CoreWeave Q2 Earnings: Who Owns the Stock Is the Wrong Question for August 11

CoreWeave reports Q2 earnings on August 11, 2026. Consensus expects revenue in the $2.45–2.60 billion range and a loss of roughly $1.07 per share. The press will frame this around who owns the stock. That is not the question that determines your allocation.

The question is whether a $50.8 billion debt load can finance a growth model that requires $16.6 billion in annual capital expenditures, as the cost of borrowing for AI infrastructure rises. The $99.4 billion revenue backlog — more than double the debt — is impressive. It also requires that backlog to convert to revenue faster than debt costs compound. That is the arithmetic that matters between now and the earnings call on Tuesday.

Here is the demand case, the architecture position, what NVIDIA's 11.5% stake actually tells us, and then the leverage signal that changes the risk/reward.

The demand case is not the debate

CoreWeave's Q1 2026 results, reported May 7, 2026, showed revenue of $2.078 billion — up 112% year-over-year from $982 million in Q1 2025. Revenue backlog was $99.4 billion as of March 31, the strongest bookings quarter in the company's history. The signed commitments include a $21 billion multi-year deal with Meta, a multi-year agreement with Anthropic, and expanded contracts with Cohere, Jane Street, and Mistral.

For Q2, the company guided to $2.45–2.60 billion in revenue. Oppenheimer expects results at the high end of that range, with 2026 full-year targets maintained. Consensus revenue estimates run from $1.57 billion in Q4 2025 to $2.55 billion in Q2 2026 — that is 163% growth over five quarters.

Revenue growth is accelerating. Sequential revenue growth was 32.2% in the trailing twelve months; gross profit growth is 114.7% year-over-year. Gross margin sits at 69.4%, which is exceptionally high for an infrastructure company — it reflects the software orchestration layer on top of the hardware, not just raw compute rental.

This is not a company struggling for customers. It is a company struggling to scale fast enough to meet commitments while the financing environment for that scaling deteriorates.

The architecture position: between models and silicon

CoreWeave's CEO Michael Intrator described the company as sitting "between the models and the silicon." That positioning matters because the AI infrastructure market is shifting from training-dominated to inference-dominated demand — and inference rewards companies that can deploy the latest generation of chips first, at scale, with purpose-built cooling and networking.

CoreWeave has a first-mover record on NVIDIA architecture generations. The company was the first AI cloud provider to offer NVIDIA HGX H100, H200, and GB200 NVL72 access. In July 2025, it announced the industry-first deployment of the NVIDIA GB300 NVL72 — a liquid-cooled rack integrating 72 Blackwell Ultra GPUs, 36 Grace CPUs, and 18 BlueField-3 DPUs. The platform delivers a 5x improvement in throughput per watt and 50x increase in output for reasoning model inference compared to the previous Hopper generation.

More importantly, CoreWeave's CEO noted that Blackwell provides the "lowest cost architecture for inference." That is the economic argument for why inference workloads will flow to operators who can deploy Blackwell-scale racks efficiently. And CoreWeave is already planning deployment of the Rubin platform, the next NVIDIA generation after Blackwell.

This architecture cadence is what separates CoreWeave from generic data center operators. The company's purpose-built AI data centers feature dense GPU infrastructure, advanced cooling, high-performance networking, and custom software orchestration through tools like SUNK (Slurm on Kubernetes) and CoreWeave Mission Control. Power infrastructure has expanded past 1 GW of active capacity, with total contracted power now exceeding 3.5 GW and a stated target of 8 GW by 2030.

But here is where NVIDIA's equity stake changes the analysis.

What NVIDIA's 11.5% stake tells us

In January 2026, NVIDIA invested $2 billion in CoreWeave Class A common stock at $87.20 per share, acquiring approximately 23 million additional shares and nearly doubling its stake to over 10%. Current institutional ownership data shows NVIDIA holding 47.2 million shares, representing 11.5% of outstanding shares.

This is not a passive investment. The partnership agreement includes CoreWeave's adoption of NVIDIA CPUs (Vera) and storage platforms (Bluefield), joint validation of CoreWeave's AI-native software for inclusion in NVIDIA's reference architectures, and collaboration on building more than 5 GW of AI factories by 2030. NVIDIA is leveraging its financial strength to accelerate CoreWeave's procurement of land, power, and shell infrastructure.

Put plainly: NVIDIA is effectively guaranteeing CoreWeave priority access to the next two generations of its chips. In an industry where GPU supply constraints determine who can scale, this is the equivalent of a long-term supply guarantee. The investment is both a strategic lock-in and a signal that NVIDIA sees CoreWeave as a distribution channel for its infrastructure stack beyond what the hyperscalers can absorb.

The other major shareholders tell a different story. Magnetar Financial — the largest institutional shareholder — held 23.3% as recently as October 2025 but has since trimmed by approximately 16.8%. CEO Michael Intrator retains an 11% stake; board member Brian Venturo holds 6.3%. Vanguard sits at 7.2%. Institutional ownership overall is approximately 61% of outstanding shares.

The early investors are taking some profits while the strategic investor — NVIDIA — is doubling down. That divergence is worth noting before the earnings call.

However: the leverage signal has changed

This is where the article shape needs to shift from "strong demand" to "strong demand with rising leverage risk."

CoreWeave's total debt stands at $50.8 billion against total equity of just $4.8 billion — a debt-to-equity ratio of 522%. Cash and equivalents are $2.2 billion, leaving substantial net leverage. Free cash flow is negative $10.6 billion on a trailing twelve-month basis, driven by $16.6 billion in capital expenditures that far exceed the $6.0 billion in operating cash flow.

The operating margin is -2.6% — the company is essentially breakeven on operations, which is remarkable given the scale. But EBITDA margin is 48.1%, and adjusted EBITDA was $1.157 billion in Q1 2026 at a 56% margin. The difference between EBITDA and operating income is entirely depreciation on the massive GPU and infrastructure base — which is a real cash cost that will need replacing every three to four years as architectures move forward.

The real concern is the tightening debt market. As of July 29, Bloomberg reported that CoreWeave had to sweeten terms on a $2.6 billion leveraged loan to attract investors, widening pricing by up to 1.25 percentage points to reach a yield of 9% or higher. The loan was offered at a 97-cent discount, down from 99 cents earlier, and CoreWeave amended documents to strengthen creditor protections. The cost of insuring CoreWeave's debt against default surged more than 50% in a single month, reaching its highest level since December.

The market language has shifted from strong demand for AI-related financings to weak demand and sky-high costs. CoreWeave is spending an estimated $34 billion on AI infrastructure in 2026, tapping the junk-debt market extensively. When lenders demand 9% yields and credit default swap costs spike, the cost of the leverage model that makes the backlog-to-revenue bridge possible is rising.

This is the dual signal I always watch for: surging supply commitments are evidence of demand strength and leverage risk. The $99.4 billion backlog proves demand. The 9% loan yields prove that the market has begun pricing in the risk that this demand may not convert to revenue quickly enough to service the debt.

How CoreWeave compares to its peers

A cross-section of AI infrastructure peers shows CoreWeave in a distinct position — the highest revenue growth, the best margins, but also the most aggressive leverage play.

Metric	CoreWeave (CRWV)	IREN	Nebius (NBIS)
Market Cap	$47.1B	$13.4B	$46.4B
Revenue Growth YoY	129.9%	—	—
Gross Margin	69.4%	—	—
Operating Margin	-2.6%	—	—
EV/EBITDA	23.3x	—	813.3x
P/S TTM	7.5x	17.7x	52.8x
Total Debt	$50.8B	—	Minimal

CoreWeave trades at 7.5x trailing revenue — substantially below Nebius's 52.8x and IREN's 17.7x. The valuation discount reflects the market's concern about debt sustainability, not a lack of demand. Nebius operates with no meaningful outstanding debt. IREN has a smaller scale but also a different risk profile tied to its crypto-mining heritage.

CoreWeave is the most operationally mature of this group — the highest gross margins, the best EBITDA margins, the deepest customer relationships, and the strongest NVIDIA integration. But it is also the one carrying the most financial risk.

The earnings call: what to listen for

On August 11, the market will hear whether Q2 revenue lands at the high end of the $2.45–2.60 billion range. But the numbers are secondary to three questions.

First, what is the company saying about financing conditions? The $8.5 billion DDTL 4.0 facility announced in Q1 was priced at SOFR + 2.25% for the floating tranche and approximately 5.9% fixed — investment-grade terms that looked strong in Q1. If management acknowledges that newer draws are being priced higher, that confirms the Bloomberg reporting on tightening conditions.

Second, what is the cadence of backlog-to-revenue conversion? A $99.4 billion backlog against $8.3 billion in trailing annual revenue implies the backlog represents roughly 12 years of current revenue run-rate. The actual conversion depends on delivery timelines — GPU supply, data center buildout, power availability. If management extends the conversion horizon, the debt service burden compounds during that bridge period.

Third, what is the capex trajectory for the back half of 2026? Q1 alone saw $7.7 billion in investing cash outflows, up from $1.4 billion in Q1 2025. If H2 capex follows a similar acceleration pattern, annual capex could approach $25–30 billion, widening the free cash flow deficit.

Where capital goes

I believe CoreWeave is on the right side of the training-to-inference transition. The company's architecture-first deployment record, NVIDIA equity partnership, and purpose-built infrastructure give it a structural advantage over general-purpose cloud providers for inference workloads. The $99.4 billion backlog is real demand, not pipeline optimism.

However, the debt-to-equity ratio of 522%, negative free cash flow of $10.6 billion, and the visible tightening in the AI infrastructure debt market change the risk profile. A 7.5x P/S multiple looks cheap relative to Nebius at 52.8x — but cheap leverage is still leverage, and the market is starting to price that distinction.

In my opinion, this is not a stock that warrants a large allocation at $86 per share and a $47 billion market cap. The return profile is real but back-half weighted — the debt model needs to prove it can sustain $16+ billion annual capex while interest costs rise and GPU replacement cycles loom. A smaller position — sized to the conviction in the demand story but protected against the leverage risk — is the right approach. If Q2 reinforces the backlog conversion timeline and management provides credible visibility into how the debt pile stabilizes, the thesis strengthens. If financing conditions continue to deteriorate and capex acceleration outpaces revenue conversion, the risk/reward tilts toward sidelines.

The debate is not whether CoreWeave remains important to the AI infrastructure stack. It is whether the capital structure supporting that importance is sustainable at current financing costs — and whether the return profile justifies the leverage risk relative to what can be found elsewhere in the AI trade.