NVIDIA Q4 FY2026 Earnings Play
NVIDIA Q4 FY2026 Earnings: Supply Chain Synthesis & Assessment
Report Date: February 14, 2026 | Earnings Date: February 25, 2026 (After Market Close)
Executive Summary
Based on a comprehensive analysis of earnings reports from NVIDIA’s upstream suppliers, downstream customers, and ecosystem competitors, the evidence overwhelmingly points to NVIDIA beating its $65B ± 2% Q4 revenue guidance, likely reporting in the $67–68B range. However, the stock’s reaction will hinge less on the Q4 beat (largely priced in) and more on FQ1 2027 guidance, gross margin trajectory, and Vera Rubin/Blackwell Ultra commentary. The stock currently trades at ~$183, well below its 52-week high of $212, suggesting the market is already cautious — creating asymmetric upside if guidance is strong.
1. Upstream Suppliers: All Signals Green
TSMC (Primary Foundry)
- Q4 2025 results: Revenue $33.7B (+25.5% YoY), beat consensus. Record net income of NT$505.7B (+35% YoY).
- January 2026 revenue: NT$401.3B — up 36.8% YoY and 19.8% MoM, a massive acceleration indicating robust wafer demand into CY Q1.
- AI accelerator revenue reached high-teens % of total in 2025; expected to grow at mid-to-high 30s CAGR through 2029.
- 2026 capex guidance: $52–56B (up ~37%), with 70–80% going to advanced technologies. Guided Q1 2026 revenue of $34.6–35.8B with gross margins of 63–65%.
- Key read-through: TSMC’s January revenue surge and expanding advanced packaging capacity (targeting 120K–130K wafers/month by late 2026, up from ~75K exiting 2025) confirms NVIDIA’s Blackwell production is scaling at or above plan. NVIDIA reportedly booked 50%+ of TSMC’s total advanced packaging (CoWoS) output.
SK Hynix (HBM Supplier)
- Q4 2025: Record revenue of KRW 32.8T (+66% YoY), record operating profit of KRW 19.2T (+137% YoY), 58% operating margin.
- Full year 2025: Revenue KRW 97.1T (+47% YoY), operating profit KRW 47.2T (+101% YoY). HBM revenue more than doubled YoY.
- HBM4 mass production already underway. SK Hynix secured ~two-thirds of NVIDIA’s HBM4 orders for Vera Rubin.
- Key read-through: Record HBM shipments directly correlate with Blackwell GPU production volumes. The doubling of HBM revenue confirms NVIDIA was shipping at full capacity, supporting a Q4 revenue beat.
Samsung (HBM/Memory)
- Q4 2025: Record revenue of KRW 93.8T, record operating profit of KRW 20.1T (+209% YoY). Memory division at KRW 44T revenue.
- Finalized HBM4 development at 11.7 Gbps; mass HBM4 shipments to NVIDIA started February 2026.
- Samsung reportedly delivered late-stage HBM3E to NVIDIA, and also supplied Google and AMD.
- Key read-through: Both major HBM suppliers had record quarters, confirming memory supply was being consumed at record rates — consistent with NVIDIA’s Blackwell ramp.
2. Downstream Customers: Hyperscalers Spending at Unprecedented Levels
The four major hyperscalers — NVIDIA’s largest customers — have collectively signaled ~$675–720B in 2026 capex, the most aggressive AI infrastructure spending cycle in history.
Microsoft (Largest NVIDIA Customer)
- Q2 FY2026 (Jan 28, 2026): Revenue $81.3B (+17%), Azure grew 39%. Capex hit $37.5B in the quarter (+66% YoY).
- Commercial backlog reached $368B of contracted business; RPO at $625B (+110% YoY).
- Still capacity-constrained through H1 FY2026, meaning GPU demand > supply.
- 2026 signal: Capex growth rate to slow vs FY2025 but absolute spend continues rising. Over $30B guided for next quarter.
Alphabet/Google
- Q4 2025 (Feb 4, 2026): Revenue $113.8B (+18%), Google Cloud $17.7B (+48% YoY). Cloud operating margin expanded to 30%.
- 2026 capex guidance: $175–185B — nearly double the $91.4B spent in 2025 and ~$60B above Street expectations.
