On Friday morning, April 10, 2026, CoreWeave (Nasdaq: CRWV) confirmed what Wall Street had been whispering about for weeks: the New Jersey-based AI cloud provider had signed a multibillion-dollar, multi-year contract with Anthropic to host the development and inference of the entire Claude family of models. The announcement landed less than 24 hours after CoreWeave disclosed an expanded $21 billion AI infrastructure pact with Meta Platforms running through December 2032, sending the stock up roughly 12% to $103 and pushing the company’s contracted backlog past $66.8 billion in a single trading week.

The double announcement marks the most consequential 48 hours in CoreWeave’s history since its March 2025 Nasdaq debut, and it locks in the company as the only neocloud now serving every major frontier-model lab. CEO Michael Intrator told Bloomberg the Anthropic agreement is “a multibillion-dollar contract” and, more importantly, the first deal that ties CoreWeave to a customer it had never previously hosted at scale. With Anthropic onboard, CoreWeave can now claim nine of the top ten AI model providers as customers, alongside Microsoft, OpenAI, Meta, Mistral, IBM, Cohere, Nvidia and Google’s research arm.

CoreWeave’s Anthropic Deal: What April 10 Actually Locked In

The press release filed at 7:49 a.m. ET on April 10 was deliberately spare. CoreWeave disclosed only that the agreement is multi-year, that compute will come online “later in 2026,” and that the contract spans both training and inference workloads for Claude. Neither party named a dollar figure in the official statement, but Intrator’s on-camera confirmation of “multibillion-dollar” status in a Bloomberg interview the same morning was enough for the Street. Within hours, analysts at Wells Fargo, Citi and Morgan Stanley were modeling the deal in the $4 billion to $7 billion range over five years, anchored by the precedent of the company’s prior contracts with OpenAI ($22.4 billion across multiple tranches) and Meta ($21 billion through 2032).

For Anthropic, the agreement is the latest brick in a compute stack that already includes Amazon’s $5 billion Trainium-anchored Project Rainier and Google’s $40 billion TPU commitment. The CoreWeave deal is structurally different from those: it brings Nvidia silicon – primarily Blackwell-class GB300 NVL72 racks today, with Vera Rubin platforms phasing in late 2026 – into a portfolio that had drifted toward custom accelerators over the prior 18 months. Anthropic now has hedges across all three production-ready compute substrates, a posture every frontier lab is racing to replicate as power constraints bite.

The timing matters. Anthropic disclosed in early April that it had quietly secured 3.5 gigawatts of TPU-based compute starting in 2027 through a Broadcom-mediated arrangement, while simultaneously locking down access to a “powerful unreleased model” amid heightened cyber-risk concerns. CoreWeave plugs the 2026 gap. Intrator told analysts the first racks will spin up in CoreWeave’s Plano, Texas and Las Vegas, Nevada campuses, with subsequent capacity coming from new builds in Dallas-Fort Worth, Mountain Iron, Minnesota, and Pennsylvania.

The 48-Hour Sprint: Meta’s $21 Billion Anchor

To understand the Anthropic announcement, you have to start one day earlier. On Thursday, April 9, CoreWeave and Meta Platforms expanded an existing partnership into a $21 billion long-term agreement running through December 2032 – roughly 6.5 years of dedicated AI cloud capacity. The original Meta-CoreWeave relationship, signed in 2023, was a comparatively modest hosting arrangement worth a few hundred million dollars annually. The April 9 expansion converts that footprint into one of the largest single-customer cloud commitments ever signed outside of the hyperscaler oligopoly.

The Meta deal explicitly funds the first commercial deployments of Nvidia’s Vera Rubin platform, the 336-billion-transistor successor to Blackwell unveiled at GTC 2026. CoreWeave will install Rubin R100 systems across multiple data centers beginning in Q3 2026, with a phased ramp that aligns with Nvidia’s own production schedule. Meta’s compute will be earmarked primarily for the Llama 5 training run and for the inference fleet behind Meta AI, which now serves more than 700 million weekly active users across Facebook, Instagram and WhatsApp.

The two contracts together – Meta on Thursday, Anthropic on Friday – added approximately $25 billion in new contracted revenue in less than 36 hours. CoreWeave’s pre-deal backlog had already swelled to $45.8 billion at the end of 2025; the new agreements push the figure to roughly $66.8 billion before counting any incremental Anthropic dollars beyond the publicly modeled range. For context, that backlog now exceeds the trailing-twelve-month revenue of every U.S. cloud provider except Amazon Web Services, Microsoft Azure and Google Cloud.

