NVDA -- Earnings Review Q4 FY26 + MS TMT Conference Combined Analysis

Date: 2026-03-05 | WSM | Initiating 5-7% position


Verdict

All three re-entry gates cleared. Initiating. Revenue $68.1B (> $67B trigger), Q1 FY27 guide $78B (> $75B trigger by $3B), Non-GAAP GM 75.0% (> 74% trigger). Morgan Stanley's Joseph Moore called it "the largest, cleanest beat and raise in the history of the semis industry." I agree.


Revenue Trajectory -- The Primary Lens

Q3 FY24 Q4 FY24 Q1 FY25 Q2 FY25 Q3 FY25 Q4 FY25 Q1 FY26 Q2 FY26 Q3 FY26 Q4 FY26 Q1 FY27G
Revenue ($B) 18.1 22.1 26.0 30.0 35.1 39.3 44.1 46.7 57.0 68.1 78.0G
QoQ % -- +22% +18% +15% +17% +12% +12% +6% +22% +20% +15%G
YoY % -- -- -- -- +94% +78% +69% +56% +63% +73% +77%G
QoQ $ Add ($B) -- +4.0 +3.9 +4.0 +5.1 +4.2 +4.8 +2.6 +10.3 +11.1 +9.9G

YoY trajectory: 56% -> 63% -> 73% -> 77% guided. Three consecutive quarters of reaccelerating growth at $68B quarterly revenue. Companies this large don't accelerate. NVIDIA is.

Q2 FY26 dip (+6% QoQ) now fully confirmed as Blackwell transition artifact. Q3 +22%, Q4 +20%, guide +15% -- normal cadence restored at dramatically higher baseline.


Beat vs Expectations

Metric Consensus/Guide Actual Beat
Revenue 65.0B(guide)/ 66B (Street) $68.1B +$3.1B vs guide (+4.8%)
EPS [Non-GAAP] $1.54 $1.62 +$0.08 (+5.2%)
Data Center ~$58-60B $62.3B +$2-4B
Networking ~$9-10B $11.0B +$1-2B
Q1 FY27 Guide ~$72-73B (Street) $78.0B +$5-6B
Q1 FY27 GM Guide ~74% 75.0% +1pt

The Q1 FY27 guide beat is the headline. 78Bvs 72-73B Street consensus is a $5-6B raise. "The largest, cleanest beat and raise in the history of the semis industry."


Data Center Deep Dive

Q4 FY25 Q1 FY26 Q2 FY26 Q3 FY26 Q4 FY26
DC Revenue ($B) 35.1 39.9 40.9 51.2 62.3
DC YoY % -- -- -- +66% +75%
DC QoQ % -- +14% +3% +25% +22%
Networking ($B) ~3.0 -- -- 8.2 11.0
Networking YoY -- -- -- +162% 3.5x
DC % of Total 89% 90% 88% 90% 92%

Networking at $11B jumped off the page. +34% QoQ, 3.5x YoY. NVLink + Spectrum X at this scale = mega-clusters deploying. $31B+ for FY26 total (10x vs FY21). This is the best leading indicator of infrastructure scale.

Customer diversification improving:

Three new demand vectors (from MS TMT Conference, Mar 4):

  1. OpenAI expanding from Azure/OCI to AWS -- NVIDIA "ramping AWS like mad"
  2. Anthropic expanding aggressively at both AWS and Azure -- factory-limited
  3. MSL (Meta's new AI lab) -- "net new on top of Meta," millions of GPUs
  4. Also: OpenClaw agentic AI framework surpassed Linux downloads in 3 weeks

Margins -- The Concern That Wasn't

Q1 FY26 Q2 FY26 Q3 FY26 Q4 FY26 Q1 FY27G
GM [GAAP] 60.5% 75.1% 73.4% 75.0% 74.9%G
GM [Non-GAAP] 64.9% 75.7% 73.6% 75.2% 75.0%G
Op Income ($B) -- -- -- $44.3 --
Op Margin [GAAP] 48.8% 64.9% 63.2% 65.1% --
Net Income [GAAP] ($B) -- -- $31.9 $43.0 --
Net Income [Non-GAAP] ($B) -- -- -- $39.6 --

Q4 Non-GAAP GM of 75.2% -- highest of FY26 (excluding anomalous Q2 timing). Q1 FY27 guided at 75.0% +/-50bps. "Mid-seventies" is holding. Rubin ramp hasn't started yet (samples shipped, production H2), so GM compression is FY28 risk at earliest.

Colette: "Delivering generational leads to customers is the single most important lever of gross margins." Innovation velocity = margin moat. Vera Rubin at 10x lower inference cost vs Blackwell = customers pay premium. Blackwell Ultra at 50x vs Hopper per SemiAnalysis. As long as each generation is an order of magnitude better, pricing power holds.


Free Cash Flow

Q3 FY26 Q4 FY26 FY26 Total
FCF ($B) $22.1 $34.9 $96.6
FCF Margin 38.8% 51.2% 44.7%
Operating CF ($B) -- -- $102.7

$34.9B FCF in a single quarter (+58% QoQ from Q3's $22.1B). FY26 total: $96.6B. Run-rate: $139.6B.

