Meta Platforms has signed a sweeping new agreement with AI cloud provider Nebius Group worth up to $27 billion over five years, dramatically scaling up a partnership first forged in November 2025. Shares of Nebius surged approximately 14% in premarket trading Monday, while Meta's stock climbed nearly 3%, as Wall Street welcomed one of the largest single AI infrastructure contracts in history.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">Meta Doubles Down on Third-Party AI Compute</h2>
Meta Platforms sealed a five-year AI infrastructure agreement with Nasdaq-listed Nebius Group on March 16, 2026, committing up to $27 billion for advanced GPU cloud capacity — a massive expansion of the initial $3 billion deal the two companies signed just four months ago. The contract represents one of Meta's largest single vendor commitments on record and underscores the social media giant's urgency to secure dedicated AI compute outside its own internal data center buildout.
A spokesperson for Meta confirmed the transaction, describing it as part of the company's broader strategy to accelerate the development of next-generation large language models and AI-driven products across its platforms. The new agreement supersedes the November 2025 contract, which Nebius had disclosed was capped below demand — meaning Meta was prepared to commit to even greater capacity than Nebius could supply at the time.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">Nebius: From $3 Billion to $27 Billion in Four Months</h2>
The scale of the upgrade is striking. When Nebius first announced the Meta relationship in November 2025, the $3 billion figure was framed as a floor constrained by available GPU cluster capacity, with Nebius pledging to deploy the required infrastructure within three months. Monday's announcement confirms that once capacity was secured, Meta moved decisively to lock in far greater volume.
Combined with Nebius's previously announced $17.4–$19.4 billion Microsoft contract signed in September 2025, the company now holds more than $46 billion in committed contract backlog from just two hyperscaler-class customers — a figure that dwarfs its full-year 2025 revenue of $529.8 million and anchors one of the most aggressive revenue ramp trajectories in enterprise technology history.
The company's 2026 revenue guidance of $3.0–$3.4 billion — itself a nearly six-fold increase over 2025 — now appears conservative in light of the expanded Meta commitment. Nebius has also projected annualized recurring revenue of $7–$9 billion by year-end 2026 and adjusted EBITDA margins of approximately 40%.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">Why Meta Chose Nebius Over Hyperscalers</h2>
The decision by Meta — a company that operates some of the world's most sophisticated private data center infrastructure — to outsource AI training compute at this scale to a third-party provider reflects a structural reality now acknowledged across the industry: hyperscaler GPU capacity is simply insufficient to meet demand. AWS, Microsoft Azure, and Google Cloud have each faced chronic wait times for H100 and H200 clusters throughout 2025 and into 2026.
Nebius, built on the engineering foundation of Yandex's former international operations and led by co-founder Arkady Volozh, offers H100 GPUs at $2.00–$3.06 per GPU-hour, undercutting the industry standard, delivered via a full-stack AI cloud purpose-built for large-scale model training. The platform runs InfiniBand networking at 3.2 Tb/s, enabling thousands of GPUs to operate as a single coherent training fabric — a capability critical for the frontier model development Meta is pursuing.
The deal also benefits from Nebius's rapidly expanding global data center footprint. The company has secured 16 sites worldwide, including a flagship 1.2-gigawatt campus in Independence, Missouri, a 300-megawatt facility in Vineland, New Jersey (where Microsoft's deployment is anchored), and additional owned sites in Oklahoma, Alabama, and Minnesota. Total contracted power capacity has been raised to 3 gigawatts, with 800 megawatts to 1 gigawatt targeted to be connected and operational by year-end 2026.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">NVIDIA Partnership Amplifies Nebius's Strategic Position</h2>
The Meta deal arrives just five days after Nebius announced a landmark expanded partnership with NVIDIA, including a fresh $2 billion strategic investment — bringing NVIDIA's total stake to $2.7 billion. Under that agreement, Nebius and NVIDIA are targeting deployment of more than 5 gigawatts of NVIDIA systems across the Nebius platform by 2030, with early access to the NVIDIA Rubin platform, Vera CPUs, and next-generation Blackwell GB200 NVL72 clusters.
NVIDIA CEO Jensen Huang described Nebius as "an AI cloud designed for the agentic era, fully integrated from silicon to software" — the most explicit public endorsement NVIDIA has extended to any neocloud operator. The backing ensures Nebius priority GPU allocation as competition for NVIDIA hardware among cloud providers intensifies globally.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">Market Reaction and Broader AI Capex Context</h2>
Markets responded decisively. Nebius shares jumped approximately 14% in premarket trading on Monday, adding to a year-to-date run that has already seen the stock climb sharply following the NVIDIA investment announcement on March 11. Meta's stock rose nearly 3% premarket, reflecting confidence that the deal supports the company's AI development roadmap without disproportionate capital burden given the multi-year payment structure.
The agreement fits within a broader hyperscaler capital expenditure surge. Major technology companies — including Meta, Microsoft, Alphabet, and Amazon — are collectively expected to invest approximately $650 billion in data centers and AI hardware in 2026 alone, per industry estimates. Goldman Sachs projects total hyperscaler AI capex at $527 billion for the year, with cloud revenues on a trajectory toward $2 trillion by 2030.
Nebius's capacity to absorb demand from the world's most resource-intensive AI developers positions it alongside CoreWeave and IREN as one of the defining neocloud infrastructure companies of the current AI build-out cycle — a segment McKinsey has identified as growing on the back of structural GPU supply gaps that hyperscalers have been unable to fully close.
<h2 class="text-2xl font-bold text-landing-text-primary mt-10 mb-6">A Company Transformed</h2>
Nebius's trajectory has been extraordinary by any measure. Emerging from the $5.4 billion divestiture of Yandex's international assets in July 2024 with a single data center in Finland and $2.5 billion in cash, the company has in under two years secured more than $46 billion in contracted infrastructure revenue, attracted $2.7 billion from NVIDIA, and posted 479% revenue growth in fiscal year 2025. In Q4 2025, its core AI infrastructure business reached a 24% adjusted EBITDA margin — the first quarter of positive adjusted EBITDA in the company's independent existence.
With the Meta agreement now dramatically expanding its backlog, Nebius enters the second quarter of 2026 as a critical pillar of the global AI compute supply chain, its fate intertwined with the ambitions of two of the most powerful technology platforms on earth.
---
Mentioned tickers: META, NBIS, MSFT, GOOGL, AMZN, NVDA, CRWV

