[MSX Institute · Daily Observation on U.S. Stocks and RWA] is a flagship daily report by MSX, the leading RWA trading platform. Leveraging our robust macro research capabilities, we capture the key trends in global traditional U.S. stocks, liquidity shifts, and the RWA tokenization market to help you strategically position yourself in high-quality assets.
Today's Observations
A notable shift is emerging in the AI computing power market: while cloud providers continue to expand their data centers, rental prices for certain high-end GPUs are not declining with increased supply—they continue to rise.
According to the price adjustment notice sent by Nebius to its customers, starting October 1, 2026, the company plans to increase prices for multiple on-demand GPU instances by approximately 17% to 21%. The proposed hourly rates per GPU are as follows: H100 will rise from $3.85 to $4.50, H200 from $4.50 to $5.40, B200 from $7.15 to $8.50, and B300 from $7.85 to $9.50.
Even so, this round of price adjustments still sends a significant signal: the pace of growth in AI computing demand may still outstrip the rate at which high-end GPUs, power, and data center capacity are coming online.
Unlike traditional cloud computing, the current core scarce resources for AI cloud services are not just servers, but also GPU availability, power access, liquid cooling systems, networking equipment, and data centers capable of supporting high-density racks. If any one of these components encounters a bottleneck, new computing power cannot be rapidly deployed.
Nebius management recently stated at the Goldman Sachs Technology Conference that the company’s visibility into customer demand has extended from approximately 18 months to over 24 months, with some customers already reserving compute capacity for the first half of 2028. The company previously conducted price testing for scarce Blackwell capacity, achieving transaction prices approximately 15% higher than previous peak prices and about 20% higher than standard sales pipeline quotes.
This means that, at this stage, the AI cloud market may not yet be engaged in price competition, but rather in a phase where whoever can deliver computing power faster holds the pricing power.
Data per minute
- Customer notice indicates that Nebius plans to increase prices for certain on-demand GPU instances by approximately 17% to 21%, effective October 1st;
- The price of the H100 is planned to increase from $3.85 per GPU hour to $4.50, an increase of approximately 16.9%;
- H200 is proposed to be increased from $4.50 to $5.40, a 20% increase;
- B200 is planned to be increased from $7.15 to $8.50, an approximate increase of 18.9%;
- B300 is proposed to be increased from $7.85 to $9.50, an approximate increase of 21%;
- The Nebius website still displays the pre-price adjustment rate; the new price has not yet taken effect on the public pricing page.
- Management stated that customer demand visibility has exceeded 24 months, with some compute demand extending into the first half of 2028;
- The winning price for the Blackwell capacity auction test was approximately 15% higher than the previous highest price and about 20% higher than the standard sale quote;
- Nebius reported group revenue of approximately $582 million for the second quarter of 2026, a 454% year-over-year increase;
10. Among these, Nebius AI Cloud generated approximately $575 million in revenue, a 514% year-over-year increase;
11. The group's adjusted EBITDA for the second quarter was approximately $236 million, with Nebius AI Cloud's adjusted EBITDA margin at approximately 50%;
12. At the end of the second quarter, the annualized revenue run rate for Nebius AI Cloud was approximately $3 billion;
13. The company maintains 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 lies not just in the ability to charge approximately 20% more per GPU hour, but in confirming that supply and demand in the AI computing market still favor suppliers.
One of the main concerns in the past market regarding AI infrastructure companies was that the simultaneous launch of numerous data centers could eventually lead to an oversupply in the GPU leasing market. If compute prices fall, these companies’ data centers, built under the assumption of higher prices, may face lower utilization rates, extended payback periods, and asset impairment pressures.
But Nebius is currently sending the opposite signal: the company is not only raising prices for its new B200 and B300 models, but also increasing prices for the H100 and H200, which have been on the market for several years. This indicates that demand is not concentrated solely on the latest architecture. For model fine-tuning, inference, scientific computing, and certain enterprise workloads, the previous-generation GPUs still offer sufficient performance and cost efficiency.
