-layer circuit board called the PCB midplane. That board, SemiAnalysis said, "remains challenging from a manufacturability standpoint." In plain terms, Nvidia cannot yet build it at scale. The knock-on effects spread from there. A larger system, the NVL576, links eight racks with optical cables. It is likely delayed too, or held to tiny volumes. Nvidia did not comment. No plan B Nvidia had a backup. It planned to bolt two current-generation racks together for similar power. Cloud customers hated it. They pushed back hard on the "odd design and heavy operational burden," and Nvidia scrapped the idea. That leaves a gap. SemiAnalysis says Nvidia now has "no proven solution" to scale up its most powerful Rubin Ultra systems. For a company that ships a new architecture every year, missing a rung on the ladder is rare. A window for rivals The timing helps the competition. Google and AMD already win work from top AI labs with their own chips. A stumble at the very high end hands them a rare technical opening, SemiAnalysis notes. It also feeds the current nervousness in the AI chip trade. Asian technology and circuit-board stocks slid on the report, Bloomberg noted. None of this dents Nvidia's near-term grip. Its current Rubin systems are in full production. They start shipping this autumn to eight cloud partners, including AWS, Microsoft Azure and Google Cloud. SemiAnalysis even sees Nvidia's data-centre compute revenue running 20 per cent above Wall Street forecasts. That covers the second half of its 2027 financial year. Nvidia shares barely moved. The real squeeze The story underneath is manufacturing. Nvidia's yearly release pace is colliding with the limits of what its suppliers can make. That runs from advanced circuit boards to the memory that feeds each GPU. The factories assembling these racks are already running hot. Rivals building their own custom silicon are betting that pace, not design, is where Nvidia is most exposed. For once, the hardest part of the AI boom is not the chip. It is the box around it.