# Meta Commits Up to $27 Billion to Nebius in Landmark AI Infrastructure Pact

**Meta Platforms has signed a sweeping new agreement with AI cloud provider Nebius Group worth up to $27 billion over five years, dramatically scaling up a partnership first forged in November 2025. Shares of Nebius surged approximately 14% in premarket trading Monday, while Meta's stock climbed nearly 3%, as Wall Street welcomed one of the largest single **AI infrastructure contracts** in history.**

## Meta Doubles Down on Third-Party AI Compute

Meta Platforms sealed a **five-year AI infrastructure agreement** with Nasdaq-listed Nebius Group on March 16, 2026, committing up to $27 billion for **advanced GPU cloud capacity** — a massive expansion of the initial $3 billion deal the two companies signed just four months ago. The contract represents one of Meta's largest single vendor commitments on record and underscores the social media giant's urgency to secure **dedicated AI compute** outside its own internal data center buildout.

A spokesperson for Meta confirmed the transaction, describing it as part of the company's broader strategy to accelerate the development of next-generation **large language models** and **AI-driven products** across its platforms. The new agreement supersedes the November 2025 contract, which Nebius had disclosed was capped below demand — meaning Meta was prepared to commit to even greater capacity than Nebius could supply at the time.

## Nebius: From $3 Billion to $27 Billion in Four Months

The scale of the upgrade is striking. When Nebius first announced the Meta relationship in November 2025, the $3 billion figure was framed as a floor constrained by available **GPU cluster capacity**, with Nebius pledging to deploy the required infrastructure within three months. Monday's announcement confirms that once capacity was secured, Meta moved decisively to lock in far greater volume.

Combined with Nebius's previously announced **$17.4–$19.4 billion Microsoft contract** signed in September 2025, the company now holds more than **$46 billion in committed contract backlog** from just two hyperscaler-class customers — a figure that dwarfs its full-year 2025 revenue of $529.8 million and anchors one of the most aggressive revenue ramp trajectories in enterprise technology history.

The company's **2026 revenue guidance of $3.0–$3.4 billion** — itself a nearly six-fold increase over 2025 — now appears conservative in light of the expanded Meta commitment. Nebius has also projected **annualized recurring revenue** of $7–$9 billion by year-end 2026 and **adjusted EBITDA margins** of approximately 40%.

## Why Meta Chose Nebius Over Hyperscalers

The decision by Meta — a company that operates some of the world's most sophisticated private data center infrastructure — to outsource **AI training compute** at this scale to a third-party provider reflects a structural reality now acknowledged across the industry: **hyperscaler GPU capacity** is simply insufficient to meet demand. AWS, Microsoft Azure, and Google Cloud have each faced chronic wait times for H100 and H200 clusters throughout 2025 and into 2026.

