$CRWV at $81.75 is the screaming buy in AI infrastructure right now π₯π₯ It is also the leader in U.S. Sovereign AI πΊπΈ Don't lack the "situational awareness" to see it. Here's the bull thesis in a nutshellπ
Here's the number that should stop you cold: CoreWeave will exit 2027 at a $30B revenue run-rate, and more than 75% of it is already contracted. Today's entire market cap is ~$44.6B. You are paying roughly 1.5x market-cap for signed, take-or-pay contracts on six times its 2025 revenue, eighteen months out. Those are locked commitments.
And you're buying it on sale. A $99.4B contracted backlog against that ~$44.6B cap is ~0.45x market-cap-to-backlog. Revenue grew 112% YoY last quarter. The market handed you a big two-day drop on a Meta headline.
This is the fastest-scaling AI cloud in history. Revenue went from roughly $16M in 2022 to $1.92B in 2024 to $5.13B in 2025, and management guides to $12β13B this year on the way to a $30B run-rate exiting 2027. That is a company going from essentially nothing to a $30B revenue pace in about five years. It signed more than $40B of new bookings in a single quarter and pushed backlog to $99.4B. No infrastructure company has ever ramped contracted revenue at that dollar pace.
Power is the real bottleneck in AI, and CoreWeave has already locked up years of it. They hold ~3.5 GW of power under contract and are converting it to live capacity as fast as they can build, more than doubling active power to 1.7 GW by year-end 2026 and scaling toward 8 GW by 2030. Every gigawatt they energize is already spoken for, so growth here comes down to execution on power they've already secured. First self-build sites are coming online now to capture even more of the economics.
CoreWeave is first to market on GB200, GB300, and Vera Rubin. The only Platinum ClusterMAX provider, two rounds running. NVIDIA put $2B into the stock at $87.20 last quarter, so today you're buying it cheaper than NVIDIA did. Rosenblatt's John McPeake reiterated a $250 target and called this exact selloff a buying opportunity, and Roth called the neocloud selloff overdone.
The single-customer risk that shadowed the IPO is gone. Microsoft was ~62% of revenue in 2024. Today CoreWeave counts nine of the top ten AI model developers as customers, and management says no single one will exceed roughly 35% of contracted revenue. Look at the roster: OpenAI (~$22.4B), Meta (~$35B through 2032), Anthropic (multi-year Claude deal), plus Microsoft, Mistral, Cohere, and IBM. Then it expanded well beyond the AI labs. Jane Street signed a $6B compute deal and put $1B of equity in at $109 a share, pulling in quant finance, with financial-services backlog now near $10B. And NVIDIA backstopped $6.3B of capacity through 2032 on top of its equity stake. When nine of the ten companies building frontier AI all pay you rent, you own the toll road.
And the demand is the sticky kind. Inference is now over 50% of consumed compute, recurring production workloads that repeat every day. That's why pricing rose quarter-over-quarter across A100, H100, H200, and L40S. Rising prices in a supposedly commoditized market prove real pricing power. CoreWeave has it.
2027 is where the operating leverage shows. Management called Q1 the margin trough. Margins expand every quarter from here, exiting 2026 at low double-digit operating margins with $18β19B of run-rate revenue on the way to $30B . This is the year the earning power becomes visible on the page.
Then the moat nobody's pricing: sovereign AI.
$CRWV stood up CoreWeave Federal for U.S. government and the Defense Industrial Base, joined the DOE's Genesis Mission for national security, and is pursuing FedRAMP authorization. Classified and defense compute has to run on U.S. soil, inside U.S. jurisdiction. Washington will keep that work with the domestic leader, and a Netherlands-domiciled company built out of Russian-origin Yandex sits outside that door.
And the debt everyone frets about is the most misunderstood part of the story. Every big loan is backed by a signed take-or-pay contract and the GPUs themselves. The $8.5B facility earned the first investment-grade rating of its kind, with no maturities until 2029, and interest already covered more than 2x on adjusted EBITDA this early in the build. The same contracts that fill the backlog service the debt. Cost of capital keeps falling, S&P moved the outlook to positive, and CoreWeave just joined the Nasdaq-100.
So here is the real question, and it decides everything. If you do not believe AI is a generational shift, if you think the demand is a mirage and the models are near their ceiling, then you should not own any neocloud! Zero. Not CoreWeave, not
$NBIS, not one of them. This whole sector rests on one conviction: that AI keeps getting more capable and the world keeps needing more compute to run it. And the biggest companies on earth are voting with real money. The Mag 7 and the hyperscalers are pouring hundreds of billions of their own cash into AI infrastructure,
$Meta alone budgeting $115 to $135 billion in capex this year.
Leopold Aschenbrenner's Situational Awareness makes the case for where this leads (he hardly needs a citation but here:
situational-awareness.ai/). He traces the same scaling that carried models from preschooler to smart high-schooler in four years and argues it points to AGI by around 2027. His more staggering claim comes next. Once AI can do AI research, hundreds of millions of those systems could automate the work themselves, compressing a decade of algorithmic progress into a single year and running at many times human speed. That feedback loop, smarter AI building smarter AI, is the intelligence explosion, and it turns compute into the most strategic resource on the planet. He calls it the most extraordinary techno-capital acceleration ever set in motion, with trillions flowing into GPUs, data centers, and power before the decade is out, and a fierce scramble to lock up every power contract still available. That scramble is happening right now, and CoreWeave is winning its share of it. It's THE primier United States Neocloud,
Whether the timeline lands in 2027 or later, the direction is set, and the entire race runs on one thing: compute. Every model, every breakthrough, every step toward that future gets trained and served on infrastructure someone has to own and operate. I believe the demand is real, I believe it is durable, and I believe CoreWeave is the leading entirely U.S.-based company building that infrastructure at the frontier, first to every new chip, with the power and the contracts already locked. If you share that conviction, you want the leader. This is it.
$CRWV 30B contracted for 2027. ~3.5 GW power under contract, converting fast. $99.4B backlog. Nine of the top ten AI labs. Sovereign moat. NVIDIA in at $87.20. Priced at discount.
Long and sized for it. And I plan to HODL to realize a very minimal 5-10x. I think as a growth story in the Nasdaq 100, with the success of the AI industry, CoreWeave is an obvious winner and a company of very great importanceπ
