Nebius’ second-quarter earnings report pushed its stock sharply higher on Aug. 12, but the central takeaway from a Futurum Equities podcast was not revenue alone. Shay Boloor, chief market strategist at Futurum Equities, argued that the quarter offered evidence that AI compute scarcity is being translated into pricing power, faster payback, stronger margins, and better financing terms.
Nebius reported $582 million in second-quarter revenue, up 454% year over year. AI revenue rose more than 500% to about $575 million. On the day of the release, the stock climbed from its Aug. 11 close of $193.23 to an intraday high of $243.89, a gain of about 26%. Boloor said short covering likely amplified the move, with short interest close to 25%.
Boloor said on FE Earnings Edge that the report showed something more durable than a one-quarter beat. In his words, 「这份财报证明了一件事：AI 算力短缺正在转化为更好的价格、更快的回本、更高的毛利和更好的融资条款。」 He also disclosed that Nebius is one of the top 10 positions in his family portfolio and that he added heavily during a market pullback several months ago. His portfolio also holds Bloom Energy. The commentary was presented as a holder’s interpretation of the earnings report, not independent investment advice.
Margins, not just growth, drew the strongest reaction
Boloor’s main point was that the report mattered because Nebius began to show the economics behind AI infrastructure demand. Revenue came in slightly above market expectations of about $570 million, but he focused more closely on profitability.
Group adjusted EBITDA margin reached 41%, up from 32% in the prior quarter. The core AI cloud business posted a 50% adjusted EBITDA margin. Boloor said that for a company spending tens of billions of dollars on infrastructure, proving that those assets can generate this level of profitability is what matters most for shareholders.
That, in his view, helps explain why the stock rallied more than 20% after earnings. The move reflected a reassessment of the underlying business model, not just a reaction to top-line growth or per-share loss figures.
Urgent compute and standard contracts are now priced very differently
The earnings discussion also laid out a clearer pricing structure for Nebius capacity. Standard contracts, typically one to three years in length, generate $20 million to $25 million per megawatt per year. Capacity that must be delivered within three to six months is priced far higher, at $40 million to $50 million per megawatt, and sometimes above that range.
Boloor framed that spread as a function of timing. Companies training the next generation of large models are not only buying GPUs; they are buying access to working compute at the exact moment they need it. He said that getting GPUs six months later may no longer be useful if development cycles are moving at an exponential pace.
Nebius also held its first compute auction during the quarter. Boloor said the winning price came in 15% above the company’s previous historical high, which he described as a real price-discovery event rather than management commentary about demand.
Customer prepayments are reshaping the funding model
Financing was another major part of the discussion. One of the common short arguments against NeoCloud companies has been that rapid expansion requires repeated capital raises, leaving shareholders to absorb dilution.
Boloor said this earnings report outlined a different structure. About 70% of contracts signed in the second quarter included prepayments. Nebius expects to receive more than $9 billion in customer prepayments in 2026, enough to cover 50% to 60% of the related capital expenditures. On top of that, the company secured $775 million in asset-backed financing, borrowing against contracted infrastructure and customer cash flows.
He described the sequence this way: customers fund roughly half of a data center, banks lend against contracted revenue, and Nebius equity becomes the last dollar in the stack rather than the first. In his view, that shift materially improves capital efficiency.
Boloor also pointed to a possible asset-light version of the model. If outside investors build and own the data centers while Nebius provides the cloud platform, software, demand, and customer relationships, the company could still earn economics from infrastructure it does not fully fund itself.
5GW contracted power target and management’s 2027 comment
Management also raised its contracted power target for the end of 2026 to 5GW. According to the podcast summary, that target has been moving up by roughly 1GW nearly every quarter for close to a year. Boloor said power sits at the center of the current buildout race, so repeated increases in that target are meaningful on their own.
The more aggressive signal came from management’s statement that, at current contract economics, Nebius could sell out all of its planned 2027 capacity today but is choosing not to do so. The idea, as presented by Boloor, is to keep some capacity available for immediate demand, where pricing is better and returns are higher.
He added a note of caution on execution. Only about 800MW to 1GW is expected to be truly online by the end of 2026. That power still has to be paired with GPUs, networking, cooling, and software before it begins generating revenue. In his framing, demand is no longer the key question. Execution is.
Long-term differentiation still depends on software
Boloor repeatedly said GPU scarcity will not last forever, which means Nebius cannot rely indefinitely on being a landlord of scarce compute. He pointed instead to the company’s software layer, including Token Factory and the broader software stack, as the longer-term source of customer retention and platform value.
He said the 50% AI cloud EBITDA margin may be an early sign of operating leverage, suggesting Nebius is doing more than renting expensive hardware at thin spreads. The discussion also mentioned the company’s New Jersey project, which switched to Bloom Energy fuel cells. Boloor said fuel cells offer dependable onsite power without delaying the project schedule.
For him, details like permitting, power access, and construction timing are where the next phase of the story will be decided. Delays in those areas can push revenue recognition further out even if demand remains strong.
A holder’s case for the stock, with the core risk made explicit
Boloor ended the discussion by stating the investment case plainly. He is long the stock and believes the current evidence still points to demand growing faster than supply. He cited customers willing to pay $40 million to $50 million per megawatt for urgent capacity, an average payback period of about one year and 10 months on contracts signed in the second quarter, customer funding that covers more than half of related capital spending, a 15% premium in the first compute auction, and management’s claim that 2027 capacity could be sold out already.
He also made the risk clear: Nebius is deliberately betting that scarcity will persist. His view is that it will last for many years, longer than the market expects. The broader thesis, as he framed it, is that compute scarcity is no longer just driving demand. It is also creating pricing power, financing leverage, and the ability to decide when to monetize capacity.