Nebius, built on the engineering foundation of Yandex's former international operations and led by co-founder **Arkady Volozh**, offers **H100 GPUs at $2.00–$3.06 per GPU-hour**, undercutting the industry standard, delivered via a **full-stack AI cloud** purpose-built for large-scale model training. The platform runs **InfiniBand networking at 3.2 Tb/s**, enabling thousands of GPUs to operate as a single coherent training fabric — a capability critical for the **frontier model development** Meta is pursuing.

The deal also benefits from Nebius's rapidly expanding **global data center footprint**. The company has secured 16 sites worldwide, including a flagship 1.2-gigawatt campus in Independence, Missouri, a 300-megawatt facility in Vineland, New Jersey (where Microsoft's deployment is anchored), and additional owned sites in Oklahoma, Alabama, and Minnesota. Total **contracted power capacity** has been raised to 3 gigawatts, with 800 megawatts to 1 gigawatt targeted to be connected and operational by year-end 2026.

## NVIDIA Partnership Amplifies Nebius's Strategic Position

The Meta deal arrives just five days after Nebius announced a landmark **expanded partnership with NVIDIA**, including a fresh $2 billion strategic investment — bringing NVIDIA's total stake to $2.7 billion. Under that agreement, Nebius and NVIDIA are targeting deployment of more than **5 gigawatts of NVIDIA systems** across the Nebius platform by 2030, with early access to the **NVIDIA Rubin platform**, Vera CPUs, and next-generation Blackwell GB200 NVL72 clusters.

NVIDIA CEO **Jensen Huang** described Nebius as "an AI cloud designed for the agentic era, fully integrated from silicon to software" — the most explicit public endorsement NVIDIA has extended to any neocloud operator. The backing ensures Nebius priority **GPU allocation** as competition for NVIDIA hardware among cloud providers intensifies globally.

## Market Reaction and Broader AI Capex Context

Markets responded decisively. Nebius shares jumped approximately **14% in premarket trading** on Monday, adding to a year-to-date run that has already seen the stock climb sharply following the NVIDIA investment announcement on March 11. Meta's stock rose nearly **3% premarket**, reflecting confidence that the deal supports the company's **AI development roadmap** without disproportionate capital burden given the multi-year payment structure.

The agreement fits within a broader hyperscaler **capital expenditure surge**. Major technology companies — including Meta, Microsoft, Alphabet, and Amazon — are collectively expected to invest approximately **$650 billion** in data centers and AI hardware in 2026 alone, per industry estimates. Goldman Sachs projects total hyperscaler AI capex at $527 billion for the year, with cloud revenues on a trajectory toward $2 trillion by 2030.

Nebius's capacity to absorb demand from the world's most resource-intensive AI developers positions it alongside **CoreWeave** and **IREN** as one of the defining **neocloud infrastructure** companies of the current AI build-out cycle — a segment McKinsey has identified as growing on the back of structural GPU supply gaps that hyperscalers have been unable to fully close.

## A Company Transformed

Nebius's trajectory has been extraordinary by any measure. Emerging from the $5.4 billion divestiture of Yandex's international assets in July 2024 with a single data center in Finland and $2.5 billion in cash, the company has in under two years secured more than **$46 billion in contracted infrastructure revenue**, attracted $2.7 billion from NVIDIA, and posted **479% revenue growth** in fiscal year 2025. In Q4 2025, its core AI infrastructure business reached a **24% adjusted EBITDA margin** — the first quarter of positive adjusted EBITDA in the company's independent existence.

With the Meta agreement now dramatically expanding its backlog, Nebius enters the second quarter of 2026 as a critical pillar of the global **AI compute supply chain**, its fate intertwined with the ambitions of two of the most powerful technology platforms on earth.

---

**Mentioned tickers:** META, NBIS, MSFT, GOOGL, AMZN, NVDA, CRWV

Meta Platforms has signed a sweeping new agreement with AI cloud provider Nebius Group worth up to $27 billion over five years, dramatically scaling up a partnership first forged in November 2025. Sha...

Meta Commits Up to $27 Billion to Nebius in Landmark AI Infrastructure Pact

More News

CoreWeave (CRWV) Stock Overview: Moderate Buy Amid Deep Discount and Growing Risks

Adobe Posts Record Q1 Earnings as CEO Narayen Announces Departure After 18 Years

Atlassian Axes 1,600 Jobs in AI-Driven Restructuring Overhaul

Nvidia Commits $2 Billion to AI Neocloud Nebius in Landmark Hyperscale Partnership

Oracle Surges 12% as Q3 Earnings Shatter Expectations, FY27 Revenue Target Lifted to $90B