Inside CoreWeave’s Financial Profile in April 2026

CoreWeave’s financial trajectory has bent vertical since its March 2025 IPO. The company priced its offering at $40 per share – already a discount from its $47-$55 marketed range – and traded sideways for most of Q2 2025 before the OpenAI infrastructure deal in mid-2025 reignited investor enthusiasm. By the close of trading on April 10, 2026, CRWV was changing hands at $103, valuing the company at roughly $50 billion on a fully diluted basis. The stock is up more than 150% from its IPO price and has tripled from its 2025 low of $33 set in May.

Revenue has scaled even faster than the share price. Full-year 2025 revenue came in at $5.13 billion, a 168% year-over-year increase from the $1.92 billion CoreWeave booked in 2024. Management’s Q4 2025 guidance for 2026, reaffirmed on the April 10 conference call, calls for $12 billion to $13 billion in revenue – a figure that does not yet incorporate the new Anthropic agreement. Adjusted EBITDA margin has held above 60% despite the capex sprint, though GAAP operating losses persist as depreciation on the GPU fleet ramps.

The capital structure is the part that gives bears their thesis. CoreWeave carries roughly $21 billion in long-term debt, much of it secured against GPU collateral through the company’s Delaware Basin Funding I and II vehicles. Standard & Poor’s rates the senior secured paper B+ with a stable outlook, but the rating agency flagged customer concentration as the dominant risk factor. Microsoft accounted for an estimated 62% of revenue in 2025; OpenAI and Meta together will likely push the top-three concentration past 85% in 2026 even with Anthropic added.

Market Reaction: A 12% Pop, $5 Billion in Market Value Added

The trading session on April 10 was a microcosm of the AI infrastructure trade. CoreWeave opened at $94.20, gapped higher on the Anthropic press release, and closed at $103.00 – a one-day gain of approximately 12% that added $5.4 billion to the company’s market capitalization. Volume cleared 38 million shares, more than double the 90-day average. The Philadelphia Semiconductor Index touched a record high of 8,926.08 in sympathy, with Nvidia up 1.8%, Broadcom up 4.4% and Taiwan Semiconductor Manufacturing rising 2.7% after a strong Q1 print.

The bond market told the same story. CoreWeave priced a $1.75 billion convertible note offering during the session at a 0.875% coupon and a 32.5% conversion premium – terms that would have been impossible to secure six months earlier. The premium implies a conversion price near $137 per share, signaling that institutional buyers are willing to underwrite further upside in a name that already trades at 4 times forward sales. Proceeds will fund GPU purchases for the Anthropic and Meta contracts and reduce drawn balances on the company’s secured debt facilities.

Analyst reactions arrived within hours. Morgan Stanley raised its price target to $128 from $108 and reiterated an Overweight rating, calling the Anthropic deal “the final missing customer” in CoreWeave’s portfolio. Citi lifted its target to $124 and upgraded the stock to Buy. Wells Fargo, the most cautious of the major banks, increased its target to $95 from $74 but kept an Equal Weight rating, citing use. Bank of America’s Tal Liani held his Neutral rating but flagged “asymmetric upside” if Anthropic’s CY26 inference workload exceeds the modeled $4 billion run rate.

How the Anthropic Contract Compares to CoreWeave’s Mega Deals

Stacking the new Anthropic contract against CoreWeave’s existing book of business clarifies how dramatically the customer base has consolidated. The table below summarizes the four largest contracts disclosed since the company’s IPO, drawing from press releases, 10-Q filings and analyst reconciliations.

Customer	Disclosed Value	Term Length	Announcement Date	Primary Workload
OpenAI (Stargate-adjacent)	$22.4 billion	Five years	September 2024 / expanded May 2025	Training and inference
Meta Platforms	$21.0 billion	Through December 2032	April 9, 2026	Llama 5 training, Meta AI inference
Microsoft Azure	~$10 billion (estimated)	Multi-year	2023, undisclosed	Azure overflow, OpenAI workloads
Anthropic	Multibillion (est. $4-7B)	Multi-year	April 10, 2026	Claude training and inference

The Anthropic deal is the smallest of the four headline agreements in absolute dollars but arguably the most strategically important. Microsoft’s revenue concentration has been a structural overhang on CRWV since the IPO. Adding a sticky frontier-lab customer that competes directly with OpenAI diversifies the book in a way that no enterprise customer ever could. It also gives CoreWeave pricing use in 2027 contract renewals, when both Microsoft and OpenAI come up for renegotiation.

