Nvidia Revenue by Segment 2026: Where $70B Comes From

# Nvidia Revenue by Segment 2026: Where $70B Comes From

Nvidia doesn't make just one product. It makes chips for data centers, gaming PCs, professional workstations, and self-driving cars. But they're not equal. One segment dominates everything else. Understanding where Nvidia's $70B in annual revenue comes from is crucial to valuing the stock and predicting the next growth wave.

This article breaks down Nvidia's business by segment, with real SEC filing data, margin profiles, and what's driving growth heading into 2027.

The 4 Nvidia Business Segments

1. Data Center: The Cash Cow (75% of Revenue)

• H100 Tensor GPUs (2023-2024 peak demand)

• H200 Tensor GPUs (2024, higher bandwidth)

• L40S GPUs (inference workloads)

• Blackwell B100/B200 (2024-2025, next generation)

• Peak H100 pricing: $40,000 per GPU (wholesale)

• H100 lead time (2024): 6-12 months (constraint = shortage premium)

• H100 lead time (2025-2026): 4-8 weeks (normalized supply)

• Typical data center buys 10,000+ GPUs per order ($400M+)

• Hyperscalers: AWS, Microsoft (Azure), Google Cloud, Meta

• Each spends $1-5B annually on Nvidia chips for AI inference/training

• Concentration risk: Top 5 customers = 75% of Data Center revenue

• COGS per H100: ~$4,000-5,000 (manufacturing cost)

• Selling price: $35,000-40,000 (2024 peak)

• Gross margin: 87-88% (exceptional)

• As competition enters (AMD MI300X, Intel Gaudi): margin compresses toward 72-75%

• LLM inference scaling (more servers needed post-training)

• GenAI adoption (ChatGPT, Claude, Gemini training)

• Enterprise AI (Copilots, retrieval-augmented generation)

• Next-gen models (Blackwell adoption, 2025-2026)

---

2. Gaming: The Stable Performer (10% of Revenue)

• GeForce RTX 4090 (2022-2024)

• GeForce RTX 5090 (2025-2026, flagship)

• RTX 4070/4080 (mid-range, volume driver)

• RTX 5090 launch price: $1,999

• RTX 4090 used price (2026): $800-1,200

• Average gaming GPU selling price: $600-1,000

• Annual shipped units: 6-8 million GPUs

• Enthusiast gamers (high-end builds)

• Streamers/content creators (CUDA for rendering)

• ML hobbyists (Nvidia dominates because of CUDA software lock-in)

• COGS per RTX 5090: ~$700-800

• Selling price: $2,000 (at launch)

• Gross margin: 60% (much lower than Data Center)

• Used market cannibalization reduces new sales

• PC gaming cycle (GPU refreshes every 3-4 years)

• Ray tracing/DLSS adoption (tech differentiation)

• AI upsampling (DLSS 4, Frame Generation)

---

3. Professional Visualization: The Niche (8% of Revenue)

• RTX 6000 Ada (flagship workstation)

• RTX L40 (data center inference, double-counted in some segments)

• RTX 5880 Ada (VFX studios, architectural visualization)

• RTX 6000 Ada pricing: $6,800-9,000

• Annual workstation GPU units shipped: 500K-800K

• Concentration: 50%+ revenue from top 20 studios/enterprises

• VFX studios (Pixar, ILM, Disney, etc.)

• Architectural firms (Autodesk 3ds Max, Revit)

• Design houses (product rendering, automotive design)

• CAD workstations (aerospace, manufacturing)

• COGS: ~$1,200-1,500

• Selling price: $7,000-9,000

• Gross margin: 82-85% (higher than you'd think)

• Software bundling increases ASP (average selling price)

• 3D content explosion (streaming, VR, metaverse hype)

• Ray tracing in production rendering

• AI upsampling in post-production (faster renders)

---

4. Automotive: The Growth Bet (7% of Revenue)

• Drive Orin (2022-2026, in-car AI)

• Drive Hyperion (full stack: chip + software + sensor fusion)

• Drive Thor (next gen, 2026+)

• Drive Orin chip cost: $300-500 per chip

• Selling price to car makers: $800-1,200 per vehicle

• Annual vehicle ramp (2024-2026): 500K → 2M+ vehicles shipping Orin chips

• Gross vehicles equipped: ~30-50% of EV market (Tesla 100%, others ramping)

• Tesla (dominant customer, 60%+ of Automotive revenue)

• Lucid, Polestar (luxury EV makers)

• Chinese EV makers (BYD, Li Auto, others licensing tech)

• Traditional automakers ramping (Ford, GM planning for 2027+)

• COGS: $400-600 (higher than you'd think, custom integration)

• Selling price: $1,200 (OEM price, not retail)

• Gross margin: 50-55% (lower due to customization, support costs)

• EV adoption acceleration (50M EVs sold annually by 2026)

• Autonomous driving race (Level 3-4 features require Nvidia compute)

• Chinese market (BYD alone ships 1.5M+ EVs/year)

---

The Segment Reality: Data Center is Everything

What this means:

• Nvidia's P/E multiple (30-40x) is justified entirely by Data Center growth

• Remove Data Center? Stock might trade at 15-20x (like AMD, Intel)

