NVDA

Q2 2026

Date: 2025-08-27Rank: #9Forward Promise: very_bullish

NVIDIA reported record revenue of $46.7 billion for Q2 FY2026, driven by a 56% year-over-year increase in Data Center revenue to $39.6 billion. The Blackwell platform reached record levels, growing 17% sequentially, while Networking revenue surged 98% year-over-year to $7.3 billion. Despite a $4 billion decline in H20 revenue due to geopolitical factors, the company exceeded its outlook, with non-GAAP gross margins at 72.7%. Looking ahead, NVIDIA guided for Q3 revenue of $54 billion (+/- 2%), implying over $7 billion in sequential growth, with non-GAAP gross margins expected to rise to 73.5%. Management highlighted the transition to the GB300 rack-based architecture, which is now in full production at a rate of 1,000 racks per week, and confirmed that the Rubin platform remains on schedule for volume production next year.

Bullishness Score

96.05

μ Mean

101.74

σ Uncertainty

1.90

Forward Promise

8.8

Management Tone

Management exhibited extremely high confidence and assertiveness throughout the call, particularly during the Q&A session. Jensen Huang was emphatic about the inevitability of the AI revolution, using phrases like 'The AI race is on' and 'A new industrial revolution has started.' There was no shift in tone between prepared remarks and Q&A; the defensiveness often seen regarding China was replaced by a clear, albeit conditional, explanation of the H20 opportunity and a strong advocacy for US technology leadership.

Confidence: HIGH — Management used definitive language regarding future growth ('$3 to $4 trillion in AI infrastructure spend'), provided specific production metrics ('1,000 racks per week'), and dismissed competitive threats (ASICs) with detailed technical rebuttals.

Strategic Signals

Management emphasized the 'seamless' transition to the GB300 rack-based architecture, noting that factory builds were successfully converted in late July/early August. This signals that the feared production bottlenecks associated with the new Blackwell form factor are largely resolved, with production now at 1,000 racks per week and accelerating. This operational execution supports the thesis that NVIDIA can sustain hyper-growth through complex product transitions.

The strategic pivot to 'Reasoning Agentic AI' was highlighted as the primary driver for the next wave of compute demand. Huang explained that 'long thinking' AI models require 100x to 1000x more computation than previous 'one-shot' chatbots. This frames the current demand surge not as a bubble, but as an early stage in a structural upgrade cycle where NVIDIA's NVLink 72 architecture provides a necessary efficiency advantage.

NVIDIA is positioning itself as a holistic 'AI Infrastructure Company' rather than just a chip vendor. By detailing the six chips required for the Rubin platform (CPU, GPU, NIC, Switch, etc.), management reinforced the moat around its full-stack solution. This complicates the competitive narrative for ASICs, as Huang argued that while many projects start, very few reach production due to the extreme complexity of full-stack co-design.

The company is aggressively expanding its networking portfolio with the introduction of Spectrum XGS for 'scale-across' connectivity, linking multiple data centers into 'super factories.' With Spectrum X revenue now exceeding $10 billion annualized, networking is becoming a significant growth driver and a strategic lever to increase the overall efficiency and revenue generation of customer AI factories.

Sovereign AI and physical AI/robotics were identified as major new verticals. Management cited a $20 billion opportunity in Sovereign AI for the year and noted that the 'Thor' robotics platform is seeing rapid adoption. This diversification reduces reliance on the hyperscalers and opens up new TAMs as enterprises and nations build their own AI infrastructure.

