Prediction: CoreWeave's Contracted Backlog Passes $150 Billion Before 2028

CoreWeave ended June with a $104.2 billion revenue backlog, up 246% year over year, and added more than $25 billion of net new customer commitments early in Q3. The article argues backlog can exceed $150 billion before the end of 2027, requiring roughly $46 billion of net additions over six quarters, or about 28% annual growth from June levels. Demand visibility is strong, but execution risks remain substantial: only 41% of remaining obligations convert within 24 months, customer concentration is high, and Q2 net interest expense totaled $640 million as the company funds data-center capacity buildout.

Analysis

CRWV's equity value will be set less by headline backlog than by the spread between contracted revenue yield and its all-in cost of capital. With a material portion of obligations extending beyond two years, the backlog supports asset financing only if counterparties retain demand and capacity is commissioned on schedule; delays create a damaging combination of idle debt service, depreciation, and deferred revenue. The key underwriting variable is therefore conversion velocity—revenue per deployed GPU and utilization—not the aggregate RPO balance.

Customer concentration makes the stock a leveraged call on a small number of AI capex budgets. META has both the scale and technical capability to internalize incremental compute over time, so a renewal or expansion should not be valued like irreversible third-party demand; the same is true for sophisticated trading customers that may optimize workloads aggressively. This concentration can produce upside surprises from one large signing, but it also raises the probability that a single procurement pause drives multiple compression before reported revenue weakens.

Near term (days to 3 months), additional contract announcements can support CRWV sentiment, but the more consequential catalyst is the next disclosure on commissioned capacity, utilization, capex, and funding costs. Over 6-18 months, incremental GPU demand remains favorable for NVDA, yet CRWV is a lower-quality expression because it absorbs financing, execution, power, and data-center availability risk while NVDA monetizes the hardware sale earlier. The contrarian view is that a larger backlog could worsen equity risk if it requires funding growth with increasingly expensive secured debt before cash conversion catches up.

Thesis falsification: sustained improvement in operating cash flow after interest, stable or declining net interest expense as a percentage of revenue, and rising contracted-capacity utilization would demonstrate that the financing burden is scaling efficiently. Conversely, any reduction in annualized revenue outlook, delivery-date slippage, customer concentration increase, or widening debt spreads should be treated as evidence that RPO is not translating into equity value.

Key Decisions for Investors

• Maintain a relative-value long NVDA / short CRWV position over the next 1-3 months, sized beta-neutral. NVDA captures continued AI infrastructure demand with materially less project-finance exposure; cover the CRWV short if commissioning and utilization disclosures improve enough to show positive cash generation after interest.
• Do not add directional CRWV exposure solely on backlog announcements. Upgrade to a long only after management provides evidence that new capacity is funded at stable rates and revenue growth outpaces interest expense; absent that, contract wins are a trading catalyst rather than an investable earnings catalyst.
• Use META as a watch item rather than a direct beneficiary: track its disclosed AI capex and any commentary on owned versus rented compute. A pivot toward internally owned infrastructure is a negative read-through for CRWV renewal economics over 6-18 months, while accelerating external-capacity commitments would support the short leg being reduced.
• Set a downside alert around CRWV's next earnings release: reduce risk or add short exposure if annualized revenue guidance is maintained only through higher capex/debt, or if net interest expense rises faster than revenue. The risk/reward is asymmetric because delivery delays impair both earnings and refinancing capacity simultaneously.