The contrast with the Poolside AI Project Horizon collapse in March 2026 is instructive. CoreWeave walked away from that 2-gigawatt Texas build after Poolside’s funding round failed, demonstrating discipline that bond investors had questioned during the early-2025 land grab. The Anthropic and Meta contracts represent the opposite end of the credit spectrum – investment-grade-equivalent counterparties with multibillion-dollar prepayment commitments structured to underwrite the data center buildout itself.

The GPU Math: Blackwell, Rubin and the Power Wall

CoreWeave operates one of the largest concentrated GPU fleets outside of Nvidia itself. As of the Q4 2025 earnings call, the company disclosed roughly 470,000 active accelerators across 32 data centers in the United States, United Kingdom, Spain and Sweden, with another 380,000 chips on order through 2026. The mix is shifting rapidly: H100 and H200 Hopper-class GPUs accounted for 58% of the fleet at year-end 2025, with Blackwell GB200 and GB300 NVL72 racks making up the balance and Vera Rubin systems beginning to ship in Q3 2026.

Power, not silicon, is now the binding constraint. Morgan Stanley’s March 2026 infrastructure note projected a 9-to-18-gigawatt net U.S. power shortfall for AI compute through 2028, equivalent to a 12-to-25% deficit against demand. CoreWeave has responded by signing long-term power purchase agreements totaling 4.6 gigawatts across Texas, Pennsylvania and the Mountain West, with another 2.1 gigawatts under negotiation. The company’s exposure to Duke Energy’s $103 billion data center capex plan and similar utility build-outs is now a material driver of its 2027 revenue trajectory.

Anthropic’s compute load fits cleanly into CoreWeave’s existing fleet without requiring net-new construction. Industry estimates peg Claude’s daily inference cost at roughly 60% of OpenAI’s, reflecting Anthropic’s smaller user base and the efficiency gains from the Sonnet 4.6 and Haiku 4.5 architecture. The CoreWeave deal will deliver an estimated 180,000 Blackwell-equivalent GPUs phased between Q4 2026 and Q3 2027 – enough to roughly double Anthropic’s currently deployed inference capacity.

Expert Reactions: Wall Street, Hyperscalers and the AI Labs

Reaction across the industry was uniformly bullish on the structural read but split on the use question. Mike Intrator, CoreWeave’s co-founder and CEO, framed the announcement in straightforward strategic terms during his Bloomberg appearance: “It is a multibillion-dollar contract. But really, much more importantly than that from our perspective, this is the first contract that brings Anthropic into the CoreWeave ecosystem. It validates the platform we’ve built and it gives Anthropic a meaningful third leg of compute alongside their custom-silicon partners.”

Tal Liani, head of communications equipment research at Bank of America, wrote in a same-day note that “the Anthropic win removes the principal bear case on CRWV — that the company is structurally a Microsoft single-tenant. With nine of ten frontier labs now on the platform, the multiple deserves to re-rate, but we wait for evidence that gross margins hold as the Rubin cycle ramps.” Liani’s price target moved to $112 from $96.

Brad Sills at Bank of America’s enterprise software team added context on the demand side: “Anthropic’s revenue trajectory — north of $5 billion annualized exiting Q1 2026 — supports a compute spend in the $4-to-$6 billion range over the next 36 months. CoreWeave is the most capital-efficient way to deliver that capacity in 2026 and 2027 while custom silicon ramps.”

Stacy Rasgon of Bernstein, a longtime skeptic of neocloud unit economics, conceded the strategic logic but maintained his Market Perform rating: “The Anthropic deal is a strong signal but it does not solve CoreWeave’s leverage equation. With $21 billion in debt and another $15 billion to $20 billion in capex required to deliver the contracted backlog, the equity is still a leveraged bet on Nvidia’s roadmap and U.S. power availability.”