- 60% allocated to servers (GPUs, TPUs), 40% to data centers/networking. Named NVIDIA as a key partner for Vera Rubin.
- Gemini app at 750M MAUs; 10B+ tokens/minute processed.
Amazon/AWS
- Q4 2025 (Feb 5, 2026): Revenue $213.4B (+14%), AWS $35.6B (+24% YoY) — fastest growth in 13 quarters.
- 2026 capex guidance: ~$200B, up from ~$131B in 2025 and far above the ~$147B Street estimate.
- CEO Jassy: spending will “predominantly” go to AWS. Demand exceeds supply; monetizing capacity as fast as they can install it.
Meta
- Q4 2025 (Jan 29, 2026): Revenue ~$59.9B, EPS $8.88, both beat. Ad impressions +18% YoY.
- 2026 capex guidance: $115–135B, roughly doubling from 2025. R&D alone was $17.1B in Q4.
- Building out massive data center footprint; launched “Meta Compute” to secure capacity.
Combined Hyperscaler Signal
Company 2025 Capex 2026 Capex Guide YoY Growth Microsoft ~$89B (FY) ~$120–130B est. +35–45% Alphabet $91.4B $175–185B +91–102% Amazon ~$131B ~$200B +53% Meta
$65B $115–135B +77–108% Total **$376B** $610–650B +62–73% This is the single most bullish signal for NVIDIA. The majority of this capex goes to GPUs, networking, and servers — NVIDIA’s core products. All four remain capacity-constrained and are competing to buy GPU capacity faster than supply can expand.
3. Ecosystem Competitors: Confirming the Boom, Not Threatening Share
AMD
- Q4 FY2025: Record revenue $10.3B (+34% YoY). Data Center segment $5.4B (+39% YoY).
- MI350 ramping; MI450/Helios rack-scale launching H2 2026. Targeting “tens of billions” in AI revenue by 2027.
- Guided Q1 around $9.8B — softer, suggesting NVIDIA’s dominance persists near-term.
- Read-through: AMD’s growth confirms AI demand is broad-based and not zero-sum, but its data center revenue is ~10% of NVIDIA’s, showing NVIDIA retains 85–90% accelerator share.
Broadcom (Custom AI Chips / Networking)
- Q4 FY2025: Revenue $18.0B (+28%). AI semiconductor revenue $6.5B (+74% YoY). AI backlog of $73B.
- Guided Q1 AI revenue of $8.2B (doubling YoY). Now has 5 XPU customers.
- Read-through: Broadcom’s custom chip growth is real but complementary to NVIDIA, not substitutive. Hyperscalers are buying both NVIDIA GPUs and custom ASICs. The $73B backlog confirms multi-year demand durability.
4. Networking & Infrastructure Partners: Confirming Full-Stack Adoption
Arista Networks (Ethernet AI Networking)
- Q4 2025 (Feb 12, 2026): Revenue $2.49B (+28.9% YoY), beat guidance of $2.3–2.4B. Record $9B full-year revenue.
- Surpassed $1B in quarterly net income for the first time. Non-GAAP operating margin 47.5%.
- AI networking revenue hit $1.5B in 2025; now guiding $3.25B in AI networking for 2026 (doubled from $1.5B), up from prior $2.75B target.
- International growth (Asia + Europe) grew north of 40% annually.
- CEO Ullal expects to double AI networking revenue from 2025 to 2026, driven by Etherlink adoption and 1.6T switching.
- Supply-chain constraints around memory cited as worsening — this applies broadly to the AI ecosystem.
- Key read-through: Arista’s AI networking acceleration directly correlates with GPU cluster deployments. If Arista is seeing 2x AI networking growth, NVIDIA’s GPU shipments must be supporting that throughput. Also confirms that customers aren’t just buying GPUs — they’re buying full fabric, which supports NVIDIA’s own NVLink/InfiniBand/Spectrum-X networking revenue growth (162% YoY in Q3).
Broadcom (Ethernet Switches, Custom ASICs, Optics)
- AI networking backlog exceeds $10B for Tomahawk switches alone; total AI order backlog of $73B.
- AI semiconductor revenue guided to $8.2B in Q1 FY2026 (doubling YoY).
