Here Is What Wall Street Analysts Are Forecasting About NVIDIA’s GB200, B200, And NVL Rack Server Shipments Ahead Of The Earnings Next Week

This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy.

Wall Street analysts remain split on the extent to which NVIDIA's slower production ramp-up of the GB200 GPU would end up affecting its upcoming earnings and the outlook for the year ahead. While Baird analyst Tristan Gerra is dismissive of all "lingering concerns around [the] GB200 mix for the year," his counterpart at HSBC is much more cautious, lowering his forecast for NVL rack server shipments for the year from 35,000 units to 30,000 units.

To wit, Baird analyst has leveraged his supply chain meetings to note that GB200 shipments are now finally gaining momentum after a "late and slow start." In what is a critical tidbit, Gerra notes that apart from "architectural novelties" that took time to optimize, the GB200 GPUs were also initially constrained by the available data center infrastructure (electricity/space). The analyst goes on to note:

"Investors should not assume that these initial delays will lead customers to skip to the next-generation products. We expect GB200 to represent the majority of NVIDIA's GB mix in 2H25 and into 2026. GB300 crossover is expected not until 2H26."

Meanwhile, as mentioned above, HSBC analyst Frank Lee is much more cautious on NVIDIA, going so far as to take the rare step of lowering his stock price target for the GPU manufacturer from $185 to $175, while maintaining an overall 'Buy' rating. Lee believes GB200's "supply chain issues" could remain an "overhang going into H1FY26." Bear in mind that NVIDIA's fiscal year concludes in January, hence its mismatch with the December-ending calendar year.

Coming back, Lee has now lowered his "FY26e NVL rack server assumption from 35k to 30k [units]," while noting that NVIDIA will be able to "offset the slower GB200 ramp with upside in B200 GPU shipments in H1 FY26e," enabling the company to still "book Blackwell revenues of cUSD5bn and cUSD10bn in 4QFY25 and 1QFY26, respectively, leading to total revenue of USD40bn/42.2bn in 4QFY25/1QFY26 vs consensus of USD38.2bn/USD42bn."

As for the upcoming earnings, Lee expects NVIDIA to show $40 billion in total sales for Q4'25 against a company-guided figure of $37.5 billion and Wall Street's consensus estimate of $38.2 billion. The HSBC analyst concludes his note with the following projection:

"We also expect 1QFY26/2QFY26 sales of USD42.2bn/USD55.4bn vs consensus of USD42bn/USD46.2bn, implying H1 FY26 is unlikely to disappoint the market despite ongoing supply side issues."

Elsewhere, KeyBanc analyst John Vinh concedes that "manufacturing constraints are limiting shipments of GB200 NVL server racks," but predicts that "this will be more than offset by the following: 1) given the lower initial manufacturing yields of GB200 NVL, we believe customers have been able to push out orders of GB200 and backfill with HGX-based B200 servers with x86 head nodes; 2) DeepSeek, as well as limited supply of Huawei’s Ascend AI ASIC, has created a surge in demand for H20 GPUs from China CSPs; 3) we believe NVDA’s customers, esp CSPs, are financing its inventory at EMS providers, so effectively sell-in shipments from NVDA to EMS are recognized as revenues."

Finally, we bring you Oppenheimer's investment note on NVIDIA, where analyst Rick Schafer projects that the demand for H200 GPUs remains stronger than expected and will constitute ~30 percent of the company's sales mix this year.

On the GB300 GPU, Schafer notes:

"We believe GB300 (expected late CY25) likely a single rack (NVL72) configuration, suggesting DC cooling kinks being ironed out."

The Oppenheimer analyst goes on to note that "DeepSeek [is] a positive for NVDA as CSPs rapidly integrate innovations into Western models." Schafer also thinks that the CapEx of hyperscalers will be up 40 percent this year to over $300 billion vs. the $220 billion figure for 2024.

And, in what aligns with our previous post on the topic, Schafer thinks AI ASICs will coexist with GPUs:

"CSP custom AI ASIC projects coexist with GPUs, in our view—ASICs primarily used for internal workloads while GPUs used for broader, dynamic, higher performance AI applications. NVDA remains best positioned in AI, benefiting from full-stack AI hardware/software."

Do you think NVIDIA will be able to provide another triple beat (revenue, EPS, guidance) when it reports its earnings next week? Let us know your thoughts in the comments section below.

Follow Wccftech on Google to get more of our news coverage in your feeds.