Version 2

NVIDIA cancelled the original 4-die Rubin Ultra GPU announced at GTC 2026 due to manufacturing concerns; the replacement carries the same name but delivers roughly half the original design's performance [1]. SemiAnalysis, which reported the cancellation, also projects NVIDIA datacenter compute revenue 20% above consensus for 2H FY2027, attributing the Rubin ramp delay to HBM4 supply issues now resolved [5]. WCCFtech reports Rubin and Rubin Ultra face broader design and spec issues, with AMD's MI500 positioned as a competitor for late 2027 [2]. Separately, Anthropic runs substantial Claude Code inference on AWS Trainium and trains Claude models on Google TPUs, backed by a $100 billion-plus, ten-year commitment to AWS technologies [6][8].

NVIDIA faces simultaneous pressure on two fronts: internal product execution difficulties that forced a significant Rubin Ultra redesign, and growing frontier-lab adoption of alternative silicon from AWS and Google. How quickly NVIDIA executes on the revised Rubin roadmap — and whether AMD, Trainium, and TPUs convert product gaps into durable share — will determine whether NVIDIA's datacenter revenue leadership remains intact through 2027.

NVIDIA announced the 4-die Rubin Ultra at GTC 2026 in March, then cancelled the original design roughly three months later due to manufacturing execution concerns [1]. The replacement product keeps the 'Rubin Ultra' name but is approximately half the size and delivers roughly half the real-world performance of the announced configuration [1]. WCCFtech reports that both the Rubin and Rubin Ultra platforms face broader design and specification issues, and identifies AMD's MI500 as a competitor positioned for the second half of 2027 [2]. Tom's Hardware separately reported that Rubin CPX accelerators were removed from NVIDIA's roadmap [3], and NVIDIA also cancelled some planned rack configurations, with SemiAnalysis reporting material downstream implications for HBM memory demand [4].

Despite the Rubin Ultra product setback, SemiAnalysis projects a strong NVIDIA second half overall. In a June 30, 2026 post, SemiAnalysis estimated NVIDIA datacenter compute revenue would run 20% above sell-side consensus for 2H FY2027, explaining that the Rubin ramp had been delayed by HBM4 supply issues that are now resolved and that front-end wafer supply has been built up sufficiently to support the ramp [5]. This is a more nuanced position than straight-line pessimism: SemiAnalysis sees the Rubin Ultra product as diminished but views overall NVIDIA revenue trajectory as well above what analysts expect — consistent with a market where demand exceeds available supply even for a smaller product.

The competitive landscape on the alternatives side centers on Anthropic's departure from NVIDIA-centric infrastructure. SemiAnalysis reports that Anthropic runs a substantial share of Claude Code inference on AWS Trainium and trains Claude models on Google TPUs [6], a shift it characterized as something that would have been unimaginable a year ago. Reddit discussion confirms Claude Opus was trained on AWS Trainium2 [7], consistent with Anthropic's April 2026 announcement committing over $100 billion over ten years to AWS technologies including Trainium generations 2 through 4, with nearly 1 GW of Trainium2 and Trainium3 capacity expected online by end of 2026 [8]. Amazon deepened its financial stake simultaneously, committing up to $20 billion in additional investment on top of a prior $8 billion [8].

The Trainium narrative remains contested. An analysis titled 'Amazon Trainium Is A Disaster; Strategy Reset Needed' argues the platform has not delivered on its promise [9], directly contradicting the bullish framing from both SemiAnalysis and Anthropic's public commitments. Neither view has been validated with published performance benchmarks. SemiAnalysis's dual position — critical of specific NVIDIA product execution while projecting above-consensus NVIDIA revenue — illustrates that product-level setbacks and business-level outcomes are not the same thing, and that even a diminished Rubin Ultra may satisfy enough of the demand backlog to sustain strong revenue growth.