Bernstein and BofA Clash Over CoreWeave's Funding Gap as Product Push Takes a Back Seat

Two of Wall Street's more closely watched research desks are looking at the same CoreWeave balance sheet and reaching wildly different conclusions — not about whether customers want the company's AI cloud capacity, but about what it will cost to keep building it.

Bernstein's Madison Rezaei cut her price target to $51 from $74 on Wednesday while reiterating an Underperform rating, sharpening the firm's cautious stance on the stock. BofA's Tal Liani, by contrast, reaffirmed a Buy and a $140 target the same day. The gap between the two calls is striking enough on its own. What separates them even more is the underlying arithmetic: Bernstein pegs CoreWeave's additional capital needs at roughly $90 billion, while BofA puts the figure near $20 billion.

That divergence matters more than the headline targets, because it changes how heavily financing costs would weigh on the returns from new capacity. Rezaei is not writing off the business — she points to the large contract backlog and the prospect of power supply as providing a valuation floor. Her argument draws a line between the commercial opportunity and the financial burden of delivering on it, warning that fresh debt is likely to come at a steep price. The concern, in other words, extends beyond existing obligations to how the next phase of growth gets funded.

Liani takes the opposite view of where that spending leads. Additional infrastructure, he argues, will drive revenue growth in the triple-digit percentage range and lift operating margins, with the buildout supporting expansion over several years. Even in his bullish case, though, funding costs and capital requirements remain among the central risks.

Should investors sell immediately? Or is it worth buying CoreWeave?

A Backlog That Proves Demand, Not Financing

At the start of the third quarter of 2026, CoreWeave's remaining performance obligations stood at about $129 billion. BofA expects capital expenditure of $12.5 billion for that quarter, and when the results land in early November, the bank anticipates a 558% jump in quarterly investment compared with the year-earlier period. Those figures frame the debate neatly: a hefty backlog signals demand, but it does not pay for the servers, power or real estate needed to serve it. How quickly CoreWeave converts contracted work into revenue and profit — and how much extra capital that conversion still requires — will be the yardstick for both analysts' assumptions.

The stock's slide has run alongside this argument. Shares were down 8.2% on the day, though that move cannot be pinned solely on Bernstein's note; insider selling and broader skepticism toward debt-financed AI infrastructure have also weighed on the name.

Launchables Aim at the Friction Behind the Hype

While the analysts spar over the balance sheet, CoreWeave has been working the product side. At NVIDIA GTC Berlin on Tuesday, the company unveiled its Physical AI Launchables — tested templates designed to spare robotics and autonomous-driving teams the laborious work of configuring cloud environments, letting them fire up compute workloads with a single command.

The templates target teams running simulations, generating data, or training and evaluating models, and they also simplify GPU usage and resource-limit management. At launch, CoreWeave supports NVIDIA Isaac Sim, CARLA and MuJoCo, alongside NVIDIA AlpaSim and NVIDIA Cosmos 3. Developers can bring their own scenes and data, and a blank template is available for each tool, so applications are not confined to preset examples. Workloads can run either on reserved capacity or as short-notice compute jobs.

CoreWeave at a turning point? This analysis reveals what investors need to know now.

CoreWeave has also wired in Claude Code, Codex and other coding assistants through dedicated functions that can launch and monitor compute jobs and retrieve results — extending the simplification beyond setup into parts of the downstream workflow. A Cosmos Data Factory offering adds tools for augmenting, auto-labeling and curating data, including retrieval. Partners named include Cadence, Encord and PhysicsX, with more templates and workflows planned; the company gave no timeline for those additions.

For investors, the announcement is best read as an expansion of usability rather than a proven productivity gain. CoreWeave offered no general performance benchmark and no universal time saving for setup, so a one-command launch does not by itself translate into a blanket efficiency win. The real test is whether the combination of custom data, prepared environments and assistant-driven control delivers lasting relief for development teams — and whether the November quarter shows the backlog turning into revenue fast enough to justify whichever funding estimate proves closer to reality.

Ad

CoreWeave Stock: New Analysis - 22 October

Fresh CoreWeave information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.

Read our updated CoreWeave analysis...