Nvidia’s 2028 Kyber Delay Lifts 3 AI Chip Rivals as Customers Reassess Infrastructure Bets
Updated
Updated · Simply Wall St · Jul 6
Nvidia’s 2028 Kyber Delay Lifts 3 AI Chip Rivals as Customers Reassess Infrastructure Bets
3 articles · Updated · Simply Wall St · Jul 6
Summary
Nvidia’s more than 12-month delay to the Kyber NVL144 rack system has put Marvell, Tower Semiconductor and Lattice Semiconductor in focus as potential beneficiaries in AI infrastructure.
The setback, tied to manufacturing problems with Kyber’s PCB midplane, gives rivals extra time to refine products, deepen customer ties and pitch alternatives beyond Nvidia-centric systems.
Marvell, with a US$214.6 billion market value and US$8.7 billion in revenue, is seen as the closest value-chain play through custom XPUs, optical interconnects and silicon photonics, though rich valuation and customer concentration remain risks.
Tower and Lattice offer narrower but credible angles: Tower has about US$1.3 billion in 2027 silicon photonics contracts, while Lattice is pushing low-power FPGAs for server control, security and edge AI.
The broader implication is that a delay in Nvidia’s flagship rack-scale rollout could widen investor and customer interest across specialty chipmakers supplying the AI buildout.
NVIDIA redesigned its flagship chip and delayed its super-rack. Is this a sign of weakness, or a calculated move in the AI arms race?
NVIDIA’s superchip hits a manufacturing wall. Can rivals like AMD and Google truly seize this rare opening to challenge its AI dominance?
With the power grid years behind AI's needs, is NVIDIA's hardware delay a crisis or a necessary reality check for the entire industry?
Nvidia’s Kyber Rack Delay and Rubin Ultra Dual-Die Pivot: Advanced Packaging Bottlenecks Disrupt AI Roadmaps
Overview
Nvidia's flagship Kyber rack AI system has been delayed, with its core Rubin Ultra GPU shifting from a planned quad-die to a dual-die design due to major manufacturing and packaging challenges. These issues, including yield and thermal problems from packaging partners, have forced Nvidia to adjust its ambitious plans. As a result, the delay is impacting the timelines for large-scale AI model training and deployment, affecting the competitive landscape and investment strategies across the AI sector. The situation highlights the difficulties of pushing technological limits and the ripple effects such changes have on the broader industry.