CoreWeave, Inc.CRWV

CoreWeave, Inc. closed its March-2026 quarter with $2.08bn in revenue, a headline that sounds robust until you notice the net loss of $740m and an EPS of $-1.40-figures that would make even the most optimistic seed-stage founder reach for the Tums. The company’s $55.57bn in total assets are dwarfed by $50.81bn in liabilities, leaving shareholders with just $4.76bn of equity to split among 527m diluted shares. Cash on hand sits at $2.24bn, a number that looks healthier than it feels once you recall that $7.7bn was ploughed into property, plant, and equipment during the same three-month window.

CoreWeave isn’t a cloud provider in the way your granddad’s bank is a ‘financial services provider’-it sells purpose-built AI infrastructure: GPU, CPU, and networking stacks optimised for the kind of workload that makes Fortune 500 CIOs sweat through their slide decks. The company’s managed Kubernetes service promises ‘50% fewer interruptions per day’ and ‘little to no fleet management overhead’, language so targeted it reads like a love letter to harried AI teams who’ve just survived a week of outages and vendor lock-in nightmares. CoreWeave’s storage and networking pages double down on low-egress, high-throughput promises, all built on hardware that most mortals can’t afford to power, let alone cool.

The company’s March-2026 10-Q shows revenue from contracts with customers (ex-tax) of $2.08bn, offset by an operating loss of $144m and a net loss attributable to parent of $740m. Operating cash flow clocked in at $2.98bn, but that cash burn was funded largely by capex of $7.7bn, R&D spend of $104m, and a share base that ballooned to 527m diluted shares. EPS came in at $-1.40, a figure that sits somewhere between ‘disappointing’ and ‘existential’ when stacked against the growth story the pitch deck still tells. (source: 10-Q, period ending 2026-03-31, https://www.sec.gov/Archives/edgar/data/1769628/000176962826000099/000176962826000099.htm)

The company’s product pages-storage, networking, Kubernetes, and CPU/GPU compute-are heavy on benchmarks (DeepSeek-V3 trained in 2.02 minutes on 8,192 GB300 NVL72s, fastest in MLPerf v6.0) and light on unit economics. The pricing page cheerfully promises ‘details on our GPU cloud pricing’, but the evidence stops at the landing page; no concrete price lists, discounts, or utilisation bands are supplied in the source rail, so the reader is left guessing whether ‘exclusive to inference platform customers’ means ‘cheap’ or ‘existentially expensive’. The storage page touts ‘no egress fees’ and a ‘[0]EM program’ that reduces transfer friction, yet the same page offers no public pricing table, no sample bills, and no customer counts beyond the obligatory ‘partners like VAST Data and WEKA’. (source: CoreWeave Cloud Pricing, https://www.coreweave.com/pricing; Storage for AI, https://www.coreweave.com/products/storage; Networking Services That Enhance Performance, https://www.coreweave.com/products/networking; Kubernetes Management for GenAI, https://www.coreweave.com/products/coreweave-kubernetes-service; CPUs for AI Acceleration, https://www.coreweave.com/products/cpu-compute)

CoreWeave’s SEC submissions JSON lists 144 filings since inception, including a recent Form 144 filed on 2026-07-21, a procedural paper that rarely arrives without insiders eyeing the exit. Five Form 4 filings between 22-23 July 2026 show multiple insiders disposing of shares, a signal that reads less like ‘diversification’ and more like ‘liquidity event’. The company’s market narrative still revolves around ‘AI acceleration at scale’, yet the balance sheet’s real headline is leverage: $50.81bn in liabilities against $55.57bn in assets is the kind of ratio that keeps credit committees polite but wide awake. (source: SEC submissions JSON, https://data.sec.gov/submissions/CIK0001769628.json; CRWV 144 (2026-07-21), https://www.sec.gov/Archives/edgar/data/1769628/000195004726007250/xsl144X01/primary_doc.xml; CRWV 4 (2026-07-22/23), https://www.sec.gov/Archives/edgar/data/1769628/000176962826000323/xslF345X06/form4.xml)