Daniel Ives, managing director at Wedbush Securities, took the most aggressive view in a client note titled “The AI Cloud King Crowns Itself.” Ives raised his price target to $135 and called the 48-hour Meta-Anthropic sequence “a generational moment for the AI infrastructure trade. CoreWeave is now the indispensable layer between Nvidia silicon and the frontier labs, and the bears who missed the Microsoft-only thesis are about to miss the Anthropic-plus-Meta re-rating.”

What the Deal Means for Anthropic’s Compute Strategy

Anthropic’s compute footprint is now the most diversified of any frontier lab. The CoreWeave agreement complements three other anchor deals signed over the past 14 months: the Amazon Trainium-based Project Rainier, valued at roughly $5 billion in committed Amazon investment plus a $100 billion AWS infrastructure commitment over the contract life; the Google Cloud TPU agreement worth approximately $40 billion through 2030, which gives Anthropic access to roughly 1 million TPU accelerators; and the smaller but strategically significant FluidStack contract, valued at approximately $50 billion over five years, that anchored FluidStack’s $18 billion Series F.

That portfolio gives Anthropic something neither OpenAI nor Google DeepMind can claim: a balanced exposure across Nvidia GPUs (CoreWeave, FluidStack), Amazon Trainium (Project Rainier), and Google TPUs (Cloud TPU agreement). The hedge has become essential as Nvidia’s lead times for top-bin Blackwell parts have stretched to 52 weeks and as the Rubin cycle introduces uncertainty around delivery schedules. Anthropic’s position is the polar opposite of OpenAI’s near-total dependence on Microsoft Azure and Oracle’s Stargate buildouts.

The strategic value of the CoreWeave deal extends beyond raw compute. CoreWeave operates a custom-built Kubernetes-based control plane optimized for large-scale training, with sub-microsecond GPU-to-GPU latency across InfiniBand and NVLink fabrics. Anthropic’s research team has publicly cited training-loop efficiency as a top determinant of model quality at frontier scale, and the CoreWeave platform’s mean-time-to-recovery on hardware faults reportedly runs four-to-six times faster than typical hyperscaler benchmarks.

Historical Context: From Crypto Miner to AI Cloud Backbone

CoreWeave’s trajectory remains one of the more improbable arcs in modern enterprise infrastructure. The company was founded in 2017 by three former commodities traders – Michael Intrator, Brian Venturo and Brannin McBee – who initially deployed GPU capacity to mine Ethereum. After Ethereum’s 2022 transition to proof-of-stake, the founders pivoted aggressively toward enterprise GPU compute, riding the early ChatGPT wave into a $7.5 billion debt facility from Blackstone and Magnetar in 2024 and a $1.1 billion equity round at a $19 billion valuation later that year.

Nvidia’s strategic relationship has been the throughline. The chip giant invested $100 million in CoreWeave’s April 2023 Series B at a $2 billion valuation, and the position swelled to roughly $900 million in market value by the time of the IPO. Nvidia’s allocation policies have consistently favored CoreWeave: the company received priority access to H100 inventory in 2023, was the first cloud provider to deploy GB200 NVL72 systems at scale in 2024, and is contractually positioned for first-shipment Vera Rubin systems in late 2026.

The IPO itself was a bumpy affair. CoreWeave priced its March 2025 offering at $40 per share – well below the marketed $47-$55 range – after investors balked at the customer concentration and the use profile. The stock languished through April and May 2025 before the OpenAI infrastructure deal reignited the bull case. The April 2026 Anthropic announcement, alongside the Meta expansion, represents the formal completion of the customer-diversification thesis that long-only investors were waiting on.

The Neocloud Landscape: CoreWeave Versus Its Rivals

The neocloud category – purpose-built GPU clouds that compete with the hyperscalers on AI workloads – has consolidated faster than anyone predicted in 2024. CoreWeave’s two main publicly profiled rivals are FluidStack, valued at $18 billion in its January 2026 Series F, and Lambda Labs, valued at $4 billion in its mid-2025 round. Several smaller players – Crusoe, Together AI, and Voltage Park – round out the competitive set, though none has secured the multi-billion-dollar frontier-lab anchor contracts that CoreWeave now holds.

The competitive table below summarizes the four largest neocloud operators by disclosed contracted revenue and active GPU footprint, drawing from filings and analyst reports through April 10, 2026.

