MSX US Stock Daily Watch: Nebius Plans to Raise GPU Rental Prices by Up to 21%, Further Strengthening Supply-Side Bargaining Power in AI Compute
- Key Takeaway: Nebius plans to raise on-demand GPU instance prices for H100, H200, B200, and B300 by approximately 17%–21%, indicating that AI compute demand is still growing faster than the pace at which high-end GPUs, power, and data center capacity come online, with the market still in a supply-driven phase.
- Key Elements:
- The price adjustment takes effect on October 1, 2026, with H100 rising from $3.85/hour to $4.50/hour and B300 rising from $7.85/hour to $9.50/hour, an increase of approximately 21%.
- Nebius's customer demand visibility has extended from approximately 18 months to over 24 months, with some customers having already booked compute capacity for the first half of 2028.
- Blackwell capacity auction clearing prices are approximately 15% higher than previous peak prices and approximately 20% higher than regular sales quotes.
- Nebius's Q2 2026 group revenue was approximately $582 million, up 454% year-over-year, with cloud business ARR run rate at approximately $3 billion.
- The company signed a contract worth up to $27 billion with Meta, received a $2 billion investment from Nvidia, and customer prepayments can cover approximately 50%–60% of related project capital expenditures.
- 2026 capital expenditure guidance is $20 billion–$25 billion, far exceeding the revenue guidance of $3 billion–$3.4 billion, with capital intensity and customer concentration remaining primary risks.
[MSX Research Institute · Daily US Stock RWA Observation] is a signature daily report produced by MSX Maitong, a leading RWA trading platform. Leveraging our strong macro research capabilities, we help you capture the core pulse of global traditional US stocks, liquidity changes, and the RWA tokenization market, enabling you to position ahead in quality assets.
Today's Observation
A noteworthy shift is emerging in the AI compute market: even as cloud providers continue to expand data centers, rental prices for some high-end GPUs have not fallen due to increased supply — instead, they continue to rise.
According to a price adjustment notice Nebius sent to customers, starting October 1, 2026, the company plans to raise prices for multiple on-demand GPU instances by approximately 17% to 21%. Specifically, the H100 price per GPU hour is set to increase from $3.85 to $4.50, the H200 from $4.50 to $5.40, the B200 from $7.15 to $8.50, and the B300 from $7.85 to $9.50.
Even so, this round of price adjustments still sends an important signal: the pace of AI compute demand growth may still be outpacing the speed at which high-end GPUs, power, and data center capacity come online.
Unlike ordinary cloud computing, the core scarce resources in AI cloud services today are not just servers, but also GPU supply, power access, liquid cooling systems, networking equipment, and data centers capable of housing high-density racks. If any one of these links becomes a bottleneck, new compute capacity cannot be brought online quickly.
Nebius management recently stated at a Goldman Sachs technology conference that the company's visibility into customer demand has extended from about 18 months previously to more than 24 months, with some customers already booking compute capacity for the first half of 2028. The company previously conducted price testing for scarce Blackwell capacity, with transaction prices coming in about 15% higher than the previous highest price and about 20% higher than quotes in the regular sales pipeline.
This means the AI cloud market may not yet have entered price competition at this stage; instead, it is in a phase where "whoever can deliver compute faster has pricing power."
Data in a Minute
- Customer notices show that Nebius plans to raise prices for some on-demand GPU instances starting October 1, by approximately 17% to 21%;
- The H100 price is set to rise from $3.85 to $4.50 per GPU hour, an increase of about 16.9%;
- The H200 is set to rise from $4.50 to $5.40, an increase of 20%;
- The B200 is set to rise from $7.15 to $8.50, an increase of about 18.9%;
- The B300 is set to rise from $7.85 to $9.50, an increase of about 21%;
- Nebius's official website still shows pre-adjustment prices, and the new prices have not yet taken effect on the public pricing page;
- Management said customer demand visibility has reached more than 24 months, with some compute demand extending into the first half of 2028;
- Winning bids in Blackwell capacity auction tests were about 15% higher than the previous highest price and about 20% higher than regular sales quotes;
- Nebius's Q2 2026 group revenue was approximately $582 million, up 454% year-over-year;
10. Of that, Nebius AI Cloud revenue was approximately $575 million, up 514% year-over-year;
11. Q2 group adjusted EBITDA was approximately $236 million, with Nebius AI Cloud adjusted EBITDA margin at approximately 50%;
12. At the end of Q2, Nebius AI Cloud's annualized revenue run rate was approximately $3 billion;
13. The company maintained its guidance of $3 billion to $3.4 billion in revenue for 2026, $7 billion to $9 billion in year-end ARR, and $20 billion to $25 billion in capital expenditures.
