Opening question: A company with a market cap in the hundreds of billions, full-year revenue growth of 53%, net income growth of 55%, non-GAAP net income growth of 75%—how many such companies exist globally?
AI king NVIDIA's Q4 earnings arrived late; summed up in one word: "record." The report overall matched yesterday's forecast; refer back to "Earnings Preview | Share Price Consolidates, Can NVIDIA 'Turn Around' on Earnings?".
NVIDIA Q4 Earnings Summary:
Q4 revenue $5.003B, up 61% year over year, consistent with my forecast yesterday, above Bloomberg consensus of $4.818B, marking the third consecutive quarter of record highs;
Q4 GAAP gross margin 63.1%, non-GAAP gross margin 65.5%;
Q4 GAAP net income $1.457B, up 53% year over year; non-GAAP net income $1.957B, up 67% year over year, both marking the second consecutive quarter of record highs.
Since the views expressed in yesterday's preview were validated one by one, I won't repeat them. Just a few key points by segment:
Gaming (Q4 revenue share 50%):
Gaming Q4 revenue $2.495B, up 67% year over year, marking the second consecutive quarter of record highs; Q1 guidance calls for sequential growth, setting another record!
RTX 30 series shipments far exceed prior-generation same-period performance; inventory remains low; GeForce Now cloud gaming registered users reach 6M;
NVIDIA cites analyst estimates that Q4 mining card revenue was $100M-$300M; officially expects Q1 CMP mining card revenue of $15M, a small share.
Data Center (Q4 revenue share 38%):
Data center Q4 revenue $1.903B (yesterday's forecast $2B), up 96.6% year over year, marking the third consecutive quarter of record highs; Q1 guidance calls for sequential growth, another record! Vertical industry data center revenue exceeds 50% of the segment.
Mellanox Q4 revenue ~$500M, up 31.5% year over year, down 19% sequentially, in line with guidance. Excluding Mellanox, data center grew 45% year over year; Q1 guidance calls for both to grow sequentially; (starting Q1, Mellanox revenue no longer disclosed separately)
Notably, Jensen Huang believes A100 is still in early shipment stages with room to grow. Most importantly, he is very bullish on BlueField DPU prospects: future data center security is critical, and DPU can serve as a security chip. At Q3 earnings, NVIDIA indicated DPU TAM could reach $10B.
Pandemic-Impacted Businesses—Professional Visualization (Q4 revenue share 6%) and Automotive (Q4 revenue share 3%):
Professional visualization Q4 revenue $307M, down 7.3% year over year, up 30% sequentially. Professional design notebook shipments hit a record; also won major medical deals from GE, Siemens, and Oxford.
Closely watched automotive Q4 revenue $145M, down 11% year over year, up 16% sequentially. AI cockpit revenue grew year over year; traditional infotainment declined;
Software: AI cockpit has secured multi-billion dollar long-term orders (Mercedes) to be recognized in future; Hesai, Baraja, Innoviz, Magna, Ouster and other LiDAR vendors all use NVIDIA sensor solutions;
Hardware: SAIC, NIO, Li Auto new models all use NVIDIA Orin compute platform.
Capacity constraints exist, but perhaps not that severe
Many media cite capacity constraints to explain NVIDIA's post-earnings share decline; ignore short-term price swings. The view remains: capacity is an industry-wide issue, and NVIDIA's split production across Samsung and TSMC is actually an advantage.
On the earnings call, when asked about capacity, Jensen Huang said demand is indeed exploding, but NVIDIA has the world's best operations team, Q4 made it through, Q1 will only be better. Hence Q1 guidance slapped Wall Street in the face.
There's another often-overlooked aspect of capacity: when NVIDIA, AMD, and Intel are all capacity-constrained, everyone is affected, and NVIDIA's GPU position is more critical.
As for the Arm acquisition, it remains on track; of course, relative to the 18-month timeline set last September, it's still early, but there is hope.
Finally, I fully agree with Jensen Huang's three-stage AI view on the call: stage one was the birth of the AI computing platform seven or eight years ago, stage two was the birth of cloud computing, stage three is AI industrialization, and we are now in stage three. Please wait patiently for the good news~