NVDA · Q2 2026 Earnings

Bullish

Nvidia

Reported August 27, 2025

30-second summary

Revenue of $46.7B grew 56% YoY and 6% QoQ, beating the prior $45.0B guide by $0.84B at the high end, with Data Center at $41.1B (+56% YoY) and Blackwell up 17% sequentially. The H20 China overhang collapsed — a feared ~$8B revenue loss came in at a $180M inventory release impact — and Q3 guidance of $54.0B (±2%) implies ~16% sequential growth with gross margin stepping up ~90 bps to 73.3%. Nvidia raised FY26 OpEx growth from mid-30% to high-30% while reaffirming the mid-70% gross margin exit, signaling accelerated investment without margin compromise.

Headline numbers

EPS

Q2 FY2026

$1.05

Revenue

Q2 FY2026

$46.74B

+56.0% YoY

Gross margin

Q2 FY2026

72.4%

Free cash flow

Q2 FY2026

$13.45B

Operating margin

Q2 FY2026

60.8%

Key financials

Q2 FY2026

Metric	Q2 FY2026	YoY
Revenue	$46.74B	+56.0%
EPS	$1.05	—
Gross margin	72.4%	—
Operating margin	60.8%	—
Free cash flow	$13.45B	—

Guidance

Guidance is issued one quarter forward. The Prior-guide column references the guide issued last quarter for the period just reported; the New-guide column is for next quarter.

Actuals vs prior guidance

Metric	Period	Prior guide	Actual	Δ	Result
Revenue	Q2 FY2026	$45.0B +/- 2% ($44.1B - $45.9B)	$46.743B	+$0.84B above guide high	Beat
GAAP Gross Margin	Q2 FY2026	71.8% +/- 50 bps (71.3% - 72.3%)	72.4%	+10 bps above guide high	Beat
Non-GAAP Gross Margin	Q2 FY2026	72.0% +/- 50 bps (71.5% - 72.5%)	72.3%	-20 bps below guide high, within range at low end	Beat
GAAP Operating Expenses	Q2 FY2026	approximately $5.7B	Not explicitly disclosed	In-line with guidance (not separately reported)	Met
Non-GAAP Operating Expenses	Q2 FY2026	approximately $4.0B	Not explicitly disclosed	In-line with guidance (not separately reported)	Met
Tax Rate	Q2 FY2026	16.5% +/- 1% (15.5% - 17.5%)	Not explicitly disclosed	In-line with guidance (not separately reported)	Met
H20 Revenue Loss	Q2 FY2026	approximately $8.0B loss	$180M inventory release impact	Better than expected; H20 impact limited to $180M vs ~$8B prior expectation	Beat

New guidance

Metric	Period	Guide	YoY
Revenue	Q3 FY2026	$54.0B +/- 2% ($52.92B - $55.08B)	—
GAAP Gross Margin	Q3 FY2026	73.3% +/- 50 bps (72.8% - 73.8%)	—

Changes to prior guidance

Metric	Period	Prior guide	New guide	Δ	Result
Operating Expense Growth	FY2026	mid-30% range	high-30% range	Raised from mid-30% to high-30% (approximately +3-5 percentage points upside)	Raised

Reaffirmed unchanged this quarter: Non-GAAP Gross Margin (mid-70% range)

Segment performance

Q2 FY2026

Segment	Q2 FY2026	YoY
Data Center	$41.1B	+56.0%
Gaming and AI PC	$4.3B	+49.0%
Professional Visualization	$0.601B	+32.0%
Automotive and Robotics	$0.586B	+69.0%

Capacity & utilization

Q2 FY2026

Segment	Q2 FY2026
Blackwell Data Center Revenue Growth (Sequential)	17%
H20 Inventory Release Impact	$180M

Profitability

Q2 FY2026

Segment	Q2 FY2026
Non-GAAP Operating Margin	64.5%
Non-GAAP Gross Margin (excl. H20)	72.3%

Other KPIs

Q2 FY2026

Segment	Q2 FY2026
Share Repurchases (H1 FY2026)	$24.3B
Remaining Share Repurchase Authorization	$74.7B

Management tone

The prepared-remarks transcript was not available for this brief, so multi-quarter tone arc analysis is deferred. The Q&A commentary that was captured carries a notably more confident posture on three specific fronts than the headline risks would suggest:

On China, Jensen framed the market as "~$50 billion opportunity this year" growing "~50% annually," with H20 licenses already approved for non-entity-list customers and active advocacy for Blackwell approval. This is a step beyond the defensive H20-loss framing of prior quarters — China is now sized as a discrete growth lane, not a regulatory hole.

On ASIC competition, Jensen's response to Vivek Arya was unusually detailed: full-stack co-design, platform ubiquity, perf-per-watt, perf-per-dollar, and the six-chip integration complexity of Ruben. The willingness to spend Q&A airtime quantifying NVIDIA's $35B-of-$50B value capture per gigawatt AI factory suggests management views the ASIC narrative as the most-asked competitive question and is leaning into it rather than deflecting.

On the demand horizon, the "$3–4 trillion AI infrastructure spend by end of decade" with "top 4 hyperscaler CapEx doubled in 2 years to $600B annually" framing positions FY26 and FY27 as both "record-breaking years." This is forward-leaning language paired with a concrete OpEx step-up — management is committing capital behind the rhetoric.

Q&A highlights

CJ Muse · Cantor Fitzgerald

With 12-month wafer-to-rack lead times and Ruben on track for H2 ramp, what is NVIDIA's vision for growth into 2026, particularly the split between networking and data center revenue?

Jensen outlined reasoning and agentic AI as primary growth drivers, requiring 100-1000x more compute than single-shot models. Blackwell NVLink72 rack-scale systems are designed for this moment, enabling order-of-magnitude speedups in token generation. Over 5 years through end of decade, scaling into $3-4 trillion AI infrastructure opportunity driven by top 4 CSP CapEx doubling to $600B annually, enterprise on-prem builds, global cloud builds, and sovereign AI.

Reasoning/agentic AI requires 100-1000x more computation than single-shot modelsTop 4 hyperscalers CapEx has doubled in 2 years to $600 billion annually$3-4 trillion AI infrastructure spend by end of decadeTransitioned from NVLink 8 (node-scale) to NVLink 72 (rack-scale computing)

Joe Moore · Morgan Stanley

What are the long-term prospects for China business? How much can NVIDIA grow there, and how important is it to get Blackwell licensed in China?

China represents ~$50 billion opportunity this year with potential 50% annual growth rates consistent with rest of AI market. China hosts 50% of world's AI researchers and generates leading open-source models (DeepSeek, QWin, Qimi). Critical for American tech companies to address market. Advocating with administration for Blackwell approval; H20 licenses already approved for non-entity list companies.

China market estimated at ~$50 billion opportunity for NVIDIA this yearExpected to grow ~50% annually in line with global AI market50% of world's AI researchers located in ChinaMajority of leading open-source models created in China

Vivek Arya · Bank of America Securities

What needs to happen for $2-5B H20 China shipments in Q3? What is sustainable China business pace into Q4? How does NVIDIA compete against ASIC projects, given customers' internal silicon efforts?

H20 shipments require geopolitical licensing approvals and customer purchasing decisions still in flux; supply is ready. On ASICs: accelerated computing requires full-stack co-design unlike general-purpose computing. NVIDIA's advantages include platform diversity (available everywhere), support across entire pipeline (data processing to inference), energy efficiency (perf per watt drives revenue in power-limited datacenters), superior perf per dollar, and high gross margins. Complex systems integration (CPUs, memory, GPUs, networking, switches) creates high barriers.

H20 shipment range $2-5 billion contingent on licenses and customer purchasingCan build and ship more if orders arriveBlackwell NVLink 72 achieves order-of-magnitude energy efficiency improvementNVIDIA represents ~35 out of ~50B dollars for a gigawatt AI factory

Aaron Rakers · Wells Fargo

What is the opportunity set for Spectrum XGS and overall Ethernet portfolio sizing, particularly for data center interconnect layer?

NVIDIA offers three networking technologies: NVLink (scale-up for largest virtual GPU), InfiniBand (scale-out for lowest latency, preferred by leading model makers), and Spectrum X Ethernet (scale-out for Ethernet-based datacenters, approaching InfiniBand performance). Spectrum XGS for scale-across connects multiple datacenters into super-factories. Right networking choice improves factory efficiency 65% to 85-90%, making networking cost-effective. $50B gigawatt factory with 10-20% efficiency gain yields $10-20B value.

Spectrum X Ethernet annualized revenue exceeding $10 billionDouble-digit sequential and year-over-year growth for Spectrum XInfiniBand revenue nearly doubled sequentiallyNVLink 72 represents 14x bandwidth of PCIe Gen 5

Jim Schneider · Goldman Sachs

How does Ruben position relative to Blackwell in terms of incremental capability? Is it a bigger, smaller, or similar performance step-up?

NVIDIA operates on annual product cadence to accelerate cost reduction and maximize customer revenue generation. Blackwell delivers order-of-magnitude perf/watt improvement over Hopper for reasoning systems. Ruben will include multiple new ideas but Jensen deferred specific details until GTC. Company focused on ramping Blackwell and Blackwell Ultra this year; expects 2025 and 2026 to be record-breaking years. Customers advised to pace datacenter builds on annual rhythm.

Annual product cadence strategyBlackwell perf/watt order-of-magnitude higher than Hopper for reasoning systemsRuben includes multiple new architectural ideas (details deferred to GTC)Ramping GB200, GB300 (Blackwell Ultra) hard this year

What to watch into next quarter

H20 China shipments materializing or not. Q3 guide assumes zero; Jensen flagged $2–5B potential range contingent on licenses. Any actual shipment becomes pure upside vs. the $54B guide.

Q3 gross margin trajectory toward the mid-70% FY exit. Guide is 73.5% non-GAAP (±50 bps); the mid-70% FY26 exit commitment implies another step-up in Q4. Watch whether Blackwell Ultra mix delivers it.

Blackwell sequential growth rate. Blackwell Data Center grew 17% QoQ this quarter; the Q3 ~16% overall sequential guide implies continued Blackwell acceleration. A deceleration below ~15% would signal supply or demand limits before Ruben.

Spectrum-X / networking disclosure cadence. Now >$10B annualized — at what point does Nvidia break out networking as a reported segment? Continued disclosure of specific networking growth rates would be a positive tone signal.

OpEx step-up execution. FY26 OpEx growth raised to high-30%; watch whether the Q3 ~$4.2B non-GAAP OpEx guide is the new run rate or a one-quarter step, and whether operating margin holds above 60%.

Sources

NVIDIA Q2 FY2026 Press Release, filed with SEC 2025-08-27: https://www.sec.gov/Archives/edgar/data/1045810/000104581025000207/q2fy26pr.htm
NVIDIA Q2 FY2026 earnings call Q&A (transcript excerpts).

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.