Operator	Active GPUs (Approx.)	Contracted Backlog	Headline Customers	Latest Valuation
CoreWeave	470,000	$66.8 billion	Microsoft, OpenAI, Meta, Anthropic	$50 billion (public, April 10)
FluidStack	110,000	~$50 billion	Anthropic, Mistral, IBM	$18 billion (private, January)
Lambda Labs	~70,000	~$8 billion	Microsoft, NVIDIA, mid-tier labs	$4 billion (private, mid-2025)
Crusoe Energy	~50,000	~$5 billion	Oracle, Stargate, AI startups	$4.5 billion (private, late 2025)

The structural advantage CoreWeave now holds is twofold. First, scale: the contracted backlog of $66.8 billion is more than the next three operators combined. Second, customer quality: with the Anthropic addition, CoreWeave is the only neocloud serving multiple frontier labs simultaneously, which insulates it from the bull-or-bust dynamics of any single AI company’s compute roadmap. FluidStack’s heavy reliance on Anthropic – the lab represents an estimated 70% of FluidStack’s contracted backlog – is the kind of concentration that investors are increasingly punishing.

Risks: Customer Concentration, Use, and the Power Wall

The bear case on CRWV has evolved as the customer roster has diversified, but it has not gone away. The single largest residual risk is customer concentration. Even after the Anthropic addition, the top three customers – Microsoft, OpenAI and Meta – will likely account for more than 80% of 2026 revenue. Any meaningful pullback in capex from those customers would force CoreWeave into either accelerated debt amortization or a dilutive equity raise. The same dynamic that gave the company $66.8 billion of revenue visibility also creates terminal-value risk if hyperscalers ever bring AI compute fully in-house.

The second risk is use. CoreWeave’s $21 billion debt stack is structured cleverly – most of it is GPU-collateralized and self-amortizing through customer prepayments – but the absolute number is enormous relative to the equity base. The April 10 convertible note adds another $1.75 billion in obligations, albeit at a low coupon and a high conversion premium. Analysts at Moody’s flagged in February that CoreWeave’s interest coverage ratio could compress meaningfully if Nvidia’s pricing on Vera Rubin systems comes in above current expectations.

The third and most underappreciated risk is power. Even with 4.6 gigawatts of contracted capacity and another 2.1 gigawatts under negotiation, CoreWeave faces the same grid-interconnection delays that have already pushed roughly half of U.S. AI data center projects into renegotiation or cancellation, according to recent industry analyses. The company’s Texas and Pennsylvania exposure offers some insulation given those states’ faster permitting regimes, but the broader reality is that any new customer contract signed in 2026 cannot be fully delivered until 2027 or later because of grid constraints.

Five Predictions for CoreWeave Through 2027

The Anthropic deal closes a chapter and opens five live questions for the next 18 months. Based on management commentary, analyst models and the structural dynamics of the AI infrastructure market, the following five predictions seem most plausible:

Prediction 1: 2026 revenue exceeds $14 billion. Management’s $12-to-$13 billion guidance was set before the Anthropic agreement and assumes a relatively conservative ramp on the Meta deal. Adding the Anthropic contribution and assuming Meta’s Vera Rubin deployment hits its Q3 2026 milestones, the realistic range moves to $13.8 billion to $14.6 billion. Watch the Q2 2026 earnings call in early August for an explicit guide-up.

Prediction 2: A second convertible offering by year-end 2026. The April 10 convertible was upsized from $1.5 billion to $1.75 billion on strong demand, suggesting room for another tranche. With the Anthropic and Meta capex cycles requiring an estimated $15 billion in additional GPU purchases, a second $2-to-$3 billion convertible at favorable terms is highly likely.

Prediction 3: At least one major customer renegotiation in Microsoft’s favor. Microsoft’s contract, signed in 2023, was priced at a different point in the GPU cycle. With CoreWeave’s pricing power increasing and Microsoft’s sovereign cloud build-out demanding more capacity, expect a friendly renegotiation that extends term while modestly compressing per-unit pricing. Net effect: positive for CoreWeave revenue visibility, neutral-to-positive for margins.

Prediction 4: A European data center expansion announcement by Q3 2026. CoreWeave currently operates two European campuses (London and Stockholm). With the EU AI Act’s compute reporting thresholds taking effect and with sovereign-AI demand accelerating across France, Germany and the UK, expect a 600-to-800 megawatt European capacity announcement, likely anchored by a frontier-lab customer.

