What are CoreWeave's Q1 2026 numbers?

CoreWeave reported Q1 2026 revenue of $2.08B, up from $982M in Q1 2025 — a 112% year-over-year increase. Revenue backlog hit $99.4B, representing committed future revenue from long-term GPU rental contracts with hyperscalers and AI companies. Active computing power topped 1 GW (1 gigawatt) — the equivalent of powering a small city, deployed entirely to AI compute workloads. This is CoreWeave's first full quarter as a public company after its March 2026 IPO, and it far exceeded the $1.0–1.3B revenue range most analysts were modeling.

What is CoreWeave's business model and why does the backlog matter so much?

CoreWeave rents NVIDIA H100/H200 GPU clusters to AI companies on multi-year contracts. Customers include OpenAI (which uses CoreWeave for a significant portion of its training compute), Microsoft, Anthropic, and dozens of AI startups. The $99.4B backlog is the most important number in the report: it represents signed, binding commitments from customers to pay for GPU time over the next several years. At $2.08B quarterly revenue, the backlog represents roughly 47 quarters (nearly 12 years) of revenue at current rates — though actual contract durations are 3-5 years. This means CoreWeave's revenue is essentially locked in for the foreseeable future regardless of whether AI demand softens.

What does 1 GW of active computing power mean?

1 gigawatt is a power milestone. For context: a large nuclear power plant produces ~1 GW. CoreWeave is consuming the equivalent of a nuclear plant's entire output just for AI compute. This is not a future target — it's current active deployment. Each NVIDIA H100 GPU draws about 700 watts; 1 GW means CoreWeave has roughly 1.4 million GPU-equivalents deployed. This level of compute density requires massive datacenter investment, specialized cooling, and grid-scale power agreements. CoreWeave has secured long-term power purchase agreements (PPAs) with utilities across the US and Europe — the physical infrastructure is already built and running.

How does CoreWeave fit into the AI infrastructure stack?

The AI infrastructure stack from bottom to top: (1) Chip design — ARM (architectures), (2) Chip manufacturing — TSMC (implied), (3) Chips — NVIDIA (GPUs), AMD (AI accelerators), (4) Cloud infrastructure — CoreWeave (GPU clusters), AWS/Azure/GCP (hyperscalers), (5) AI platforms — OpenAI, Anthropic, Google DeepMind, (6) AI applications — PLTR (government), AppLovin (advertising), etc. CoreWeave occupies the layer between raw chips and AI platforms. Its $99.4B backlog is effectively the forward bookings of the entire AI model training and inference market for the next 3-5 years — visible demand that AMD, NVIDIA, and TSMC will need to satisfy.

What is the AI earnings week scorecard and what does it mean?

Five consecutive AI-related earnings beats in one week: PLTR (EPS +18%, Rev +85% YoY), ON Semi (AI datacenter power inflection), AMD (Data Center +57%, Q2 guided $11.2B), AppLovin (Rev +59%, NI +109%, FCF $1.3B), ARM (EPS +3.4%, licensing intact), and CoreWeave (Rev +112%, backlog $99.4B). Every layer of the AI value chain — chip architecture (ARM), chip hardware (AMD), AI cloud compute (CoreWeave), AI government software (PLTR), AI advertising (APP) — is growing at 50-112% year-over-year. The consensus concern entering 2026 that AI capex would slow has been definitively refuted. The constraint is not demand — it's supply of chips, power, and datacenter capacity.

Earnings Recap: CoreWeave (CRWV) Q1 2026 — Revenue $2.08B (+112% YoY), Backlog $99.4B, Active Power Tops 1 GW — AI Cloud Infrastructure Demand Is Unbounded

CoreWeave Q1 2026 — The Numbers

Reported: Thursday, May 7, 2026 (AMC)

Headline Results

Metric	Q1 2026	Q1 2025	YoY Change
Revenue	$2.08B	$982M	+112%
Revenue Backlog	$99.4B	—	Historic
Active Power	>1 GW	<500 MW	>2x

The $99.4B Backlog — The Most Important Number

At $2.08B quarterly revenue, the $99.4B backlog represents:

~47 quarters of current revenue already contracted
3-5 year average contract duration with hyperscalers
$99.4B in committed payments from OpenAI, Microsoft, Anthropic, and others

This backlog is essentially guaranteed revenue. Even if AI demand completely stopped today (it won’t), CoreWeave would continue collecting rent on deployed GPUs through the contract term.

1 Gigawatt — The Power Milestone

Reference	Power
Nuclear power plant	~1 GW
CoreWeave AI compute	>1 GW
Large coal plant	~600 MW
Hoover Dam	~2 GW

CoreWeave is consuming nuclear-plant-scale power to run AI workloads. This is not a company — it is a power utility for artificial intelligence.

The Full AI Earnings Week Stack

Layer	Company	Beat	Key Number
Architecture	ARM	✅ +3.4% EPS	Licensing intact
Hardware	AMD	✅ DC +57%	Q2 guided $11.2B
Cloud Infra	CRWV	✅ +112% Rev	Backlog $99.4B
Gov AI	PLTR	✅ +18% EPS	US Gov +84%
AI Ads	APP	✅ +59% Rev	FCF $1.3B

Every layer. Every beat. The AI capex supercycle is real.

What Comes Next

CoreWeave’s growth is supply-constrained, not demand-constrained. The bottlenecks:

NVIDIA GPU production — H100/H200 allocations are rationed
Power — new datacenter sites require years of permitting and grid connection
Cooling — liquid cooling infrastructure for >1 GW facilities

NVIDIA’s upcoming Blackwell GPU generation (B100/B200) will unlock the next capacity tier. When available at scale (late 2026/2027), CoreWeave’s buildout will accelerate further.

— Ray | signals.themenonlab.com