MSX View
The significance of Nebius's proposed price increase is not just that it can charge about 20% more per GPU hour, but that it validates that supply and demand in the AI compute market still favors suppliers.
One of the market's main concerns about AI infrastructure companies in the past was that after a large number of data centers broke ground simultaneously, the GPU rental market would eventually become oversupplied. Once compute prices fell, the data centers these companies had built at high prices could face declining utilization, longer investment payback periods, and asset impairment pressure.
But the signal Nebius is currently sending is exactly the opposite. The company is not only able to raise prices for the new-generation B200 and B300, but even the H100 and H200, which have been on the market for several years, are also seeing price increases. This shows that demand is not concentrated only in the latest architecture. For model fine-tuning, inference, scientific computing, and some enterprise workloads, the previous generation of GPUs still offers sufficiently good performance and cost efficiency.
If the price adjustments can be executed smoothly and GPU utilization does not decline significantly, Nebius will gain two improvements at once: first, higher revenue per unit of compute; second, the economic life and residual value of its existing GPU assets may be higher than previously expected.
However, higher on-demand prices do not mean all of Nebius's revenue will grow by 20% in tandem. Large customers typically sign multi-year contracts, with prices, minimum purchase volumes, and service terms already determined at signing and not immediately repriced according to new public rates. This adjustment most directly affects on-demand usage and short-term compute contracts, rather than long-term capacity agreements already signed with customers such as Meta.
On the other hand, higher public prices may also include cost pass-through. The costs of Nvidia systems, networking equipment, power, and data center construction may also rise. Therefore, only when Nebius's increase in revenue per unit exceeds the increase in investment and operating costs per unit will the price hike truly translate into a higher return on capital.
Nebius's greatest advantage right now is that orders, demand, and financing capacity have already formed a positive cycle. The company signed a contract worth up to $27 billion with Meta, received a $2 billion investment from Nvidia, and signed four large deals in Q2 with an average size of more than $1 billion. Some customer prepayments can cover about 50% to 60% of the related projects' capital expenditures, reducing the pressure on Nebius to expand entirely relying on its own balance sheet.
But this is still an extremely capital-intensive business. Nebius expects capital expenditures in 2026 to reach $20 billion to $25 billion, far above its revenue guidance of $3 billion to $3.4 billion for the year. The company also completed approximately $5.75 billion in convertible bond financing in August. Even if compute prices continue to rise, data center construction progress, debt costs, potential equity dilution, and customer concentration remain risks investors cannot ignore.
Therefore, what needs to be watched next is not only whether the new prices officially take effect on October 1, but also three more important indicators:
First, whether GPU utilization remains high after the price increase. If customers are still willing to accept higher prices, it shows that the compute supply bottleneck still genuinely exists.
Second, whether short-term high-priced contracts can translate into recurring revenue. On-demand compute prices are high, but revenue visibility is usually lower than for multi-year contracts; once demand cools, prices may also fall quickly.
Third, how much revenue and cash flow can be generated per unit of capital invested. The core competitiveness of AI cloud companies is ultimately not how many GPUs they own, but whether they can run those GPUs at relatively high utilization and recover their investment before the equipment is superseded.
Nebius's proposed price increase this time shows that the AI infrastructure industry has not yet entered a price war. What is truly scarce right now is still high-end compute that can be delivered on time, run stably, and provide complete software services.
But how long pricing power can last will depend on whether demand growth can continue to outpace the expansion speed of GPUs, power, and data centers. For Nebius, the price increase proves that demand is strong enough; what needs to be proven in the next stage is whether this strong demand can be converted into sustainable free cash flow and shareholder returns.
About MSX Maitong
MSX is a leading RWA trading platform dedicated to providing safe, efficient, and transparent access to global financial markets. As one of the world's earliest on-chain US stock trading platforms, MSX has always been at the forefront of the industry, leading market transformation through continuous innovation.
The platform deeply integrates blockchain technology with a compliance framework, offering spot and derivatives trading for nearly 400 tokenized stocks and pre-IPO assets, perfectly bridging the gap between traditional finance and the digital asset industry.
Around its core mission of "letting quality assets flow freely," MSX has now built a diversified digital financial services system covering US stock spot and perpetual contracts, crypto-to-crypto trading, Pre-IPO, and the Maitong Research Institute, aiming to provide global investors with all-hours, high-performance access to quality assets.
Official Website: https://msx.com/
Chinese TG:https://t.me/MSXOfficial
Global TG:https://t.me/MSXGlobal
The global App Store / Google Play are now live simultaneously.
Risk Disclosure: Macroeconomic and US stock market conditions are highly volatile. The content of this article is provided solely for academic and research observation reference by the Maitong Research Institute and does not constitute any investment advice.
