Nvidia's Moat Remains Strong
A breakdown of Nvidia’s latest earnings, why the business remains dominant, and the key risks ahead.
The day Nvidia doesn’t provide deeply positive revenue growth is the day that the AI trade will be over. But that day is not today and doesn’t seem like it will be any time soon! I will be picking up where I last left off with Nvidia. Also, to avoid any confusion, I’d like to remind you that Nvidia runs their fiscal year from February to January, so the latest reported earnings are for fiscal year 2026 Q1, covering February, March, and April 2025.
Let’s start off with the good news and save the bad news for later on.
Nvidia reported Q1 revenue of $44bn, up 12% QoQ and 69% YoY. Data Centre revenue remains the key engine, reaching $39bn, up 23% from the previous quarter and 73% YoY. It now accounts for nearly 90% of total revenue. Gaming also made a notable comeback, generating $3.8bn, up 48% QoQ and 18% YoY, supported by strong uptake of the new GeForce RTX 4070 and 4060 cards. While the headline figures are once again staggering, there are several aspects that I want to explore. The first is the ongoing shift from training to inference. Inference is the stage where AI models are no longer learning from data but instead applying what they have already learned to generate real-time outputs. This includes tasks like answering user queries, generating images, recommending content or translating language. Jensen Huang explained that the industry is now entering a new phase of AI deployment, where large language models and foundation models have already been trained, and hyperscalers are moving into production-level inference. Microsoft, Meta, and others deployed tens of thousands of Nvidia GPUs last quarter to support real-time workloads. Nvidia is now shipping close to 1,000 Blackwell racks per week, with each rack containing 72 GPUs. That amounts to over 70,000 Blackwell units deployed weekly, indicating not just demand but urgency.
Just as striking is the scale of future deployments. Management noted that the number of AI factory projects has doubled compared to this time last year, and each project is now significantly denser in terms of GPU allocation. These are no longer isolated or speculative deployments. They are large-scale infrastructure programmes built to support real-time, production-grade AI across search, recommendation engines, advertising, autonomous systems and digital twins. Nvidia has already begun sampling its next-generation GB300 systems, which are designed as a direct successor to GB200. Built on the same Blackwell architecture, GB300 offers higher inference throughput and improved memory efficiency, allowing customers to upgrade seamlessly without needing to redesign their infrastructure. Production is expected to begin later this quarter.
Networking was another bright spot. Revenue rose 64% QoQ to reach $5bn. Spectrum-X, Nvidia’s Ethernet solution designed for AI workloads, is now on pace to exceed $8bn in annualised revenue. NVLink switches alone brought in over $1bn during the quarter. These figures reflect more than hardware demand. They represent Nvidia’s expanding moat as it integrates deeper into the data centre stack. Cisco’s integration of Spectrum-X into its enterprise offerings underlines this trend. Speaking of expanding moats, Nvidia is coming for quantum computing too. And this could disrupt both IonQ and Rigetti Computing because Nvidia is positioning itself at the orchestration layer, providing the hardware and software stack that connects quantum processors with classical AI systems. While IonQ and Rigetti are focused on building the qubit hardware itself, Nvidia’s CUDA-Q platform and InfiniBand networking give it a central role in how hybrid quantum-classical workloads are actually run. If researchers and developers build around Nvidia’s tools, it risks marginalising smaller quantum hardware players who rely on others for integration. The reason I say this is because Nvidia recently announced that its GPUs and networking are powering ABCI-Q, the world’s largest research supercomputer dedicated to quantum computing. Built by Japan’s National Institute of Advanced Industrial Science and Technology, the system includes 2,020 H100 GPUs connected using Nvidia’s Quantum-2 InfiniBand. It also runs on CUDA-Q, Nvidia’s open-source platform for managing hybrid quantum-classical workloads. The system is designed to support quantum research using different processor types, including superconducting, photonic and neutral atom qubits. While the aim is research, the scale and integration highlight Nvidia’s intent to be a core enabler of future quantum development.
Another key point worth noting is sovereign demand. Nvidia highlighted AI infrastructure programmes underway in Japan, Canada, the UAE, Saudi Arabia, and several European countries. These government-backed AI factories reinforce the idea that AI infrastructure is now viewed as a national strategic asset. Unlike typical hyperscaler deployments, these initiatives are often tied to public sector objectives such as national LLMs, public health modelling, and advanced research capabilities. The scale and intent behind them suggest they are not one-off installations, but long-term infrastructure investments. Nvidia appears to be the default provider, offering not just the compute layer but also the networking, software, and system integration needed to get these programmes running quickly. In effect, it’s becoming the foundation for sovereign AI ambitions.