- Now has 5 XPU customers (custom ASIC), including Google TPU and OpenAI.
- Read-through: Broadcom’s AI growth is complementary to NVIDIA, not substitutive. Hyperscalers are multi-sourcing, buying both NVIDIA GPUs and custom ASICs from Broadcom. The $73B backlog confirms the multi-year demand pipeline is real.
5. Sovereign AI & International Customers: The Emerging Growth Vector
This is an underappreciated tailwind that Wall Street may be under-modeling. NVIDIA has become the de facto “arms dealer” for national AI strategies:
Key Sovereign Deployments
- South Korea: Government + Samsung, SK Group, Hyundai Motor deploying 260,000+ NVIDIA GPUs across sovereign clouds and AI factories. Announced at APEC Summit Oct 2025.
- Japan: Ministry of Economy committing ~$6.6B over 5 years for domestic AI; SoftBank investing ~$13.1B in data centers. All NVIDIA GPU-based.
- Saudi Arabia: $40B+ committed to AI under Vision 2030, including data center buildouts with NVIDIA infrastructure.
- UAE: MGX/BlackRock/Microsoft/NVIDIA consortium acquiring Aligned Data Centers for ~$40B, securing 5GW+ capacity across the Americas.
- Europe: Deutsche Telekom launched the world’s first Industrial AI Cloud (NVIDIA-powered, live early 2026). Orange, Fastweb, Swisscom, Telefónica, Telenor all scaling sovereign AI on NVIDIA platforms. Germany deploying tens of thousands of GPUs. Italy building sovereign infrastructure with NVIDIA.
- Canada: ~$3B CAD for Sovereign Public AI Infrastructure, including NVIDIA H200 and Blackwell GPUs.
- Singapore + Southeast Asia: Regional coalition developing Sea-Lion AI on NVIDIA compute.
Why This Matters for Q4 and Beyond
TSMC’s CEO CC Wei specifically called out “sovereign AI” as one of three AI demand vectors (alongside consumer and enterprise) during the Q4 earnings call. NVIDIA’s Q3 noted that AI factory announcements totaled 5 million GPUs across partnerships. Many of these sovereign orders are beginning to convert to revenue in Q4 FY2026 and especially FY2027, representing a diversification away from the Big 4 hyperscaler concentration risk.
6. Neo-Clouds & AI-Native Infrastructure Providers: The New Demand Layer
Beyond the hyperscalers, a rapidly growing cohort of “neo-clouds” and AI-focused infrastructure companies are becoming significant NVIDIA customers:
CoreWeave
- NVIDIA invested $2B at $87.20/share (Jan 2026), now holds ~13% stake.
- Building 5GW of AI factory capacity by 2030; early adopter of Vera Rubin platform.
- $6.3B master services agreement with NVIDIA (unused capacity guarantee through 2032).
- Total contracts with OpenAI valued at $22.4B; $14.2B commitment from Meta through 2031.
- Revenue of $1.36B in Q3 alone, scaling rapidly.
OpenAI / Stargate
- 10GW NVIDIA deployment announced as strategic partnership; NVIDIA investing up to $100B.
- $500B Stargate initiative (OpenAI + Oracle + SoftBank) — largely NVIDIA-powered.
- Also signed 6GW AMD deal, showing breadth of demand, but NVIDIA is primary.
Other Neo-Clouds
- Lambda: Building 100+ MW AI factory in Kansas City with 10,000+ GB300 NVL72 GPUs.
- Together AI: Operating AI factories with B200 GPUs, expanding with GB200/GB300.
- xAI: Colossus 2 data center in Memphis — 500,000+ NVIDIA GPUs.
- Global AI: Placed order for 128 GB300 NVL72 racks (9,000+ GPUs) — largest GB300 deployment in New York.
Significance
The neo-cloud layer represents incremental demand that the consensus may undercount. These companies raise billions in debt/equity specifically to buy NVIDIA GPUs, creating a virtuous cycle. While some analysts flag “circular financing” risks (NVIDIA invests → they buy NVIDIA gear), the dollar amounts of NVIDIA’s investments ($2B in CoreWeave) are tiny relative to the revenue they generate (CoreWeave alone is on a ~$5.5B+ annual revenue run rate). The neo-clouds also serve as a distribution channel for NVIDIA into the enterprise market, which is the next wave of AI adoption.