Capital returned: 41.1B(40.1B buybacks + $974M dividends). Cash on hand: $62.6B. Remaining buyback authorization: $58.5B.

Investing $40B in customers (OpenAI + Anthropic) AND buying back $40B of stock annually. This is what happens when FCF is nearly $100B/yr.


Valuation -- WSM Run-Rate Method

Metric Q3 FY26 (Feb 22) Q4 FY26 (Mar 5) Q1 FY27G Forward
Stock Price -- $183.30 --
Market Cap $4,505B $4,450B $4,450B
Run-Rate Revenue $228.0B $272.4B $312.0B
Run-Rate P/S 19.8x 16.3x 14.3x
Run-Rate Net Income [GAAP] $127.6B $172.0B --
Run-Rate P/E [GAAP] 35.3x 25.9x --
Run-Rate P/E [Non-GAAP] -- 28.1x --
Run-Rate FCF $88.4B $139.6B --
P/FCF 50.9x 31.9x --
PEG (P/E / YoY %) -- 0.35 --

Valuation compression in one quarter despite flat stock:

25.9x run-rate GAAP earnings for 73% YoY growth = PEG 0.35. S&P 500 trades ~21x with mid-single-digit growth. Market is pricing NVIDIA as if growth is about to cliff. The $78B Q1 guide (+77% YoY) says the opposite.

Forward run-rate P/S of 14.3x for 75% GM, 65% OpM, 51% FCF margin company growing 70%+. In SaaS land, companies with half these metrics trade at 15-20x.


Combined Thesis: MS TMT Conference + Q4 Earnings

The MS conference (Mar 4) gave the why -- three new demand vectors, OpenClaw developer lock-in, compute-equals-revenues proven, $2T software TAM converting to tokens. The Q4 earnings (Feb 25) gave the what -- $68B revenue, 75% margins, $35B FCF, $78B guide.

Together:

  1. Demand is widening -- new vectors: OpenAI@AWS, Anthropic multi-cloud, MSL, sovereign (30B), physicalAI(6B)
  2. Growth is reaccelerating -- 73% YoY, guided 77%
  3. Margins are stable -- 75% Non-GAAP, guided to hold
  4. Cash generation is expanding -- $35B FCF in Q4 (+58% QoQ)
  5. The moat is deepening -- OpenClaw = CUDA 2.0 for agentic AI
  6. Valuation is compressing -- 25.9x run-rate P/E, PEG 0.35

Product Pipeline & Forward Visibility

Product Status Revenue Impact
Blackwell Shipping, 9 GW deployed Current revenue driver
Blackwell Ultra Shipping, 50x vs Hopper (SemiAnalysis) Q1-Q2 FY27
Vera Rubin Samples shipped, production H2 CY2026 FY28 driver
Vera Rubin Ultra Roadmapped FY29+
OpenClaw Live, most downloaded OSS ever Ecosystem lock-in

Partnerships & Investments:


Risks (Honest Accounting)

  1. ASIC displacement -- Broadcom, Meta MTIA, Google TPU v6. Jensen addressed none of this at MS conference or Q4 call. 2027+ risk. Not in the numbers yet. Track quarterly.
  2. Hyperscaler CapEx cycle -- If even one major hyperscaler pauses CapEx growth, narrative flips. Demand signals strong today but this is cyclical risk in structural clothing.
  3. Rubin transition GM -- H2 CY2026 production ramp. Blackwell transition caused Q1 FY26 GM dip to 65%. Could repeat. Watch Q3-Q4 FY27.
  4. China wildcard -- Guidance excludes China DC compute. Policy change is upside/uncertainty.
  5. Gaming supply constraint -- DC allocation priority starving Gaming. If Gaming declines, perception risk despite irrelevance to thesis.

None present in current data. But at $4.5T, even a 10% correction = $450B of value destruction. Size accordingly.


Position & Action

Initiating 5-7% position. All three gates cleared:

Add triggers:

Sell triggers:


Summary

Factor Assessment
Revenue trajectory Reaccelerating -- 73% YoY, 77% guided, 3 consecutive Qs of acceleration
Leading indicators All bullish -- networking 3.5x YoY, 3 new demand vectors, OpenClaw adoption
Gross margin Strong and stable -- 75.2% Non-GAAP, 75.0% guided Q1 FY27
FCF Exceptional -- $34.9B Q4, $96.6B FY26, $139.6B run-rate
Valuation Compelling -- 25.9x run-rate P/E, PEG 0.35, 16.3x run-rate P/S
Demand visibility Widening -- sovereign, physical AI, agentic, enterprise + 3 new hyperscale vectors
Risks ASIC displacement (2027+), CapEx cycle, Rubin GM transition
Thesis Intact. Strengthening. Initiating.

The three simultaneous platform shifts (CPU->GPU, gen AI, agentic AI) are real, not marketing. NVIDIA is the only architecture running every AI model in production at scale. OpenClaw is the CUDA moment for agentic AI. And the $2T software industry is about to become a token consumer.

At 25.9x run-rate earnings with 73% growth reaccelerating, this is the most asymmetric risk/reward setup in mega-cap tech.

-wsm (Long NVDA, 5-7%. Initiating.)


Sources: