KNVIDIA shares rose on Tuesday after independent semiconductor research firm SemiAnalysis forecast that the artificial intelligence chipmaker’s data centre computing revenue would significantly exceed Wall Street expectations in the second half of its 2027 fiscal year.
The stock gained 1.7 per cent by midday in New York, trading at $198.37, recovering some ground after declining more than 11 per cent over the past month.
SemiAnalysis said its supply chain research indicates NVIDIA’s data centre compute revenue could come in around 20 per cent above current consensus forecasts, citing the resolution of earlier supply constraints and an accelerating rollout of the company’s next-generation Vera Rubin platform.
In a post published on X on Tuesday, the research firm said it expects a “huge second-half ramp” for NVIDIA, adding that previous delays caused by HBM4 memory supply issues have now been resolved. It also noted that front-end wafer supply has been built up, creating favourable conditions for increased production.
Unlike traditional sell-side forecasts, SemiAnalysis said its projections are based on extensive supply chain intelligence spanning materials suppliers, chip fabrication, component manufacturers, server integrators, hyperscale cloud providers and frontier artificial intelligence laboratories.
The firm’s bullish outlook centres on the easing of HBM4 memory bottlenecks, which had previously delayed mass shipments of the Rubin platform. With supply constraints now largely removed, SemiAnalysis expects shipment volumes to accelerate sharply during the second half of the fiscal year.
That assessment aligns with a report published by Digitimes on 25 June, which said industry suppliers expect NVIDIA’s transition from its Blackwell architecture to the Rubin platform to conclude during the second quarter of 2026, paving the way for stronger demand from the third quarter onwards.
NVIDIA officially commenced full production of its Vera Rubin platform on 1 June, with production shipments expected to begin this autumn to eight cloud computing partners: Amazon Web Services (AWS), Google Cloud, Microsoft Azure, Oracle Cloud, CoreWeave, Lambda, Nebius and Nscale.
The Rubin platform incorporates HBM4 memory capable of delivering approximately 22 terabytes per second of bandwidth per graphics processing unit (GPU), roughly three times the bandwidth offered by the HBM3e memory used in NVIDIA’s Blackwell generation. The higher memory throughput is expected to improve performance efficiency for large-scale artificial intelligence workloads.
Despite its optimistic revenue outlook, SemiAnalysis also highlighted a potential concern regarding NVIDIA’s product roadmap. The research firm said the company’s original four-chip Rubin Ultra configuration, unveiled during GTC 2026, was reportedly cancelled around three months after its introduction and replaced with a smaller design approximately half its original physical size, with performance reduced proportionately.
NVIDIA has not publicly commented on the reported design revision, and its potential impact on future revenue remains uncertain.
The upbeat supply chain outlook comes ahead of NVIDIA’s second-quarter fiscal 2027 earnings announcement, tentatively scheduled for 26 August. Current market consensus forecasts earnings per share of $2.07 on revenue of approximately $91.7 billion, although SemiAnalysis believes those estimates may ultimately prove conservative if production ramps as expected.
