CoreWeave: Orders Soar, Profits Plunge — Is the Compute Unicorn Stuck in a Profitability Trap? ---

AI cloud unicorn $Coreweave(CRWV.US) posted its Q4 2025 results after hours today (Feb 27). Overall, revenue and scale kept growing well, but profits fell more than even the lowered guide. This raises the risk that higher revenue could deepen losses for a while, adding to already-elevated concerns about medium- to long-term profitability.

1) Capacity ramp on track: After cutting Q4 revenue guidance last quarter due to supplier delays, this quarter delivered revenue of ~15.7bn (+110% YoY), meeting and slightly beating expectations. The delay issues appear largely resolved, with no further spillover impact. Market focus remains on deployed compute. Active Power reached 850MW this quarter, broadly in line with top-tier sell-side estimates. QoQ net adds were 260MW, or 2.2x last quarter's increase, signaling the ramp has clearly entered a faster phase.

2) RPO beat; customer mix improving? The leading indicator RPO came in at 66.8bn, well above Bloomberg consensus of 62bn. With no headlines of new mega contracts, the Street had expected limited additions this quarter. In reality, RPO still rose by nearly 17bn QoQ, not far off last quarter's 20bn peak. Dolphin Research suspects the increase came from numerous non-mega clients, and management confirmed the number of new customers doubled vs. the prior peak, including Cognition, Cursor, Mercado Libre, Midjourney and Runway. This is encouraging. It suggests the prior overreliance on a handful of top customers is being addressed.

3) Profit contraction worse than feared: The market already expected margin pressure during the rapid ramp and had trimmed Adj. OPM to ~8%. Actual Adj. OPM was just 5.6%, a bigger miss than feared, and despite revenue doubling, Adj. OP fell ~27%. The main driver was an underlying GPM (COGS plus Tech & Infra expensed) of only 7.2%, plunging from 18% last quarter and far below estimates. S&M as a percentage of revenue fell ~160bps QoQ, indicating margin compression was primarily due to higher energy costs, depreciation and upfront ramp spending.

4) Capex accelerating faster: Capex was 8.2bn, broadly in line and slightly above market expectations. Dolphin Research notes Capex was 3x–4x the 2–3bn levels of prior quarters, while new deployed compute rose by just 2.2x QoQ. In other words, the capacity coming online is ramping more slowly than Capex. While not unreasonable given installation and testing lead times, investment running ahead of monetized capacity will keep margins under pressure.

5) Debt load keeps rising: With Capex stepping up, total liabilities increased by ~7.5bn QoQ. By end-Q4, most remained on balance sheet as cash equivalents, so net interest-bearing debt rose by ~3.4bn. Even so, interest expense as a percentage of revenue rose from 23% to 25%. Debt pressure will likely intensify ahead.

Dolphin Research view:

1) Near-term results show a mixed picture. On the positive side, compute deployment met plan, earlier delays did not worsen, and revenue growth hit targets. New contracts were above expectations and more diversified, easing prior customer concentration risks. On the negative side, during the rapid ramp, more business is translating into deeper losses. The lag from investment to deployed capacity means Capex is rising much faster than new capacity coming online, driving a sharp margin decline and pushing GAAP profitability further out.

2) Guidance: near term, the company expects next quarter revenue of 1.9–2.0bn, up 0.4–0.5bn vs. this quarter and still accelerating. However, it is clearly below the Street’s ~2.24bn. The company guides next quarter Capex at 6.0–7.0bn, also below the Street’s ~7.1bn. The likely culprit behind the revenue shortfall is slower-than-expected deployment progress. On profits, next quarter Adj. OP is guided to 0–40mn, further down vs. this quarter and below the ~140mn consensus. That implies more revenue but less profit.

For medium to long term, 2026 revenue is guided at 12.0–13.0bn, broadly in line to slightly above current market expectations. This suggests the slower pace next quarter should be temporary. Adj. OP for 2026 is guided at 0.9–1.1bn, implying Adj. OPM of ~7.5%–8.5%, with margins improving meaningfully in 2H, broadly matching expectations.

Another worrying point: 2026 Capex is guided at 30–35bn, nearly 3x that year’s revenue. Put differently, 2026 revenue is only a bit over 2x Q4 annualized, while Capex is close to ~4x Q4 annualized. Investment growth still far outpaces revenue growth.

3) For upcoming guidance, first next quarter: GP is guided at 2.8bn (+22% YoY), maintaining solid growth and ~3% above expectations. Adj. OP is guided at 0.6bn, also above the ~0.575bn consensus. The implied OPM is roughly in line with expectations, with outperformance driven by stronger GP growth. Next quarter’s performance should remain upbeat. For long-term guidance, 2026 full-year GP is expected to grow 18% to 12.2bn, with Adj. OP around 3.2bn and OPM at 26%, up from 20.5% this quarter. Versus prior guidance and market expectations, the main change is a higher Adj. OP (from ~2.7bn previously).

4) Investment view: as discussed in our deep-dive, revenue is compounding rapidly while margins are declining, making the steady-state hard to handicap. In other words, it is difficult to see where margins stabilize. Management indicates by end-2027, annualized revenue could exceed 30bn and steady-state Adj. OPM could reach 25%–30%. At the midpoint, 2027 Adj. OP would be ~8.25bn, implying a sub-7x PE on post-ramp earnings — very cheap. The issue, as noted, is that this is management’s own case, and we lack conviction on achieving those margins. The biz. model is not great, and debt-driven growth carries high uncertainty and risk; our stance is to stay on the sidelines.

Risk disclosures and statement:Dolphin Research Disclaimer and General Disclosure