MarketMoodz
Nvidia's Kyber Rack Delayed to 2028, SemiAnalysis Says
Nvidia's Kyber NVL144 rack architecture is now expected to launch in 2028 after manufacturability problems with the PCB midplane, SemiAnalysis reports via CNBC. The slip — which also clouds the larger NVL576 system — reshuffles AI hardware timelines for cloud customers and investors.
Key Takeaways
- Kyber NVL144 rack delayed to 2028 due to PCB midplane manufacturability challenges, according to SemiAnalysis via CNBC.
- NVL576, a system that would link eight Kyber racks, is also likely delayed or could be produced only in small volumes.
- Rubin Ultra and Kyber were originally slated to debut in 2027; the timeline now pushes the launch to 2028.
- Current-generation Rubin systems are reported to ship in full production this fall to eight cloud partners including AWS, Microsoft Azure, and Google Cloud.
- SemiAnalysis projects Nvidia's data-center compute revenue about 20% above Wall Street consensus in H2 of fiscal 2027, per the report.
People Involved
- No specific individuals mentioned
Entities Involved
- Nvidia (NVDA) Developer of Kyber rack architecture and Rubin Ultra data-center chips
- SemiAnalysis Market research firm reporting the delays
- CNBC Outlet reporting SemiAnalysis's findings
- Amazon Web Services (AWS) One of eight cloud partners slated to receive Rubin systems
- Microsoft Azure One of eight cloud partners slated to receive Rubin systems
- Google Cloud One of eight cloud partners slated to receive Rubin systems
MarketMoodz Analysis
A pushed launch to 2028 matters because it delays the point at which hyperscalers can deploy next-generation rack-scale AI hardware at scale, stretching enterprise capex plans and potentially flattening near-term revenue upside tied to Kyber. For investors, the immediate questions are revenue timing and margin mix: if Kyber volumes slip, Nvidia may see later-than-expected recognition of system-level sales and weaker server-related unit growth even if chip demand remains strong. Reported projections from SemiAnalysis — that data-center compute revenue could still run roughly 20% above consensus in H2 fiscal 2027 — suggest chip demand may offset some system delays, but that view rests on estimates that should be confirmed against Nvidia's own guidance.
Operationally, PCB midplanes are complex and can become a real manufacturing choke point at rack scale; a year-long delay is plausible for a novel architecture integrating 144 chips per NVL144 rack. Historically, Nvidia has used a fast cadence of chip and system releases to hold a technology lead; a missed milestone at the rack level expands the window for competitors to test alternative architectures or win design commitments from cloud providers. Investors should watch for three signals: an official Nvidia roadmap update or earnings commentary, confirmations from cloud partners about shipment timing, and any signs of third-party suppliers resolving PCB yield issues.
Caveats: this account is sourced to SemiAnalysis via CNBC and could not be independently verified; some details, including a canceled bolt‑together backup plan and short-term stock reaction, are low-confidence. Short-term market movements will hinge on Nvidia's response and whether rivals like AMD or Google can capitalize on any gap. For portfolio managers, consider the timeline risk when sizing exposure to Nvidia and related AI infrastructure plays over the next 12–18 months.
Source: Original Article