Bernstein Maintains CRWV Underperform: short-term contracts remain strong, but long-term neocloud model faces structural pressure

Overview

This is a Bernstein company research report on CoreWeave (CRWV), focusing on investor feedback after the initial coverage regarding the new cloud business model, AI demand, hyperscaler competition, debt financing, NVDA support, and peer comparability. The report maintains an Underperform rating and a $56 target price on CRWV. The authors acknowledge that short-term AI demand is substantial, CRWV’s capacity is tight, and it may continue to sign large contracts, but they argue that long-term value capture is more likely to shift to much larger players with stronger ecosystems, lower financing costs, greater scale, and self-build capabilities.

Core views

The core view is that CRWV is not likely to see an immediate deterioration in fundamentals and may still add about $35B in contracts by the end of 2027, but as hyperscalers bring significant self-owned capacity online around 2028, GPUaaS capex ramps up, and other neocloud peers absorb demand, contract signing pace and revenue growth at CRWV may slow. The report believes CRWV’s software moat is unlikely to defend against the scale, network strength, and lower capital cost advantages of mega-clouds like Microsoft, Google, Amazon, and Oracle over the long run. High-quality customers are supportive of financing, but also create strong bargaining power and internal alternatives for clients, potentially compressing returns over time.

Analysis framework

The report is structured as grouped investor feedback with a point-by-point response: it first summarizes the key questions from bulls, bears, cloud-optimistic investors, balance-sheet-focused investors, NVDA-support proponents, and peer-comparison investors, then evaluates CRWV through AI power demand, hyperscaler self-build supply, contract signing pace, debt structure, customer bargaining power, and relative valuation framework. On valuation, it uses an EV/adjusted EBIT multiple approach, applying 2027E adjusted EBIT of $5.31 per share times 28.4x to derive a $56 target price.

Methodology notes

• valuation methodEV/adjusted EBIT multiple

  Measure CRWV valuation using enterprise value relative to adjusted EBIT

  The report uses a 2027E adjusted EBIT of $5.31 per share and a 28.4x EV/adjusted EBIT multiple to calculate the target price, arriving at $56.

• supply-demand analysisAI power demand and data center supply framework

  Compare AI demand growth with hyperscaler self-build and leased supply

  Bernstein expects global AI-related demand to be about 60GW by 2030 but notes that efficiency gains, self-build supply, and leasing supply will change scarcity for neocloud players.

• competition analysisHyperscaler substitution and bargaining power analysis

  Assess CRWV’s long-term moat relative to platforms such as AWS, Azure, and GCP

  The report says neoclouds have GPU cluster service abstraction capability, but lack the scale, diversified customer base, vertical integration, and cheaper capital advantages of traditional cloud giants.

• balance sheet analysisDebt financing and risk redistribution framework

  Judge whether lower-cost financing represents de-risking or financial engineering

  The report acknowledges CRWV has strong financing ability, but argues some structures may simply shift risk from creditors to equity holders and that this capability may not remain differentiated over time.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• CRWV.US

  coverage target

  Strengths

  Strong short-term capacity demand, mature GPUaaS platform, large signed contracts, relatively strong financing ability, and potential to continue delivering on near-term guidance.

  Weaknesses

  Customer concentration, heavy capex, high exposure to a single compute model (GPU), and a long-term lack of the scale, diversification, and vertical integration advantages enjoyed by hyperscalers.

  Comparison

  Compared with traditional cloud providers, CRWV is more like a GPU-focused neocloud infrastructure platform; compared with other neocloud peers, CRWV is more mature, but NBIS and IREN are also winning large hyperscaler contracts.

  Risks

  If hyperscaler self-built capacity is delayed, AI demand remains above expectations, CRWV’s software platform moat strengthens, or power scarcity persists, there is upside risk to the Underperform rating and target price.

• NVDA

  ecosystem investor and potential backer

  Strengths

  NVDA has strong cash flow and ample liquidity; investing in CRWV can help expand GPU distribution channels, cultivate a new cloud ecosystem, and increase negotiation leverage with hyperscalers.

  Weaknesses

  The report says NVDA may not be willing to support CRWV indefinitely, and NVDA’s investments already cover multiple neocloud players rather than backing only one winner.

  Comparison

  NVDA’s investments in CRWV, NBIS, Lambda, Crusoe, and others look more like ecosystem seeding than a single-company perpetual guarantee.

  Risks

  If NVDA continues to provide capital, supply, or commercial support, CRWV’s downside risk may be postponed.

• Hyperscalers

  customers, competitors, and potential substitutes

  Strengths

  Have proprietary networks, lower capital costs, established customer bases, broad service stacks, and vertical integration advantages.

  Weaknesses

  Hyperscaler GPUaaS commercialization and technology-stack commitment are not yet fully aggressive, so they may still rely on neocloud providers in the short term.

  Comparison

  The report believes hyperscalers are ultimately likely to compete head-on with neoclouds in enterprise GPUaaS and AI cloud services.

  Risks

  If self-build capacity launches later than expected, CRWV’s value proposition window is extended; if launches are successful, it could compress CRWV’s contract additions and margins.

Key data

• RatingUnderperformBernstein maintains a below-market stance on CRWV.
• Target price$56Based on a 28.4x 2027E EV/adjusted EBIT multiple.
• Current price$80.94Current price shown in the investment thesis table.
• Implied downsideabout -30.8%Estimated from a target price of $56 and current price of $80.94.
• 2027E adjusted EBIT per share$5.31Core input in the target price valuation.
• Expected additional contractsabout $35BEstimated contract volume CRWV could still add from current level to end-2027.
• Consensus divergence timing2027-2028The report says 2026 is broadly in line with consensus; divergence starts in 2027, and 2028 revenue forecasts are more than 30% below consensus.
• Long-term operating margin range25%-30%Analysts believe the company may reach this margin range over the long term.
• Long-term cost of capital rangeabout 7%The report believes the company’s overall cost of capital may fall to this range.
• NVDA ownership stake11%The report says NVDA currently holds about 11% of CRWV.

Impact & implications

For investors, this means CRWV trading in the near term may still be driven by AI themes and momentum and does not necessarily face immediate fundamental deceleration; however, the mid-to-long-term valuation risk comes from contract renewal and new-book velocity, hyperscaler self-build capacity, enterprise customer competition, cost of capital, and operating margin limits. The report suggests not simply extrapolating current strong demand into a permanent moat, and not treating NVDA support as sufficient downside protection.

Risks

• Delay in hyperscalers’ self-build capacity coming online may preserve CRWV’s value proposition.
• CRWV’s dedicated software platform continues to develop and sustain demand for its operated data centers.
• Continued power shortages at data centers may preserve CRWV’s pricing power.
• AI demand materially above Bernstein’s forecast, or the market tolerates prolonged low AI-profitability investment longer.
• If NVDA continues supporting CRWV, downside scenarios may be deferred or softened.

What to watch

• Whether hyperscalers increase GPUaaS investment and marketing.
• The pace of about 23GW of hyperscaler-owned capacity coming online around 2028.
• CRWV’s contract signing pace and contract size for new bookings around 2027.
• Whether NBIS, IREN, and other neocloud peers continue to win large hyperscaler contracts.
• Whether CRWV’s cost of capital truly declines, or risk is simply transferred to equity through structured financing.
• Whether enterprise demand shifts from neocloud platforms to hyperscalers’ own GPU services.