• Automotive and Gaming are profitable but not growth engines

---

The AMD Comparison: Why Nvidia Crushes

See also: NVDA Stock Analysis | AMD Comparison

---

1. CUDA lock-in - 15 years of software. ML engineers code for Nvidia. Rewriting for AMD = $50M+ per customer

2. Hyperscaler preference - AWS/Azure/Google have massive CUDA codebases. AMD adoption requires 2-3 year software port

3. Gross margin - Nvidia's 75% margin vs AMD's 50% means Nvidia can outspend on R&D, sales, support

4. Supply chain - TSMC capacity allocation favors Nvidia (TSMC customer since 2005)

---

The Earnings Trend: What's Happening in Real Time

FY 2024 (ended Jan 2025)

• Data Center: $27B (+265% YoY)

• Gaming: $6B (-30% YoY, post-RTX 40-series cycle)

• Professional: $4.5B (+22% YoY)

• Automotive: $1.2B (+41% YoY)

• Total: $39.2B (full year)

FY 2025 H1 (ended July 2025)

• Data Center: $38B+ (annualizing $76B for full year!)

• Gaming: $3.5B (half-year, RTX 50-series ramp)

• Professional: $2.8B (half-year)

• Automotive: $2.5B (half-year)

• Total: $46.8B (half-year pace)

FY 2026 Trajectory (current run rate)

• Data Center: $52-55B (maturation; growth slowing from +200% to +50-80%)

• Gaming: $7-8B (RTX 50-series cycle)

• Professional: $5-6B (AI-driven visualization boom)

• Automotive: $4-5B (EV ramp)

• Total: $68-74B (annualized)

---

The Margin Story: Data Center Compression Ahead

Nvidia's gross margin has expanded from 60% (FY2023) to 70% (FY2025) because Data Center was supply-constrained. Prices were $40K per H100 (MSRP $30K). Now in 2026, supply is normalized:

• H100 street price: $28-32K (normalized from $40K peak)

• H200 launch pricing: $32-35K (better value, some cannibalization)

• Blackwell B100 pricing: $35-40K (replacement, not price reduction)

• Data Center gross margin: 72-75% (compression from 75-78% in peak shortage)

• Company gross margin: 68-72% (still healthy but decelerating)

If AMD gains 10% market share in Data Center, gross margin could drop to 65-68%. That would compress P/E multiple significantly (40x → 28x = 30% stock downside).

---

Valuation Implications: Is NVDA Fairly Valued?

• Data Center growth decelerates (150%+ → 50% by 2027)

• AMD captures 15% Data Center market share

• Hyperscalers build 20% of compute internally (TPU, Trainium)

• Gross margin compresses to 65%

• Forward P/E should be 20-25x, not 40x

• Downside target: $80-120 (vs $170 current)

• Data Center TAM expands faster than expected (enterprise AI adoption)

• Nvidia maintains 80%+ market share (CUDA moat holds)

• Gross margin stays 70%+ (Blackwell pricing power)

• 2027 earnings = $30 per share; at 40x P/E = $1,200 stock target by 2027

• Upside target: $250-350 by end 2027

• Data Center growth decelerates to +40-60% (still strong but not +200%)

• AMD gains 8-10% share; Nvidia holds 75-80%

• Gross margin = 68-70% (stable)

• 2027 EPS = $18-20; fair P/E = 28-35x = $504-700 stock target

• Current fair value: $150-180 (stock is fairly valued to slightly overvalued)

---

The Bottom Line: Know What You're Buying

The thesis:

• Will hyperscalers spend $500B+ annually on AI compute for the next 5 years? Yes.

• Can Nvidia keep 75%+ of that market? Probably, but pressured to 70-75% by 2027.

• Will Data Center gross margin stay above 70%? Likely, but compressed from 75%.

• 40x forward P/E (assuming $25 EPS in 2026)

• 150x cash flow (insane multiple by historical standards)

• Bet fully reflects Data Center TAM + CUDA moat + no competition

• Bull case (2027): $1,500-2,000 (+50-100% return)

• Base case (2027): $1,200-1,400 (+20-40% return)

• Bear case (2027): $600-800 (-40-60% return)

The risk/reward is skewed to downside if Data Center growth disappoints (which it will, eventually). But the upside is real if AI adoption exceeds expectations. This is a conviction play, not a safe dividend hold.

---

Next Steps: Track the Key Metrics

Watch these metrics in Nvidia's quarterly earnings calls and SEC filings (10-Q):

1. Data Center revenue growth rate (currently +200%, will decelerate to +50-80% in 2026)

2. Gross margin (key early warning of competitive pressure)

3. Days sales outstanding (DSO) (channel inventory buildup = red flag)

4. Customer concentration (top 5 = 75%; if drops, good sign of diversification)

For detailed segment data, check Nvidia's quarterly 10-Q filings under "Business Segment Results."

---

Want More Analysis on Nvidia & Semiconductors?

Explore Stock Market ROI to compare Nvidia (NVDA) revenue segments with AMD, Intel, and emerging GPU makers. Also read Best Banks for Investors 2026, Apple Services Revenue Analysis, and Top 10 Global Dividend Payers 2025 for broader portfolio context. Track Data Center ASP trends, gross margins, and competitive dynamics as the AI race unfolds.

Screen stocks and compare valuations with our Stock Screener.