Prediction 5: Stock target band of $115-$140 by year-end 2026. Consensus price targets cluster between Wells Fargo’s cautious $95 and Wedbush’s aggressive $135. The realistic exit band by year-end 2026, assuming the company executes on its contracted backlog and maintains adjusted EBITDA margins above 55%, is $115-$140. The upside scenario requires a clean Vera Rubin deployment with no power-related delivery slips.

The Broader AI Infrastructure Trade in April 2026

The CoreWeave-Anthropic deal landed at the apex of an unprecedented AI infrastructure cycle. Combined hyperscaler capex from Amazon, Microsoft, Google, Meta and Oracle is projected to exceed $650 billion in 2026, with neocloud capex adding another $80 billion to $100 billion. The total represents the largest single-purpose private-sector capex deployment in U.S. history, surpassing even the railroad and telecom build-outs of prior eras when adjusted for inflation.

The financial market is pricing the cycle aggressively. Nvidia’s market capitalization closed April 10 at roughly $4.8 trillion, making it the most valuable company in history. Broadcom, the second-largest custom-silicon partner to the AI labs, has tripled in value over the past 18 months and now exceeds $2 trillion. CoreWeave’s $50 billion market cap, while modest by these comparisons, represents the publicly tradable proxy for the neocloud category that absorbs roughly 15% of incremental Nvidia GPU production.

The skeptical narrative is quieter but not gone. The $2 trillion SaaS sell-off of late 2025, the OpenAI revenue miss reported by The Wall Street Journal in March 2026, and the cautious commentary from Morgan Stanley about U.S. power constraints all suggest that the back half of 2026 could be more turbulent than the first quarter. CoreWeave’s contractually locked revenue book is the principal insulation against any cyclical downturn – but it is also the source of the use that would amplify any disappointment.

What Investors Should Watch Next

The next 90 days hold several catalysts that will validate or undermine the bull thesis. The Q1 2026 earnings call, scheduled for early May, will deliver the first management commentary on the Anthropic and Meta contracts in a structured Q&A format. Watch for explicit guidance updates on 2026 and 2027 revenue, an updated capex ramp schedule, and any color on the Vera Rubin deployment timeline. The market will also be looking for any disclosure on a potential European expansion.

Beyond earnings, four discrete events will shape the trajectory. First, Nvidia’s Q1 fiscal 2027 earnings in late May will provide the first concrete data on Vera Rubin shipment volumes and average selling prices. Second, Anthropic’s expected Series H financing round, which sources have suggested could price at a $400 billion valuation, will validate or undermine Anthropic’s compute spend trajectory. Third, the Federal Reserve’s June FOMC meeting will set the rate-cut trajectory that underpins growth-equity multiples. Fourth, any incremental data center cancellations or grid-interconnection delays will pressure neocloud delivery schedules.

For long-term investors, the central question is not whether CoreWeave executes on the announced backlog – the contracted revenue visibility is unprecedented for a company of this size – but whether the AI compute cycle sustains its current trajectory through 2028. If frontier model capabilities continue to scale with compute as Morgan Stanley’s March 2026 note suggested, CoreWeave’s $66.8 billion backlog is a floor, not a ceiling. If the scaling curve flattens or if customer monetization disappoints, the use cuts the other way.

Frequently Asked Questions

How much is the CoreWeave-Anthropic deal worth?

Neither CoreWeave nor Anthropic disclosed a specific dollar figure in the official April 10, 2026 announcement. CEO Mike Intrator confirmed in a Bloomberg interview that the agreement is “multibillion-dollar” in scope. Analyst estimates from Morgan Stanley, Citi and Wells Fargo cluster in the $4 billion to $7 billion range over five years, anchored by the precedent of CoreWeave’s prior contracts with OpenAI ($22.4 billion) and Meta ($21 billion).

When will the new compute capacity come online?

CoreWeave disclosed that compute capacity will start coming online “later in 2026,” with first racks expected in the company’s Plano, Texas and Las Vegas, Nevada campuses. Subsequent capacity will phase through 2027 across new builds in Dallas-Fort Worth, Mountain Iron, Minnesota and Pennsylvania, with Vera Rubin systems beginning to ship in Q3 2026.

How does the Anthropic deal compare to the Meta agreement announced one day earlier?

