Nvidia delivered impressive Q4 figures on 26 February, exceeding its forecasts and setting new records in core segments, especially in the data centre. With revenue of $39.3bn in the fourth quarter and full-year revenue of $130.5bn in fiscal year 2025, Nvidia has entered a phase of profound technical innovation and market expansion. In addition to the outstanding figures in the core business, Nvidia’s focus is on the rapid introduction of the Blackwell chip and potential benefits to glean from the groundbreaking DeepSeek artificial intelligence (AI), laying the foundations for future growth.
But this development costs money, and profits have exceeded expectations by only a small margin, as they last did two years ago. People are used to big hits from Nvidia, so its latest quarterly figures, while impressive, are therefore unlikely to be a major driver for the overall market.
Detailed fourth quarter results
Nvidia’s Q4 revenue of $39.3bn represents sequential growth of 12% and an impressive increase of 78% year-on-year, and its $130.5bn full-year revenue represents a 114% increase year-on-year. These figures underscore not only the success of the current products, but also the effectiveness of the company's strategic direction.
The data centre segment had annual sales of $115.2bn and a quarterly contribution of $35.6bn, driven by the massive Blackwell rollout, allowing Nvidia to meet strong demand for modern data centres. Other business areas such as gaming, professional visualisation and automotive — the latter with record sales of $570m — complement the positive overall picture. At the same time, gross margins are temporarily in the low 70s due to initial production costs and the complexity of the new systems. As production optimisation progresses, these margins could rise into the mid-70s.
Technological milestones: Blackwell and DeepSeek
The Blackwell architecture
A key driver of Nvidia’s current success is the introduction of the Blackwell architecture, which had the fastest product launch in company history and generated $11bn in revenue in Q4. According to Nvidia, Blackwell’s speed and scalability allow it to equip large data centres with hundreds of thousands of GPUs — a decisive competitive advantage in an era with increasingly complex and computationally intensive AI models.
Blackwell is characterised by several technical innovations, including a revolutionary Transformer engine that offers up to 25 times higher “token throughput” (essentially, the number of processing units the chip can process per second) while reducing costs by 20 times compared to the previous generation. An NVLink Domain system, which delivers 14 times the performance of conventional PCIe Gen 5 connections, ensures the necessary responsiveness and efficiency for massive data exchanges in modern data centres.
DeepSeek – driving force for reasoning AI
Another decisive innovation boost comes from the DeepSeek R1 technology. DeepSeek has attracted a lot of attention worldwide as it is a world-class open-source AI model that sets a new standard in the field of reasoning AI and used 2,048 Nvidia H800 graphics cards to train its V3 model.
“Reasoning AI” is advanced AI that analyses and solves complex problems using multi-stage thought processes and logical conclusions. The technology makes it possible to significantly accelerate conventional computing tasks, which require up to 100 times more computing power in test-time inference. Companies and developers worldwide are already using DeepSeek or related chain-of-thought methods to increase the performance of their AI models.
This innovation is a technological advancement and a clear indicator of how Nvidia is addressing the challenges of the next generation of AI. Not only are traditional pre-training and post-training tasks mastered, but complex inference processes are also efficiently supported. This positions Nvidia as a key player in a market where quickly scaling and adapting to new computing requirements is crucial to success.
Nvidia CEO Jensen Huang emphasised the existence of three central scaling laws:
- Pre-training: Despite continued growth rates, pre-training remains a critical factor in the foundation of many modern AI models.
- Post-training: Subsequent refinement, especially using reinforcement learning, often requires far more computing power than the original pre-training.
- Inference: So-called “test time scaling” or reasoning AI requires up to 100 times more computing power for individual queries than conventional one-shot inference methods.
This differentiated view underlines that Nvidia is investing in optimising the entire AI pipeline rather than relying on traditional training. The ability to efficiently serve all three areas — pre-training, post-processing and inference — is a decisive competitive advantage, reflected in the rapid introduction of Blackwell products.
System architecture and production optimisation
Nvidia is making great efforts to ensure a smooth transition between different GPU generations. While the current Blackwell generation is already being successfully delivered on a large scale, the groundwork for the company’s next chip, Blackwell Ultra, was also emphasised. This new generation is expected to impress with improved network integration, optimised storage solutions and new processors without changing the basic system architecture.
Nvidia places great emphasis on a modular and flexible architecture. This allows existing systems to be upgraded gradually and new products to be introduced simultaneously without causing friction in the supply chain. This strategy minimises production downtime and ensures continuous product availability to meet high customer demand.
Demand forecasting and market stability
Nvidia expects robust demand in both the short and medium term. Huang pointed to concrete signs: forecasts and orders from top partners underline the long-term trend that most future data centre capex will be invested in AI infrastructure. In addition, investments in data centres and innovative start-ups show that demand is manifesting in new application areas such as agentic and physical AI as well as more traditional areas.
Agentic AI refers to systems that make decisions and carry out actions independently as acting agents, while physical AI is integrated into real, material systems — such as robots or autonomous vehicles — to interact directly with the physical environment.
