Why Jensen Huang's 1-Gigawatt $50 Billion Valuation Benchmark Lifted Nebius 7.5%, September 2026
NVIDIA CEO Jensen Huang's assertion that a single 1-gigawatt AI data center is worth $50 to $60 billion gave investors a formal framework for pricing Nebius's gigawatt-scale buildout, sending shares up 7.5% to $226.39 on September 8, 2026.
- 최근 회계연도 설비투자
- $4.07B (FY2025)
- 전년 대비
- 403.4% · $808M → $4.07B
- 투자 / 매출
- 767% (FY2025, $530M)
- 최고 기록 기간
- $4.07B · 2025-12-31
When Jensen Huang told investors that a single 1-gigawatt AI data center carries an intrinsic worth of $50 billion to $60 billion, he did more than quantify the infrastructure cycle. He provided a publicly attributed valuation anchor for a class of companies whose equity story is precisely the bet that owning compute at that scale is worth building toward. Among neoclouds, Nebius responded most visibly: shares closed at $226.39 on September 8, up 7.5% on the day, extending a year-to-date run that already stood at 218% through late August. The move illustrates how tightly neocloud equity valuations have become coupled to NVIDIA's own narrative about where AI infrastructure value accrues — and how much weight the market assigns to the CEO of the dominant GPU supplier when he chooses to speak in dollar terms.
The operating fundamentals beneath the stock's 2026 run are substantial. Nebius reported Q2 2026 total revenue of $582 million, a 454% increase year-over-year, with adjusted EBITDA turning positive for the first time at $236 million, a 41% margin that represents a sharp reversal from a $21 million EBITDA loss in the year-earlier period. To put the magnitude in context: a single quarter of revenue now exceeds the company's entire fiscal year 2025 revenue of $530 million, which itself was the baseline against which Slicast's SEC XBRL compilation measured capex intensity. The company set full-year 2026 guidance of $7 billion to $9 billion in revenue — a range that presupposes reliable conversion of the contracted backlog. Two anchor customers define that visibility: Microsoft committed to a deal valued at up to approximately $19 billion for dedicated GPU capacity supporting Azure, while Meta reportedly signed an additional contract; multiple sources place the total contracted backlog above $40 billion, though that aggregate has not been independently audited at the line level.
To build the infrastructure required to service those commitments, Nebius executed what appears to be the largest capital raise in neocloud history to date. On August 22 the company announced a $4.5 billion convertible senior note offering; by pricing the size had been upsized to $5.0 billion, and the transaction ultimately closed with gross proceeds of approximately $5.75 billion. The announcement triggered an immediate 14% sell-off as markets weighed dilution risk, before the stock recovered and surpassed pre-announcement levels — a sequence that itself illustrates the volatility baked into the asset class. The proceeds are funding a build-out across two continents: in Europe, Nebius has announced expansion into Estonia and a second campus in Mäntsälä, Finland, alongside a colocation deployment with Vantage Data Centers in Newport, South Wales, the first in the UK government's designated South Wales AI Growth Zone. In the United States, sites in New Jersey and Butler Township, Pennsylvania have cleared key zoning hurdles, while a proposed campus in Independence, Missouri generated enough civic friction — over $6 billion in approved tax incentives — that voters recalled the council member who had backed the deal. On the technology side, Nebius became an early production deployer of NVIDIA's LPX ultra-low-latency inference rack platform, reporting a fourfold improvement in AI agent response latency.
Slicast's compilation of Nebius's SEC XBRL filings quantifies the capital commitment with a precision that no single earnings report captures. Capital expenditure reached $4.07 billion in fiscal year 2025, a 403.4% increase from $808 million in fiscal year 2024, and equivalent to 767% of that same year's $530 million in revenue — the highest capex intensity among the seven neoclouds Slicast tracks. In absolute spending, Nebius ranks second, behind CoreWeave, which reported a $104 billion contracted backlog and Q2 2026 revenue of $2.58 billion, up 112% year-over-year. The intensity ratio will compress sharply as Nebius's revenue scales toward its full-year guidance — that compression is the core of the investment thesis — but it also defines the vulnerability: if demand softens or financing conditions tighten, the fixed obligations carried by $5.75 billion in convertible notes will weigh materially. Rate sensitivity was on display in late August, when a broader neocloud sell-off triggered by interest rate concerns sent Nebius shares down roughly 5% alongside CoreWeave at roughly 7%.
The competitive picture carries genuine uncertainties that the stock's 2026 run does not fully reflect. Reports in early September indicated Meta is accelerating its own captive compute buildout; separately, SpaceX's potential entry as a third-party cloud provider was cited as an additional competitive shock. Neither development has visibly impaired Nebius's contract momentum — the company signed four deals each exceeding $1 billion in the second quarter — but the structural dynamic is clear: the largest AI spending companies are also the most capable of internalizing the infrastructure they currently outsource. At the community level, opposition to data center siting is becoming a material operational variable, as the Missouri recall illustrates; Butler Township's approval was explicitly conditioned on new ordinance guardrails, a precedent other municipalities will observe. Jensen Huang's per-gigawatt valuation framework is a useful investor heuristic, but it presupposes that demand and margin hold — an assumption, not a guarantee.
Nebius enters the second half of 2026 with a contracted backlog that dwarfs its trailing revenue, $5.75 billion in capital that can sustain multiyear construction, and an early-generation edge in low-latency inference infrastructure. Three signals are worth tracking: whether Q3 and Q4 revenue conversion validates the $7 to $9 billion full-year guidance range, given that the pace of recognition will depend on how quickly new capacity is commissioned and accepted; how effectively the company manages community and regulatory friction at its US sites, where local politics have demonstrably moved the stock; and whether the early-adopter position on NVIDIA's LPX racks translates into durable pricing power against CoreWeave, Lambda, and on-premise alternatives as the hardware generation matures. The 1-gigawatt benchmark Jensen Huang articulated is a destination Nebius is building toward at an intensity no peer has matched. The open question is what margin it captures when it arrives.