CoreWeave 4Q25 Earnings Review: Profitability Will Only Gradually Recover in 2H 2026

$CoreWeave (CRWV.US)$'s 4Q25 print was strong on delivery against physical build-out and contract visibility, and the 2026 revenue guide is aggressive.

The problem is that the “cost of getting there” is now too large to treat as a footnote. The forward plan implies an even more capital-intensive 2026, leaving equity holders fearful of execution missteps and of further increases in interest expense from continued debt financing.

The Quarter in Numbers

Fourth-quarter revenue was $1.572 billion, in line with the company’s guidance.

On profitability, adjusted EBITDA was $898 million (57% margin) and adjusted operating income was $88 million.

Adjusted operating profit came in well below the midpoint of the company’s prior guidance of RMB 125 million, mainly due to a higher proportion of hardware procurement costs and continued elevated depreciation and interest expense.

Three Key Takeaways

An increasing number of SMB customers — a positive signal

The headline backlog is $66.8B as of Dec 31, 2025. The company said that, in addition to expanding its partnerships with hyperscalers, it has also added leading AI pioneers and enterprise customers, including: Cognition, CrowdStrike, Cursor, Mercado Libre, Midjourney, and Runway. This is a positive sign, as we noted in our outlook that the durability of core bookings from the mid-market and broader enterprise segment implies more stable and sustainable growth.

But the company is explicit that backlog recognition is subject to delivery and availability of service requirements. Operationally, CoreWeave added ~260 MW of active power in the quarter, ending with >850 MW, and expanded contracted power to ~3.1 GW.

2026 guidance reads like a stress test

The company’s 2026 guide makes the trade-off explicit. For 1Q26, CoreWeave guided to:

– Revenue $1.9–$2.0 billion

– Capital expenditures $6–$7 billion

– Interest expense $510–$590 million

– Adjusted operating income $0–$40 million

For FY26, it guided to:

– Revenue $12–$13 billion

– Capital expenditures $30–$35 billion

– Adjusted operating income $900–$1,100 million

That is an extreme reinvestment rate. It can work if execution is flawless and funding is abundant. It becomes fragile if either wobbles. The company also leaned on financing in the quarter, raising ~$2.6B in convertible notes and expanding its revolver to $2.5B.

The rapid increase in total debt will still push interest expense higher, but the marginally positive news is that management said on the earnings call that the cost of financing has fallen by 300 bps over the past 12 months and by 600 bps over the past two years.

In terms of impact on operating margin, management said that the low single-digit margin in Q1 will be the “low point” for the full year, with sequential improvement in Q2/Q3 and a return to low double digits in Q4. Overall, this is one notch below 2025 and is clearly below market expectations.

Long term — platformization and a “second growth curve” with higher gross margins

In the long run, CoreWeave needs to build on its highly sticky customer base by offering more third-party and higher value-added services, gradually moving toward a platform model akin to traditional hyperscalers to lift margins. The company is working in this direction, but meaningful financial improvement will likely take a long time.

Management stated on the earnings call:

About 80% of $1M+/year customers adopt at least one storage product; management also noted that cross-selling with Weights & Biases is driving “hundreds of millions of dollars” in TCV.

NVIDIA-related: NVIDIA plans to test/validate its platform (including software and reference architectures) and is exploring incorporating it into NVIDIA’s reference architecture; the company also noted that some customers have already made SKI their default cluster management platform.

The 2026 guidance does not include any potential revenue/profit upside from “licensing out its cloud stack” to external parties.

Bottom Line

To get less bearish, the company needs to show a few quarters where (a) active power ramps predictably, (b) incremental capex yields clean, on-time revenue conversion, and (c) the interest burden stops rising faster than operating earnings.

Until then, the print reads less like a compounding software story and more like a leveraged infrastructure build whose equity value is highly sensitive to financing conditions.

From a portfolio allocation perspective, as we mentioned in our Nebius commentary (Nebius Q4 Review: 2026 ARR Unchanged, Execution Still Key: Not the Best Time for NeoClouds Yet?), the best time to take a swing at the neo-clouds may not have arrived yet.

64K Views