Key Takeaways
- Nebius Group ($NBIS) has entered the market with aggressive pricing for NVIDIA’s B200 GPUs, reportedly undercutting established hyperscalers like AWS and GCP by more than 50%.
- This pricing strategy appears to be a deliberate move to capture market share, funded by a substantial cash position and enabled by a vertically integrated business model that may lower operational overheads.
- The company’s rapid revenue growth (385% YoY in Q1 2025) is currently coupled with significant operating losses, highlighting a classic high-growth strategy prioritising scale over short-term profitability.
- While hyperscalers offer a broad and deeply integrated ecosystem of services, specialised providers focused on raw compute cost could create a distinct segment in the AI infrastructure market, particularly for large-scale model training.
The arrival of NVIDIA’s Blackwell architecture has ignited a fierce battle for dominance in the AI compute market, and the initial pricing for B200 GPU access reveals a stark divergence in strategy. While major hyperscalers are setting premium rates, newcomer Nebius Group has entered the fray with a conspicuously low price point. This is not merely a price war; it is a fundamental challenge to the established cloud economic model, pitting the focused efficiency of a vertically integrated challenger against the sprawling ecosystems of incumbents.
The Great GPU Price Divide
The cost of renting high-performance GPUs is a critical variable in the AI development equation. For companies training or fine tuning large models, compute expenditure can be the single largest operational cost. Early indications of B200 hourly rates suggest a wide and strategically significant gap between providers.
Nebius Group’s reported pricing of $5.50 per hour is a clear outlier, aggressively undercutting not only the largest players but also other specialised GPU cloud providers. This raises important questions about sustainability and the value proposition being offered. Is this a temporary, loss-leading tactic to gain a foothold, or does it reflect a structural cost advantage?
| Cloud Provider | Reported B200 GPU Hourly Rate (USD) |
|---|---|
| Nebius Group ($NBIS) | $5.50 |
| CoreWeave | ~$8.60 |
| Google Cloud Platform (GCP) | $12.00+ |
| Amazon Web Services (AWS) | $14.00+ |
These figures, while subject to change as supply and demand dynamics evolve, illustrate a clear market segmentation. The hyperscalers appear to be positioning the B200 as a premium, high-margin offering integrated into their vast suite of services. Nebius, in contrast, is treating it as a commodity, using price as its primary weapon for market entry.
Nebius: A Strategy Fuelled by Capital
To understand Nebius’s ability to sustain such pricing, one must examine its financial position. The company’s recent performance paints a picture of explosive growth balanced against significant investment in that growth. While pursuing an ambitious expansion, its strategy is underwritten by a formidable balance sheet, giving it the operational runway to absorb short term losses in pursuit of long term market share.
The company reported a 385% year-on-year revenue increase in its first quarter for 2025, demonstrating strong market appetite for its services. However, this came at the cost of a negative EBITDA of $62.6 million. With a reported $2.5 billion in cash and no debt, Nebius is well capitalised to continue this aggressive strategy. The company has reaffirmed its ambitious guidance to reach between $750 million and $1 billion in Annual Recurring Revenue (ARR) by the end of its 2025 fiscal year, a target that necessitates rapid customer acquisition, for which low pricing is a powerful lever.
| Nebius Group ($NBIS) Financial Snapshot (Q1 2025) | |
|---|---|
| Revenue | $55.3 million (+385% YoY) |
| Annual Recurring Revenue (ARR) | $249 million (as of March 2025) |
| EBITDA | ($62.6) million |
| Earnings Per Share (EPS) | ($0.39) |
| Cash Position | $2.5 billion (with no debt) |
Implications for the AI Infrastructure Landscape
The strategic logic behind Nebius’s approach likely stems from its vertical integration. By controlling more of the infrastructure stack, from data centre operations to its cloud platform, the company can potentially achieve efficiencies that are unavailable to hyperscalers managing a far broader and more complex range of services. This focused model allows it to optimise for one thing: delivering raw compute power at the lowest possible cost.
This presents customers with a clear choice. For enterprises that rely on the deep integration between compute, storage, databases, and managed services, the premium charged by AWS or GCP is often justified by the reduction in operational complexity. However, for an AI company whose primary need is access to thousands of GPUs for a multi week training run, a specialised, low cost provider becomes an extremely compelling alternative.
The risk for customers is one of reliance on a less established player. The hyperscalers have built a reputation over many years for reliability, security, and support. Nebius must prove it can match this level of service, particularly as it scales. For investors, the proposition is equally distinct. It is a bet on a high-growth, high-burn model and the notion that the AI compute market is large enough to support both integrated, premium providers and focused, low cost specialists.
As a speculative hypothesis, we may be witnessing the beginning of a bifurcation in the cloud market. The hyperscalers will likely continue to own the general purpose enterprise workload, where the breadth of their service catalogue is a key advantage. Simultaneously, a new category of “AI foundries” could emerge, competing almost entirely on the price and availability of raw compute. In this scenario, the winners will be those who can manage their supply chains and operational costs with the utmost discipline, turning scale into a sustainable, albeit lower margin, competitive advantage.
References
Amazon Web Services. (2024). New Amazon EC2 Instances Powered by NVIDIA Blackwell GPUs for the Highest Performance in the AI Era. Retrieved from https://aws.amazon.com/blogs/aws/new-amazon-ec2-p6e-gb200-ultraservers-powered-by-nvidia-grace-blackwell-gpus-for-the-highest-ai-performance/
DataCrunch. (n.d.). Cloud GPU Pricing Comparison. Retrieved from https://datacrunch.io/blog/cloud-gpu-pricing-comparison
DatacenterDynamics. (2024). NVIDIA B200s now generally available via AWS. Retrieved from https://www.datacenterdynamics.com/en/news/nvidia-b200s-now-generally-available-via-aws/
mvcinvesting. (2024, July 14). [Post showing B200 GPU pricing comparison]. Retrieved from https://x.com/mvcinvesting/status/1924787362241958282
NASDAQ. (2024). Nebius Group (NBIS) Earnings Date and Reports. Retrieved from https://www.nasdaq.com/market-activity/stocks/nbis/earnings
Seeking Alpha. (2024). Nebius Group: Expect Another Move Up Post-Q2 Earnings. Retrieved from https://seekingalpha.com/article/4799852-nebius-group-expect-another-move-up-q2-earnings
