Nvidia stock watchers are worried about the company’s revenue growth rate starting to slow down. In fact, it already has, with Nvidia’s rate of revenue increase shrinking more than 50% between its fiscal first quarter of 2025 reported last April and its fiscal second quarter of 2025.
That sure sounds like a company in trouble. Although, when you consider that the fiscal Q1 revenue increase was 262% and the fiscal Q2 revenue increase was 122%, you may not feel so sorry for Nvidia and its downward sales growth. When the company reports fiscal third quarter 2025 earnings Wednesday, its revenue growth rate is expected by analysts to be in the high double digits, above 80% or around $33.1 billion (the mid-point figure of the consensus range). Nvidia was slightly more modest with its fiscal Q3 guidance of around $32.5 billion.
But as has been the case for the last few quarters, analysts like will be less focused on teh most recent growth rate than on Nvidia’s own statements about its future outlook, including the sales forecast for its incoming Blackwell GPUs, its ability to turn early AI hype into practical business progress, and its positioning for catching computing next juggernaut–quantum–by the tail.
Nvidia delivered something tantalizing on all three fronts. The latest Blackwell news came out of Microsoft Ignite, where the show host announced the first cloud private preview of its Blackwell-powered Microsft Azure ND GB200 V6 VM series, aimed at accelerating breakthroughs in generative AI. That virtual machines series “combines the Nvidia GB200 NVL 72 rack-scale design with state-of-the-art Quantum InfiniBand networking to connect tens of thousands of Blackwell GPUs to deliver AI supercomputing performance at scale,” a Microsoft Azure blog post noted.
This progress with a major customer is only going to prime Nvidia watcher’s expectation for even the briefest of comments Nvidia CEO Jensen Huang might offer during this week’s earnings call about the state of early Blackwell market demand. Insider Monkey reported this week that Raymond James boosted its stock price target for Nvidia from $140 to $170 with the expectation that the company could sell at least 100,000 Blackwell GPUs during its current fiscal fourth quarter of 2025.
Meanwhile, at the Supercomputing 2024 (SC24) event this week, Nvidia looked beyond AI chips to practical AI applications with launches of more of its NIMs (Nvidia Inference Microservices) to aid various sectors in delivering greater value from AI investments.
NIMs get this job done by enable AI model builders to “convert their models into high-performance, efficient runtimes as NIMs,” said Dion Harris, director of accelerated computing products at Nvidia, during a briefing. “The real value [of AI] lies in deploying models for inference, where they can generate insights and predictions in real time… [NIMs] deliver two to five times faster throughput than standard AI runtimes, offering faster time to science and the best total cost of ownership. By making NIMs available across various domains, we’re accelerating innovation and expanding AI’s impact across industries.”
During SC24, Nvidia announced new NIMs for the Earth-2 digital twin, a project Nvidia has been involved in for at least three years, to augment Earth-2’s AI capabilities. The first of these is CorrDi, which Harris described as a generative AI model supporting “kilometer-scale super-resolution” that can help in forecasting and seeing complex weather events over regions like Taiwan and the U.S. As for the second new NIM, Nvidia also announced the availability of the FourCastNet NIM, which enables two-week worldwide weather forecasts about 5,000 times faster than current numerical weather models.
Meanwhile, in the biopharma field, Nvidia said it is making its BioNeMo Framework available as an open-source repository on GitHub, which will broaden its availability and utility across the
healthcare industry, Harris said. The company also unveiled DiffDock 2.0 - a NIM that researchers can use to predict how drugs interact with target proteins. This second version is six times faster than the first version that was announced last year, largely due to improvements in Nvidia’s cuEquivariance library, which is used for acclerating essential mathematical operations for molecular predictions.
Turning its attention from drug discovery to digital chemistry, Nvidia also announced a group AI Lab for Chemistry and Material Innovation (ALCHEMI) NIMs to speed up the discovery of new chemical compounds.
Harris explained, “Scientists start by designing the properties they want, like strength, conductivity, low toxicity, or color, which narrows the number of materials to analyze. A generative model suggests thousands to millions of potential candidates with the desired
Properties, and then the ALCHEMI NIM sorts candidate compounds for stability by solving for their lowest energy states.” This helps scientists to work faster to identify the best compound candidates before their organizations invest in testing.
Finally, in the quantum realm (not the one Ant-Man traveled to) Nvidia announced it is working with Google Quantum AI to accelerate the design of its next-generation quantum computing devices using simulations powered by the Nvidia’s CUDA-Q platform.
Google Quantum AI is using the hybrid quantum-classical computing platform and Nvidia’s Eos supercomputer to simulate the physics of its quantum processors to help designers of quantum hardware overcome the noise and error problems that have limited the hardware’s potential. These simulations have traditionally been prohibitively computationally expensive to pursue, but with CUDA-Q, Google can employ 1,024 Nvidia H100 Tensor Core GPUs on Eos to perform one of the world’s largest and fastest dynamical simulation of quantum devices at a fraction of the cost, Nvidia stated.
“The development of commercially useful quantum computers is only possible if we can scale up quantum hardware while keeping noise in check,” said Guifre Vidal, research scientist from Google Quantum AI. “Using Nvidia accelerated computing, we’re exploring the noise implications of increasingly larger quantum chip designs.”