CoreWeave Stock, September 2026: Revenue Doubled and $104B Backlog Underpin 11.7% Rally

CoreWeave closed at $99.83 on September 9, 2026, up 11.7%, on the back of Q2 revenue of $2.58 billion (+112% year-over-year) and a $104 billion contracted backlog, though capital intensity at 201% of FY2025 revenue and emerging competition from hyperscaler in-house clouds remain structural headwinds.

CoreWeave shares closed at $99.83 on September 9, 2026, up 11.7% on the day, extending a rally that has defined the neocloud sector through a volatile summer. The move comes the day after the company filed its 10-Q with the SEC, and roughly three weeks after the broader post-earnings neocloud rerating began in August. The central question driving both the gains and the periodic selloffs remains the same: whether demand for third-party GPU infrastructure is durable enough to justify the capital structures that neoclouds are building — and CoreWeave, as the largest independent GPU cloud by contracted capacity, sits at the center of that debate.

The Q2 2026 results are difficult to argue with on the surface. Revenue reached $2.58 billion, a 112% year-over-year gain, while the contracted backlog surged 246% to approximately $104 billion — a figure representing years of locked-in cashflow at current capacity levels. Management raised full-year guidance to $12.4–$13.2 billion, implying the growth rate does not materially decelerate in the second half. Reports at the time of results cited major hyperscaler commitments as underpinning the backlog, though precise counterparty terms remain the company's proprietary disclosure and have not been independently verified. The market's initial reaction to the August earnings release was immediate; the September 9 session signals that investors are still in the process of rerating the sector's long-run value.

What makes CoreWeave structurally distinct — and structurally contentious — is the scale of capital it must continuously deploy to sustain that position. Slicast's own compilation of SEC XBRL filings shows that in FY2025, CoreWeave spent $10.31 billion on property, plant, and equipment, up 18.5% from $8.70 billion in FY2024, against FY2025 revenues of $5.13 billion — a capex intensity of 201%. Among the seven neoclouds tracked by Slicast, CoreWeave ranks first in absolute capital expenditure; Nebius leads on intensity ratio. The highest capex recorded in any single reported period was $7.70 billion for the period ending March 31, 2026. To fund this expansion, CoreWeave closed a $2.6 billion delayed-draw term loan in August, secured against existing AI and HPC customer contracts. On the hardware frontier, Dell delivered the world's first NVIDIA Vera Rubin NVL72 server rack to CoreWeave on August 30 — with NVIDIA confirming that Vera Rubin is now in full production — establishing CoreWeave's early position in the next GPU generation. Hudson River Trading's decision to use CoreWeave for its quantitative research platform adds an enterprise validation point, though a single financial-sector deployment should not be treated as a sector-wide signal.

The risk ledger is not negligible, and the summer's volatility made that clear. In early August, when Meta disclosed plans for an internal compute cloud — referred to publicly as Meta Compute — CoreWeave fell roughly 12% in a single session, demonstrating how quickly the market reprices neocloud equities when a hyperscaler signals intent to internalize capacity. A rate-sensitivity selloff on August 21 drove CoreWeave down approximately 7% alongside Nebius and TeraWulf — a structural reminder that a company spending 201% of revenues on capital equipment is acutely exposed to any shift in financing costs. The lock-up expiry on August 14 added a technical overhang at precisely the moment post-earnings insider selling pressure was most elevated. On the competitive front, Lambda's B200 spot pricing at $6.69 per GPU-hour points to commoditization in uncontracted capacity, even if CoreWeave's long-dated contractual book provides partial insulation; at least one analyst note also cited IREN as outpacing CoreWeave on a specific operational efficiency metric, a claim not yet broadly tested by the market.

Piper Sandler, initiating coverage in early August with an Overweight rating and a $151 target, argued that CoreWeave's valuation discount did not adequately reflect its lead in deployed data center capacity — a view echoed by a September 3 analysis. A separate research note, published after NVIDIA confirmed Vera Rubin's full-production ramp, set a $153 upside target against a then-prevailing price of approximately $94. At $99.83, CoreWeave has moved toward but not reached either threshold, leaving room for a constructive view — and scope for disappointment if backlog conversion slows or capital markets tighten.

Three concrete signals are worth monitoring as investors assess the durability of the current level. First, whether CoreWeave's Indonesian expansion — three data centers totaling 360 MW, targeted for deployment before 2028 — translates from announced capacity to contracted revenue; APAC diversification would reduce the U.S. hyperscaler concentration embedded in the backlog. Second, whether Vera Rubin-era contracts price at meaningful premiums over H100-era rates; the company has reportedly extended A100 contracts into 2029, demonstrating that legacy GPU capacity remains profitable, but the equity multiple depends on the assumption that next-generation throughput sustains a durable pricing premium. Third, the 10-Q filed on September 8 warrants close reading for any changes in revenue recognition timing, lease structure, or disclosures that affect the backlog's realizability. The September 9 rally is a statement of investor confidence. Whether the fundamentals support it at these levels is a question the next two quarterly reports will answer.