If the price adjustment is implemented successfully and GPU utilization does not decline significantly, Nebius will achieve two improvements: higher revenue per unit of computing power, and the economic lifespan and residual value of existing GPU assets may be higher than previously anticipated.
However, the on-demand price increase does not mean that all of Nebius’s revenue will rise by 20% in tandem. Large customers typically sign multi-year contracts with prices, minimum purchase volumes, and service terms fixed at the time of signing, which cannot be immediately repriced according to the new public rates. This adjustment most directly impacts on-demand and short-term compute contracts, rather than long-term capacity agreements already signed by clients like Meta.
On the other hand, the price increase may also involve cost pass-through. Costs for NVIDIA systems, networking equipment, power, and data center construction may also rise. Therefore, the price increase will only translate into higher capital returns if Nebius’s unit revenue growth exceeds its unit investment and operating cost growth.
Nebius's current biggest advantage is that orders, demand, and financing capabilities have formed a positive feedback loop. The company has signed contracts with Meta worth up to $27 billion, received a $2 billion investment from NVIDIA, and secured four large deals in the second quarter, each averaging over $1 billion. Prepayments from some customers can cover approximately 50% to 60% of the capital expenditures for related projects, reducing the pressure on Nebius to expand solely through its own balance sheet.
But it remains an extremely capital-intensive business. Nebius expects capital expenditures of $20 billion to $25 billion in 2026, far exceeding its revenue guidance of $3 billion to $3.4 billion for the same year. The company also completed a $5.75 billion convertible bond financing in August. Even if compute prices continue to rise, risks such as data center construction timelines, debt costs, potential equity dilution, and customer concentration remain significant concerns for investors.
Therefore, what needs to be observed next is not only whether the new price on October 1st officially takes effect, but also three more important indicators:
First, whether GPU utilization remains high after the price increase. If customers are still willing to accept the higher price, it indicates that the computing power supply bottleneck still genuinely exists.
Second, can short-term high-price contracts be converted into sustainable revenue? On-demand computing power is expensive, but its revenue visibility is typically lower than that of multi-year contracts; once demand cools, prices may also drop rapidly.
Third, how much revenue and cash flow can be generated per unit of capital invested. The core competition among AI cloud companies is not about how many GPUs they own, but whether they can operate these GPUs at high utilization rates and recoup their investments before equipment upgrades.
Nebius's proposed price increase indicates that the AI infrastructure industry has not yet entered a price war. What remains truly scarce today is high-end computing power that can deliver on time, operate stably, and provide comprehensive software services.
But how long pricing power can be maintained will depend on whether demand growth can continue to outpace the expansion of GPUs, electricity, and data centers. For Nebius, the price increase demonstrates sufficient demand; the next step is to prove that this strong demand can be converted into sustainable free cash flow and shareholder returns.
About MSX
MSX is a leading RWA trading platform dedicated to providing secure, efficient, and transparent access to global financial markets. As one of the earliest on-chain platforms for trading U.S. stocks, MSX has consistently remained at the forefront of the industry, driving market transformation through continuous innovation.
The platform deeply integrates blockchain technology with compliance frameworks, offering spot and derivatives trading for nearly 400 tokenized stocks and Pre-IPO assets, perfectly bridging the gap between traditional finance and the digital assets industry.
Guided by the core mission of enabling free circulation of high-quality assets, MSX has built a diversified digital financial services ecosystem encompassing U.S. stock spot and perpetual contracts, crypto-to-crypto trading, Pre-IPO investments, and MacTong Research Institute, aiming to provide global investors with round-the-clock, high-performance access to premium assets.
Official website:https://msx.com/
Chinese TG:https://t.me/MSXOfficial
Global TG：https://t.me/MSXGlobal
The app is now available simultaneously on the global App Store and Google Play.
Risk disclaimer: Macroeconomic conditions and U.S. stock market volatility are significant; the content of this article is for academic and research observation purposes only by MaiTong Research Institute and does not constitute any investment advice.