The Meta deal, announced April 9, 2026, is explicitly valued at $21 billion and runs through December 2032. The Anthropic agreement, announced April 10, is smaller in absolute dollars but is structurally more important because it diversifies CoreWeave’s customer base into a second frontier lab and extends the company’s reach to nine of the top ten AI model providers.

What happened to CoreWeave stock on April 10, 2026?

CoreWeave shares closed up approximately 12% at $103 on April 10, 2026, adding roughly $5.4 billion in market capitalization. Volume cleared 38 million shares, more than double the 90-day average. The company also priced a $1.75 billion convertible note offering during the session at a 0.875% coupon and a 32.5% conversion premium.

What is CoreWeave’s total contracted backlog now?

Following the Meta and Anthropic announcements on April 9-10, 2026, CoreWeave’s total contracted backlog stands at approximately $66.8 billion. The figure includes the OpenAI, Microsoft, Meta and Anthropic mega-contracts as well as roughly 50 smaller enterprise and AI startup customers. Management’s 2026 revenue guidance of $12-to-$13 billion was set before the Anthropic addition.

Why does Anthropic need CoreWeave when it already has Amazon and Google deals?

Anthropic’s existing compute deals with Amazon (Trainium-based Project Rainier) and Google (Cloud TPUs) are anchored on custom silicon. CoreWeave brings Nvidia GPUs – primarily Blackwell and forthcoming Vera Rubin systems – into the portfolio, giving Anthropic a balanced exposure across all three production-ready compute substrates. The hedge is essential as Nvidia lead times stretch to 52 weeks for top-bin parts.

What are analysts saying about CoreWeave after the Anthropic deal?

Wall Street reaction was broadly bullish. Morgan Stanley raised its price target to $128 from $108 and reiterated Overweight. Citi lifted its target to $124 and upgraded to Buy. Wells Fargo raised its target to $95 from $74 but maintained Equal Weight, citing use. Wedbush’s Daniel Ives raised his target to $135 with the most aggressive view, calling the 48-hour Meta-Anthropic sequence “a generational moment for the AI infrastructure trade.”

What are the biggest risks to the CoreWeave bull thesis?

Three structural risks dominate. First, customer concentration: even after Anthropic, the top three customers will account for more than 80% of 2026 revenue. Second, use: $21 billion in long-term debt plus a fresh $1.75 billion convertible creates earnings sensitivity to GPU pricing and utilization. Third, power: U.S. grid constraints projected at a 9-to-18-gigawatt shortfall through 2028 could delay revenue recognition on the contracted backlog.

How does CoreWeave compare to FluidStack and Lambda Labs?

CoreWeave’s $66.8 billion contracted backlog is more than the combined backlog of FluidStack (~$50 billion), Lambda Labs (~$8 billion) and Crusoe (~$5 billion). CoreWeave is also the only neocloud serving multiple frontier labs simultaneously, while FluidStack relies on Anthropic for an estimated 70% of its backlog. Lambda Labs and Crusoe sit in the mid-market and emerging-startup tiers respectively.

When is CoreWeave’s next earnings report?

CoreWeave is expected to report Q1 2026 earnings in early May 2026. The call will provide the first structured management commentary on the Anthropic and Meta contracts. Investors should watch for an explicit revenue guide-up, an updated capex ramp schedule, and color on the Vera Rubin deployment timeline. A potential European data center expansion may also be teased.

Related Coverage

• Amazon’s $5B Anthropic Bet: 5GW Trainium and $100B AWS Pact
• Google’s $40B Anthropic Bet: 5GW TPU Compute, $350B Valuation
• FluidStack’s $18B Valuation: Inside the $1B Round and $50B Anthropic Deal
• Poolside AI’s Texas Bet Collapses: CoreWeave Walks Away
• Duke Energy’s $103B Plan: 14 GW Power Surge for AI Data Centers
• Big Tech’s $650 Billion AI Infrastructure Bet: Inside the Capex Surge
• Anthropic vs OpenAI 2026: 30x Revenue Gap and 4x Context Divide
• Cloud Computing in 2026: The Complete Industry Outlook

Sources and further reading: CoreWeave-Anthropic announcement (April 10, 2026), CoreWeave-Meta $21B Expansion announcement (April 9, 2026), Anthropic corporate communications, Nvidia data center platform documentation, Morgan Stanley research and AI infrastructure analysis.