However, whilst sovereign demand is growing, operations in China are becoming a real bottleneck, and considering China was previously a key market for Nvidia, this is somewhat worrying. The company recorded a $2.2bn inventory writedown and a $2.3bn charge for supply purchase commitments, both tied primarily to its H20 chips. This $4.5bn hit reflects weakening demand and mounting regulatory pressure. The H20 was designed to comply with prior export controls, but recent tightening by the US government has further restricted access, making even these lower-spec chips difficult to ship at scale. Chinese tech giants including Alibaba, Tencent and Baidu have started testing domestic alternatives to Nvidia’s GPUs. Huawei’s Ascend chips are now being deployed, particularly for inference tasks, where demand is rising quickly. However, switching from Nvidia’s CUDA software stack to Huawei’s CANN framework comes with substantial engineering friction. Porting existing training code can take months, often requiring direct support from Huawei engineers. Some firms are adopting a hybrid approach, keeping training workloads on existing Nvidia chips while shifting inference to local processors. The broader risk is that Nvidia’s positioning in China is becoming less secure. State-backed demand is increasingly directed toward national champions, and commercial players are being pushed to follow. Nvidia has no new China-compliant chips in the market, and it remains unclear whether the next iteration, potentially due in July, will be competitive, especially without high-bandwidth memory or NVLink. The result this quarter was a GAAP gross margin of 60.5%, down sharply from 76% the prior quarter. Excluding the charges, non-GAAP margin would have been 71.3%, roughly in line with recent trends, however, the damage was done.
Price Action
I must say what an impressive bounce back after the massive drop lower throughout most of the year so far. Price pushed slightly higher after earnings but dropped further to end the week. I think that there is a lot of uncertainty going into the summer in general across all markets as some tariff extensions expire. Even though Nvidia is doing well, an overall market scare will mean that, most probably, tech gets hit hard in terms of reactions. I think it’s quite realistic that Nvidia will make ATHs at some point this year, perhaps in the summer if good news comes out of the White House, but I wouldn’t rely on that. With the last two names that I’ve covered, I’ve mentioned that they are more or less trades rather than investments but with Nvidia I’m very much invested and am not concerned about trading it. So whether Nvidia trades down to $120 or below in the next few months, it doesn’t really concern me.
Outlook
Looking ahead, Nvidia guided to $28bn in revenue for Q2, with GAAP gross margin expected at 71.8% and non-GAAP at 72.0%. Management expects margins to trend back towards the mid-70% range by year-end, as the China-related writedown fades and Blackwell volume ramps. Crucially, demand visibility remains strong. AI factory projects are increasing in number and density, and hyperscalers are continuing to deploy at scale. GB200 systems are on track to enter production this quarter, with supply expected to materially improve in the second half. Nvidia is also preparing for GB300 to scale in 2025, giving customers a forward upgrade path without changing architecture. Sovereign demand is expected to grow as more countries move to build domestic AI infrastructure. While the regulatory overhang in China remains unresolved, there are no signs of broader weakness. Capex guidance from hyperscalers continues to climb, and Nvidia remains the bottleneck for AI compute. Unless that changes, growth will continue to compound.
Final Thoughts
I’m personally very bullish on Nvidia simply due to the moat they have, along with the sheer size and speed of revenue growth. I also really love their story, tech and management team. I think Jensen is great and highly respectable. The China story is concerning but I’m sure they will figure something out, and with sovereign demand becoming increasingly heavy, any regional softness could easily be offset. The shift from model training to real-world inference is playing directly into their strengths, and with the Blackwell ramp just getting started, I still see a lot of room to run.
As always, I hope you enjoyed my analysis of Nvidia’s earnings release. If you did, please feel free to share it with colleagues, friends, or anyone else who might find it interesting. I would appreciate any help in growing the page.
Disclaimer: This article is for informational purposes only and should not be considered financial or investment advice. The views expressed are my own and based on publicly available information, market trends, and personal analysis. I own Nvidia shares which may change at any time. Readers should conduct their own research and consult a financial professional before making any investment decisions.