These statements indicate that Nvidia has a broadly diversified customer base, ranging from hyperscale cloud providers to specialised industrial companies. This diversification makes the company less vulnerable to market fluctuations in individual regions or segments. Despite increased revenue growth in the US, the share of revenue from China has remained stable due to existing export controls.
Margin dynamics and cost efficiency
Statements from CFO Colette Kress provided important information on Nvidia’s gross margins. In the current phase of the Blackwell product launch, margins are temporarily in the low 70s due to additional production costs and the complexity of the new systems. However, Kress emphasised that an upward trend is to be expected with further production optimisation and scaling of production — in the medium term, margins should be back in the mid-70s.
Nvidia is already implementing strategic measures to optimise costs and is prepared to accept short-term margin challenges to benefit from higher economies of scale in the long term. The ability to harmonise complex production processes is an essential part of the company's recipe for success.
Technological integration and ecosystem
Nvidia emphasises that its architecture is designed to cover almost all use cases — from traditional computing tasks to highly complex AI models. This versatility is also a key reason why Nvidia remains indispensable, even in heterogeneous systems where custom chips and processing units coexist.
The answers made it clear that the company's strength lies not only in its hardware, but above all in an extensive software ecosystem that greatly simplifies the implementation and optimisation of AI solutions. This ecosystem offers developers and companies a robust platform to develop new applications and optimise existing processes — a decisive competitive advantage in a market that is increasingly dominated by integrated solutions.
Outlook and strategic perspectives
With forecast sales of around $43bn in the coming quarter, Nvidia is optimistic about the future. Its expectations are based not only on the continued strong demand in the data centre and gaming sectors, but also on customers' willingness to invest in state-of-the-art AI infrastructures. Nvidia expects gross margins to pick up again despite short-term pressure through ongoing optimisation and the expansion of product configurations.
The company’s strategic orientation includes several essential elements:
- Scaling AI infrastructure: The continuous expansion of computing capacity, especially through the ongoing Blackwell product launch, ensures Nvidia a significant competitive advantage in the long term.
- Product portfolio diversification: With the planned introduction of Blackwell Ultra and the subsequent product generation Vera Rubin, Nvidia is preparing for future market requirements without jeopardising the stability of the existing system.
- Strengthening its software ecosystem: By integrating DeepSeek and other innovation-driven solutions, Nvidia is positioning itself as a key technology provider in an increasingly software-based market environment.
- Global market dynamics: Despite geopolitical uncertainties, particularly regarding US trade policy and export controls, Nvidia remains confident as the global need for AI solutions and data-driven infrastructures grows exponentially.
Challenges and regulatory influences
A critical point repeatedly discussed at yesterday evening's annual press conference concerns the effects of US trade policy. While Nvidia strictly adheres to all applicable export controls and regulations, the future design of customs duties and tariffs remains uncertain. Kress pointed out that although not all details of future trade policy are currently foreseeable, the company will react flexibly to possible changes. This adaptability is crucial to ensure stable production and supply chains despite potential external influences.
Nvidia is also working closely with its international partners to minimise potential disruptions in the global supply chain. The strategic distribution of production sites — with over 350 manufacturing sites worldwide — enables the company to better manage risks and respond flexibly to regulatory changes.
What’s next for Nvidia?
Nvidia is in a transformational phase, driven by its technological innovations and strategically aligned production and market development. The rapid scaling of the Blackwell architecture and the introduction of DeepSeek as a groundbreaking AI model are setting new standards in AI inference and data processing.
The production of the new Blackwell chips entails enormous costs, which puts considerable pressure on margins. While other companies would proudly point to margins of 71%, these figures are more likely to raise concern among Nvidia shareholders. However, Nvidia does not see DeepSeek as a threat. Rather, the democratisation of AI development is expected to fuel demand for graphics chips as more and more companies invest in corresponding applications.
Nvidia's success is not only based on impressive sales figures, but also on a clear vision for the future of AI. With a broadly diversified product portfolio, a robust global manufacturing network and a consistently innovation-driven strategy, the company is setting new standards and consolidating its leading position in one of the world’s most dynamic and forward-looking markets.
CMC Markets er en ‘execution-only service’ leverandør. Dette materialet (uansett om det uttaler seg om meninger eller ikke) er kun til generell informasjon, og tar ikke hensyn til dine personlige forhold eller mål. Ingenting i dette materialet er (eller bør anses å være) økonomiske, investeringer eller andre råd som avhengighet bør plasseres på. Ingen mening gitt i materialet utgjør en anbefaling fra CMC Markets eller forfatteren om at en bestemt investering, sikkerhet, transaksjon eller investeringsstrategi. Denne informasjonen er ikke utarbeidet i samsvar med regelverket for investeringsanalyser. Selv om vi ikke uttrykkelig er forhindret fra å opptre før vi har gitt dette innholdet, prøver vi ikke å dra nytte av det før det blir formidlet.