Key Metrics

Total Revenue$46.7B+Sequential Growth across all platforms

Data Center Revenue$39.6B (Derived)+56% YoY

Networking Revenue$7.3B+98% YoY

Gaming Revenue$4.3B+49% YoY

Non-GAAP Gross Margin72.7%+Exceeding outlook

Q3 Revenue Guidance$54B (+/- 2%)+$7B+ Sequential

GB300 Production Rate~1,000 racks/weekAccelerating

China Data Center RevenueLow single digits %-Sequential Decline

Guidance

Q3 Total Revenue: $54 billion (+/- 2%)

Q3 Non-GAAP Gross Margin: 73.5% (+/- 50 bps)

Q3 China H20 Revenue: Not included in outlook (Potential $2-5B if licenses clear)

FY Operating Expense Growth: High thirties % YoY (Increased from mid-thirties)

FY Non-GAAP Gross Margin: Mid-seventies % by year end

Bottom Line

Price Target: N/A

NVIDIA's Q2 FY2026 results and conference call reinforce its position as the undisputed leader in AI infrastructure. The company successfully navigated the complex transition to the GB300 rack-scale architecture, with production now accelerating to 1,000 racks per week, effectively silencing concerns about Blackwell yield or deployment issues. Financial performance remains stellar, with $46.7B in revenue and expanding gross margins, driven by a product mix shift toward high-end networking and compute systems. Strategically, NVIDIA is expanding its moat by moving up the stack from individual GPUs to full AI factories, integrating NVLink, Spectrum-X networking, and software into a solution that competitors cannot easily replicate. The shift toward 'Reasoning Agentic AI' serves as a powerful demand driver, as these models require orders of magnitude more compute, validating the massive CapEx spending by hyperscalers. While geopolitical risks regarding China persist, management quantified the opportunity at over $50B, suggesting significant upside potential if regulations stabilize. With the Rubin platform already in fab and the company guiding for strong sequential growth, NVIDIA remains the best way to play the long-term AI infrastructure build-out.

Forward Promise

Overall: 8.8 · very_bullish

Unapologetically bullish: 'Everything sold out' and 'AI race is on'

Management displayed supreme confidence in the duration and magnitude of the AI cycle, backed by specific production metrics and a massive TAM expansion.

Red Flags

Geopolitical friction regarding China sales remains a significant overhang on the stock. While H20 licenses have been granted, management explicitly excluded H20 revenue from the Q3 guidance, citing unresolved 'geopolitical issues' and a lack of codified USG regulations. This creates a binary risk factor where $2-5 billion in quarterly revenue is currently contingent on regulatory clarity.

The guidance for operating expenses to grow in the 'high thirties' year-over-year, up from 'mid-thirties,' signals a significant step-up in investment spend. While necessary to support the Blackwell and Rubin ramps, this rapid expense growth could pressure operating margins if revenue growth decelerates or if the supply chain ramp encounters further hiccups.

Inventory levels increased sequentially from $11 billion to $15 billion to support the Blackwell ramp. While this indicates preparation for future demand, it also introduces execution risk. If the transition to GB300 or the subsequent Rubin ramp faces delays, or if demand softens, NVIDIA could be exposed to inventory obsolescence risks, particularly given the rapid pace of product innovation.

The complexity of the new GB300 rack-scale architecture introduces a new potential point of failure. While management claims the transition has been seamless, the shift from node-scale (NVLink 8) to rack-scale (NVLink 72) represents a massive engineering and logistical challenge. Any issues in thermal management, power delivery, or software compatibility at the rack level could slow down deployment rates.

Management's reliance on the 'everything is sold out' narrative to describe current demand creates a risk that any sign of inventory availability or order push-outs could be misinterpreted by the market as a demand collapse. The 'sold out' status is currently a positive, but it sets an impossibly high bar for maintaining the perception of growth.

Macro Insights

Global CapEx: Global AI infrastructure CapEx has doubled in two years to $600B annually among the top 4 hyperscalers, with a projected $3-4T total spend by the end of the decade.

Geopolitics: US-China tech tensions continue to disrupt revenue recognition, with H20 shipments paused pending regulatory clarity, though licenses are technically approved.

Energy/Power: Power limitations in data centers are driving demand for NVIDIA's high-efficiency Blackwell chips, as 'perf per watt drives directly to revenues' for customers.

Sovereign AI: Nations and regions (EU, UK, Japan) are investing heavily in domestic AI infrastructure, creating a 'Sovereign AI' market vertical that is on track to exceed $20B for NVIDIA this year.