Nebius Auctions AI Compute Capacity at 20% Premium as Demand Outpaces Supply

Nebius Auctions Nvidia Blackwell Capacity at Premium Prices

Nebius Group N.V. has discovered just how severe the AI infrastructure shortage has become. When traditional price increases failed to balance supply and demand for AI compute, the AI infrastructure provider turned to auctions for access to Nvidia Blackwell-powered capacity. The experiment revealed striking pricing power: customers voluntarily paid 15-20% above Nebius' previous highest prices to secure guaranteed access to scarce GPU resources2

2

. Speaking at the Goldman Sachs Communacopia + Technology Conference 2026, founder and CEO Arkady Volozh explained that manual price increases alone couldn't address the imbalance. "We tried to do it manually first to raise the prices. It didn't help. The demand is still there," he said2

2

. The auction mechanism provided price discovery while giving customers certainty over capacity allocation, with winning bidders expressing satisfaction at securing guaranteed resources through this approach2

2

.

AI Infrastructure Market Shows Persistent Supply-Demand Gap

The AI infrastructure market remains deeply constrained despite ongoing capacity expansion. Volozh described demand as growing "times a year, if not tens of times a year," while physical infrastructure cannot expand at that pace3

3

. Nebius now has visibility into customer demand extending into Q1 and Q2 2028, with some clients seeking tens of thousands of vGPUs and GPUs3

3

. Management noted the company could sell all of its 2027 capacity today if it chose to pre-sell it. Global AI infrastructure capacity is expanding by approximately 5-10 gigawatts annually, potentially rising to 15-20 gigawatts by 2028. However, even planned additions from xAI and Meta, estimated at 6-8 gigawatts, would add only 20-30% incremental supply and wouldn't change the market "radically"3

3

. This persistent gap between demand for AI compute and available supply continues to support pricing power across the AI infrastructure market.

Customer Prepayments Fund Aggressive $20-25 Billion Capex Program

Nebius is pursuing massive AI infrastructure investment to meet surging demand, spending nearly $5.7 billion in capital expenditures during Q2 2026 on GPUs, GPU-related hardware and data center footprint expansion1

1

. The company reaffirmed its 2026 capex guidance of $20-25 billion, underscoring the investment intensity required to support growth ambitions1

1

. Customer prepayments are playing a crucial role in funding this buildout. Management noted that nearly 70% of second-quarter deals included upfront prepayments, with customer prepayments expected to provide more than $9 billion of funding in 20261

1

. The prepayment terms secured in Q2 cover approximately 50-60% of associated capex, reducing dependence on debt and equity financing1

1

. Nebius also generated $2.3 billion in operating cash flow during the quarter with $8 billion in cash and cash equivalents as of June 30, 20261

1

.

Contracted Power Capacity Expansion and Financing Strategy

Nebius has raised its year-end contracted power target to 5 gigawatts and plans to build more than 1 gigawatt of new capacity in 2027, while continuing to expect 800 megawatts to 1 gigawatt of connected power in 20261

1

. The company is deploying multiple funding sources beyond customer prepayments. Under its at-the-market equity program, Nebius issued 12.7 million Class A shares, generating gross proceeds of approximately $2.8 billion in Q2 2026, with 12.3 million shares remaining available under the program as of June 301

1

. The AI infrastructure provider is also supplementing these sources with asset-backed financing. Management highlighted its $775 million debt facility, priced at SOFR plus 250 basis points, backed by deployed GPU infrastructure and contracted cash flows from an investment-grade customer1

1

. The company's roughly $40 billion committed backlog is seen as supporting further asset-backed financing opportunities1

1

.

Major Customer Wins and Broadening Revenue Mix

Nebius secured four major customer wins, each averaging approximately $1 billion in contract value, won on technical performance and reliability rather than price3

3

. One customer described the proof-of-concept as "the best technical experience they have ever had with any supplier in the industry"3

3

. All four customers are already discussing next-cluster expansions and future needs. Chief Revenue Officer Marc Boroditsky outlined three main contract buckets: short-term deals of 3-6 months for scale training runs at premium pricing, medium-term deals of 1-3 years representing the "lion's share" of opportunity, and long-term deals of 5 years or more3

3

. The customer base is expanding beyond AI-native companies into enterprise accounts, supported by a new partnership with Palantir3

3

. Nebius' revenue surged 507% over the last twelve months to $1.36 billion, reflecting explosive demand for AI infrastructure capacity3

3

.

Execution Risks Amid Intensifying Competition

While customer prepayments, operating cash flow and contracted backlog provide meaningful support for Nebius' elevated investment program, execution risks remain significant. The payoff hinges on continued AI demand and the ability to scale profitably amid increasing competition from CoreWeave and tech giants like Microsoft and Amazon1

1

. Microsoft added 31 data centers across five continents during the quarter and another 1 gigawatt of capacity, while expecting to double overall capacity within two years1

1

. CoreWeave is projecting 2026 capital expenditures between $35-39 billion, up from $31-35 billion, with Q2 2026 capex reaching $9.4 billion and construction in progress increasing to $11.9 billion1

1

. Nebius is building an asset-light growth model with infrastructure partners to ease limits on physical buildout and capital3

3

. The company's shares trade at a 12-month forward price-to-sales multiple of 5.95X, above the Internet Software Services industry's 4.13X1

1

.