Nebius Group (NBIS)

AI GPU neocloud (Nebius) — full-stack NVIDIA GPU cloud for AI labs; NVIDIA-backed, Meta-anchored

AI exposure 95% — derived from reported segment revenue, not asserted.

Where it sits

AI Platforms → Hyperscalers & Cloud Platforms · figures rebased to Q2 2026, reported 2026-08-12

Pure-play AI GPU cloud operator — builds and operates the platform layer AI is delivered through (NVIDIA GPU clusters for AI labs/hyperscalers); direct peer of CoreWeave. Non-core edtech/AV arms are beyond-map.

Reported segments

• AI cloud infrastructure — 100% of revenue

Nebius Group ($582.3M Q2 2026, +454% YoY; ARR $3.0B exiting Q2 against a $7-9B exit-2026 target; FY2026 guide $3.0-3.4B) — full-stack AI cloud spun out of Yandex's international business. Nebius AI is 98% of revenue ($575M in Q2). Primary B6 Hyperscalers & Cloud Platforms (neocloud). Contracted book of more than $40B from investment-grade counterparties: Meta five-year agreement providing $12B of dedicated Vera Rubin capacity with total value up to ~$27B, and Microsoft five-year contract at $17.4B plus up to $2B of additional services. Four new contracts averaging over $1B each signed in Q2…

How the 95% is built

• AI cloud infrastructure — 100% of revenue × 95% AI-driven (estimate). Nebius's core is a purpose-built AI GPU cloud sold to AI labs/hyperscalers; a small adjacent non-cloud sliver (edtech/AV) is being separated out

Investment read

Moat. Full-stack vertical integration inherited from Yandex, NVIDIA supply priority, and — the part that actually separates it — the cheapest capital in the neocloud sector. Customer prepayments are expected to fund more than $9B of the 2026 build, net debt is roughly $2B, and a $775M senior secured facility at SOFR+250 backed by deployed GPUs and one investment-grade contract was significantly oversubscribed. In a business selling undifferentiated compute, cost of capital is the durable variable.

Bottleneck / pricing power. Allocation power while GPUs are scarce, now with a published price: four new Q2 contracts averaging over $1B each at disclosed yields of $20-25M per megawatt, an AI-cloud unit adjusted EBITDA margin of 50% and group margin stepping from 32% to 41% in one quarter. No peer has put a yield-per-megawatt number on the table. Pricing still erodes as industry capacity catches up.

Role in the AI stack. A full-stack AI GPU-neocloud layer — the balance-sheet-led challenger to CoreWeave.

Connected companies

• NVIDIA (NVDA) — NBIS is supplied by: GPUs/accelerators — Nebius builds its AI cloud on NVIDIA (Blackwell/Vera Rubin); NVIDIA took a strategic stake for supply priority
• Meta Platforms (META) — NBIS supplies: AI cloud capacity — five-year agreement providing $12B of dedicated Vera Rubin capacity across multiple locations, total contract value up to ~$27B (Mar 2026)
• Microsoft (MSFT) — NBIS supplies: AI GPU cloud capacity — five-year contract worth $17.4B with up to $2B of additional services (~$19.4B total), including a dedicated site at Vineland, New Jersey; latest planned tranche delivered as…
• CoreWeave (CRWV) — NBIS competes with: AI GPU neocloud rivals — both rent NVIDIA GPU clusters to AI labs/hyperscalers
• Bloom Energy (BE) — NBIS is supplied by: On-site fuel-cell power for Nebius AI data centers — $1.7B project investment (IDF lead developer, Oaktree minority equity; Morgan Stanley tax equity, MUFG debt) deploying Bloom Energy Servers as…
• ServiceNow (NOW) — NBIS supplies: AI cloud capacity — ServiceNow reported among the enterprise AI adopters buying Nebius compute in 2026; press-reported, not company-confirmed

Others in Hyperscalers & Cloud Platforms

• Microsoft (MSFT)
• Alphabet (GOOGL)
• Amazon (AMZN)
• Meta Platforms (META)
• CoreWeave (CRWV)

All 174 companies · What moved this quarter · How the map is built