Nvidia releases Vera Rubin benchmarks, forecasts 8.8 million GPU shipments, and announces a CFO share sale.

Nvidia's ability to physically ship its next generation of AI accelerators has become the central question for investors, overshadowing even the company's ambitious revenue targets. With the Vera Rubin NVL72 platform now posting official benchmark figures and manufacturing partners racing to expand advanced packaging capacity, the debate has shifted from whether demand exists to whether supply chains can keep pace.

**Performance gains set the stage**

Nvidia's first official performance data for Rubin, submitted under the MLPerf Inference v6.1 benchmark, showed throughput up to 3.7 times higher than the preceding GB300 NVL72 system on the Qwen3-VL model. On DeepSeek-R1, measured performance reached up to 2.5 times the prior level.

The gains have already attracted concrete orders. India's Yotta Data Services plans to deploy 80,000 units from the Vera Rubin line as part of a $12 billion expansion program, while Southeast Asian technology group Sea Limited is using the platform to run e-commerce and financial applications.

**Production capacity becomes the constraint**

Meeting global demand depends heavily on advanced packaging capacity. Analysts at UBS raised their production forecast for Nvidia's AI graphics processors in 2027 to 8.8 million units, including roughly 200,000 Rubin CPX series chips. The revision was driven by expected growth in worldwide CoWoS capacity at chipmakers such as TSMC.

Management has guided for 70 percent revenue growth in fiscal 2028, and according to CEO Jensen Huang, customer demand exceeds that figure by a wide margin — leaving real supply chain delivery capacity as the primary constraint on growth.

**Analyst initiation and price momentum**

Piper Sandler began coverage of the stock with a buy rating and a $300 price target. Analyst David O'Connor cited an expected annual revenue growth rate of 47 percent through fiscal 2030 and a projected earnings per share of $30.

In German trading, the shares changed hands at EUR 199.54, just 1.5 percent below the 52-week high of EUR 202.50. Pre-market quotes put the stock at EUR 200.25, a gap of 1.1 percent from that peak.

**Insider transactions**

On the insider front, 40,747 shares were withheld from CFO Colette Kress on September 16 to cover obligations from vesting restricted stock units. The following day, Kress sold a further 34,918 shares for roughly $7.65 million. The disposal was executed under a Rule 10b5-1 trading plan established on June 16, a mechanism designed to automate executive sales independently of current corporate news. Such moves are routine at technology companies when equity-based compensation comes due and the associated tax burden is settled directly in shares.

**Strategic expansions**

On September 14, Nvidia announced an extension of its open-source CUDA-Q platform, with the CUDA-Q Logical orchestration layer aimed at fault-tolerant quantum computing applications. The company is also preparing to close a major acquisition: Nvidia agreed to buy Hugging Face for $12.93 billion. The purchase price comprises $11.9 billion for existing shareholders plus a further $1 billion in stock as an employee retention package. Management expects the transaction to close in the first half of 2027.

**Valuation under pressure**

Market participants are watching the fundamental valuation with growing scrutiny. The price-to-earnings ratio based on expected profits has fallen sharply to less than 17 — a halving from the 2025 level and far below the multiple of 50 seen in 2023. Investors are increasingly asking whether the company can sustain its exceptional growth rates. Attention now turns to whether the new strategic steps and platforms will be enough to justify elevated expectations for the business, with delivery timelines for the new chip generation serving as the key test for revenue and margin targets.