CoreWeave (NASDAQ: CRWV) trades at $91.90 against a Wall Street consensus price target of $138.03, leaving roughly 50% between current levels and analyst expectations.

The AI infrastructure company rents specialized GPU hardware to AI labs and hyperscalers, and since its March 2025 IPO has functioned as the market’s purest proxy for whether the AI capital expenditure cycle can sustain itself.

Shares have fallen 33.4% over roughly three months, ranking among the worst performers in AI infrastructure, with the selloff triggered by the May 7 Q1 2026 earnings release when the stock sat at $137.98.

Revenue came in at $2.08 billion, up 111.7% year over year and beating consensus by 5.8%, but the bottom line told a different story entirely.

Net loss widened to $740 million from $315 million a year earlier, EPS missed at -$1.40 versus -$1.23 expected, interest expense doubled to $536 million, and capital expenditure reached $7.70 billion in a single quarter.

Free cash flow ran to -$4.71 billion while total liabilities climbed to $50.8 billion, giving investors serious pause about the pace of the company’s balance sheet expansion.

Pressure compounded through summer as a securities fraud class action was filed over data center construction delays, the CEO sold roughly $100 million in stock under Rule 10b5-1 plans, and a July selloff followed Meta’s announcement of its own “Meta Compute” cloud service.

Piper Sandler’s Harsh Kumar maintains a $151 price target on CoreWeave, implying approximately 64% upside from current levels, anchoring his thesis on tier-one allocation of NVIDIA’s Blackwell and Vera Rubin architectures, long-term GPU-as-a-Service backlog, and EBITDA margin expansion as clusters approach high utilization.

The revenue backlog reached $99.4 billion in Q1, up nearly 50% sequentially, and includes a $21 billion Meta commitment, an Anthropic multi-year deal for Claude, and expansions with OpenAI, Jane Street, and Hudson River Trading.

Management reaffirmed full-year 2026 revenue guidance of $12 billion to $13 billion and pointed to a 2027 annualized run rate of $30 billion or more, with over 75% already contracted under existing agreements.

Adjusted EBITDA hit $1.2 billion on a 56% margin in Q1, and CFO Nitin Agrawal characterized the quarter as the trough of the margin story, suggesting improvement ahead.

Of analysts tracked, 26 rate CoreWeave a Buy, 9 rate it Hold, and 2 rate it Sell, while Rosenblatt’s John McPeake holds the most bullish position on the street with a $250 price target.

Peers Nebius Group (NASDAQ: NBIS), IREN (NASDAQ: IREN), and Applied Digital (NASDAQ: APLD) all rallied during the same window in which CoreWeave fell, framing the decline as company-specific rather than sector-wide.

Nebius trades near $225.74 and is up 169.69% year to date after posting 279.6% revenue growth and signing its own $27 billion Meta agreement, with a consensus target of $258.13 implying around 14% further upside.

IREN sits at $40.85 against a consensus target of $81.73, implying roughly 100% upside, after securing a $3.4 billion five-year AI Cloud contract with NVIDIA plus up to $2.1 billion in NVIDIA investment.

Applied Digital trades at $31.27 versus a $74.14 consensus target, representing the widest implied upside gap in the peer group at approximately 137%, and CoreWeave is Applied Digital’s principal HPC hosting customer, linking the two companies directly.

CoreWeave is down 13.31% over the past year but remains up 28.33% year to date, still outpacing the S&P 500’s 13.11% return over the same window.

The bull case hinges on the August 11 Q2 earnings release delivering the sequential margin inflection management guided toward, the $99.4 billion backlog converting on schedule, and the securities lawsuit resolving without material dilution.

The bear case strengthens if interest expense creeps above the $650 million to $730 million guidance range, customer deployments slip, or Meta Compute pulls meaningful workload from the platform, risks that a $50.8 billion liability stack and -$4.71 billion free cash flow leave little room to absorb.