7. Concrete Q4 FY2026 Assessment
Revenue Estimate: $67.0–68.0B (vs. $65B ± 2% guidance, $65.5B consensus)
Rationale:
- NVIDIA has beaten guidance by an average of ~$2B+ in recent quarters
- TSMC’s January revenue acceleration confirms robust production into Q4 period
- SK Hynix’s record HBM shipments confirm memory supply kept pace with GPU production
- All hyperscaler earnings confirm demand is unconstrained — supply is the only limiter
- Blackwell (GB200 NVL72 racks) began shipping in volume during the quarter at ASPs of $2–3M per rack
- Networking revenue (NVLink, InfiniBand, Spectrum-X) likely continued its 162% YoY growth trajectory from Q3
EPS Estimate: $1.52–1.58 (vs. $1.45–1.52 consensus)
Gross Margin: ~74.5–75.0% (in line with guidance)
- The Blackwell ramp initially compressed margins but rack-level sales carry higher dollar margins
- Software revenue (NVIDIA AI Enterprise, NIMs) provides incremental margin uplift
FQ1 2027 Guidance (The Key Number): $72–75B likely
- Represents 45–55% YoY growth
- Blackwell at full ramp; Blackwell Ultra (B300) beginning to contribute
- Q1 is typically a seasonally softer quarter, but AI demand is non-seasonal
8. Stock Price Reaction Assessment
Current price: ~$183 | 52-week high: $212 | Avg analyst target: $254
Bull Case (60% probability): Stock moves to $200–215 (+9–17%)
- Revenue beat of $67B+, strong Q1 guide of $73B+
- Gross margins hold at 75%
- Jensen confirms Blackwell Ultra on track for H2 2026, Vera Rubin for late 2026
- Confirms no “demand air pocket” between Blackwell and Rubin
- Forward P/E of ~25x on FY2027E EPS of $7.50+ looks very cheap
Base Case (25% probability): Stock moves to $190–200 (+4–9%)
- Revenue meets at ~$66B, Q1 guide at $70–72B
- Margins slightly below guidance (~74%)
- Narrative is “good, not great” — continuation of recent pattern where massive beats don’t move the stock
Bear Case (15% probability): Stock drops to $165–175 (-4–10%)
- Revenue near low end of range ($63–64B) due to CoWoS supply issues
- Q1 guide disappoints at <$70B
- Gross margin compresses below 74% on Blackwell mix
- Any commentary suggesting demand moderation or customer pushback on pricing
- China export restriction headwinds mentioned more prominently
9. Key Variables to Watch on Feb 25
- Q1 FY2027 revenue guidance — This is the most market-moving number. Anything above $73B is very bullish.
- Gross margin trajectory — Must hold ≥74.5% to maintain the narrative.
- Blackwell Ultra / B300 timeline — Confirming H2 2026 dismantles the “air pocket” thesis.
- Vera Rubin commentary — Updates on customer engagement and order pipeline for late 2026/2027.
- Networking revenue — A key indicator of full-stack adoption vs. GPU-only purchases.
- China revenue — Any guidance changes due to export controls could create noise.
- Customer concentration — If top 4 hyperscalers are >50% of data center revenue, diversification becomes a topic.
10. Bottom Line
The supply chain evidence is unambiguous: we are in the most aggressive AI infrastructure spending cycle in history, and it is accelerating, not decelerating. Every upstream supplier had record results. Every downstream customer is spending more than anyone expected. NVIDIA sits at the center of this with 85–90% market share in AI accelerators, a confirmed $500B order backlog, and a product roadmap (Blackwell → B300 → Vera Rubin) that maintains its architectural lead through 2027.
The stock at $183 with a forward P/E of ~25x on FY2027 estimates does not reflect this reality. The primary risk is not demand — it’s execution on supply and the broader macro/geopolitical environment. Barring a negative surprise on those fronts, NVDA is set up for a strong post-earnings move.
Disclaimer: This is an analytical synthesis, not investment advice. Markets can move in unexpected ways regardless of fundamentals.
❤️ 0 likes (likes are read-only in this static version)