Greetings from San Francisco! 👋🏼
A warm welcome to the recent additions to our community!
Over 89,000 How They Make Money subscribers turn to us weekly for business and investment insights. Glad you're here.
Watch our quick take on NVIDIA’s Q4 earnings on our YouTube Channel. Get all the insights you need in a few minutes with our commentary and analysis alongside the visuals.
NVIDIA (NVDA) soared 16% yesterday after reporting its Q4 results.
Enough to add $277 billion in market value, the largest one-day gain in history.
With a nearly $2 trillion market cap, the company is now bigger than Amazon (AMZN) and Alphabet (GOOG). If the stock surges another 60% from here, NVIDIA will become the largest company in the world.
Jensen Huang explained in his prepared remarks:
“Accelerated computing and generative AI have hit the tipping point. Demand is surging worldwide across companies, industries and nations.”
The commentary on the near-term outlook remained optimistic, with guidance well ahead of analysts’ expectations.
If you learn one insight from today’s article, it should be that inference workloads contributed to roughly 40% of NVIDIA’s Data Center revenue in the past year.
Last quarter, I broke down the two types of AI workloads: Training vs. Inference.
AI systems operate through two core stages:
🎓 Training: Imagine AI as a student learning from vast amounts of data. In this phase, AI develops intelligence and pattern recognition. NVIDIA dominates this segment with its robust Graphic Processing Units (GPUs), accelerating the learning process.
🧠 Inference: Post-learning, AI applies its newfound knowledge to real-world scenarios and decision-making. NVIDIA faces stiffer competition from Intel, Qualcomm, and big tech in this category, but the company is making significant progress.
NVIDIA already has a lead in training, but inference is deemed more competitive. So, it was a critical soundbite to hear that such a large portion of NVIDIA’s massive Data Center revenue comes from inference.
Inference represents a larger market than training, implying no saturation in sight. It could also mean a diversification of revenue beyond the large cloud providers. Inference is closely linked to what BofA described as “revenue-bearing AI,” where companies directly use AI to generate income by delivering a service.
In a context where the crux of the thesis is the durability of the demand for NVIDIA’s AI solutions, inference will likely become more crucial to future-proof the business.
Let’s unpack the quarter.
Today at a glance:
NVIDIA’s Q4 FY24.
Recent development.
Key quotes from the call.
What to watch looking forward.
1. NVIDIA Q4 FY24
Income statement:
Here is the bird’s-eye view of the income statement.
I’m focusing on sequential growth (quarter-over-quarter), a better representation of the current growth dynamic.
Revenue jumped +22% Q/Q to $22.1 billion (vs. a $20 billion guidance).
⚙️ Data Center grew +27% Q/Q to $18.4 billion.
🎮 Gaming was flat at $2.9 billion.
👁️ Professional Visualization grew +11% Q/Q to $0.5 billion.
🚘 Automotive grew +8% Q/Q to $0.3 billion.
🏭 OEM & Other grew +23% Q/Q to $0.1 billion.
The graph below shows the past 13 quarters, which paint a challenging time in FY23 caused by the decline in the Gaming segment we covered previously, followed by the stunning rise in Data Center revenue in the past year.
Gross margin was 76% (+2pp Q/Q), compared to a 74.5% guidance.
Operating margin was 62% (+4pp Q/Q).
Non-GAAP operating margin was 67% (+3pp Q/Q).
Non-GAAP EPS $5.16 ($0.52 beat).
Cash flow:
Operating cash flow was $11.5 billion (52% margin).
Free cash flow was $11.2 billion (51% margin).
Balance sheet:
Cash and cash equivalent: $26.0 billion.
Debt: $9.7 billion.
Q1 FY25 Guidance:
Revenue +9% Q/Q to $22.0 billion ($2.0 billion beat).
Gross margin 76.3% (+0.3pp Q/Q).
So what to make of all this?
NVIDIA delivered another exceptional revenue beat of 8%. If you recall, the revenue beat was 10% in Q1, 23% in Q2 and 13% in Q3. The Q4 revenue guidance was $1.6 billion ahead of consensus. That compares to a $2.2 billion beat in the previous quarter. Despite resetting expectations higher every quarter, NVIDIA still finds a way to surprise by a wide margin.
⚙️ Data Center was 83% of overall revenue (+3pp Q/Q).
As expected, Data center sales to China declined significantly due to US export controls.
Yet, the segment continued to surge. Like previous quarters, it was led by strong demand for the Hopper GPU platforms related to a surge in demand for training and inference of large language models (LLMs), along with the InfiniBand network.
Compute revenue grew more than 5x, and networking revenue tripled from a year ago.
🎮 Gaming was flat sequentially but up 56% year-over-year and ahead of expectations. The segment has doubled compared to pre-COVID. The channel inventory levels improved in the past year. The launch of the GeForce RTX 40 SUPER Series family of GPUs also contributed. Management highlighted an installed base of over 100 million RTX PCs.
👁️ Professional Visualization accelerated to +11% sequentially, driven by higher demand for workstation GPUs based on the Ada Lovelace architecture.
🚘 Automotive accelerated +8% sequentially (from +3% last quarter). Nearly 80 automakers use NVIDIA’s AI structure for autonomous driving and other applications.
Margins improved significantly, boosted by increased Data Center revenue and more favorable component costs. The gross margin was 1.5 percentage points ahead of guidance. Management expects a gross margin in the mid-70s for FY25.
Guidance beat expectations ($2 billion beat or 10%). The continued rapid growth in Q1 FY25 will come from, you guessed it, Data Center. Gaming is expected to decline based on seasonality.
2. Recent developments
💰 NVIDIA’s equity portfolio
Nvidia recently disclosed its equity investments via its quarterly 13F filing. Here's a summary of their disclosed stakes:
📱 Arm Holdings (ARM): The largest stake was a $147 million investment in the UK chip designer. So far, it looks like a smart move. ARM surged 58% (!) after its latest earnings report. Why? A massive revenue beat and guidance raise, combined with AI optimism. Check out our full breakdown to learn more about ARM and how they make money.
🧬 Recursion Pharmaceuticals (RXRX): NVIDIA announced a $50 million investment in the small biotech company in July. Recursion uses NVIDIA DGX Cloud to train its AI models using its 23-petabyte dataset, which is focused on drug discovery.
In addition, the filing disclosed smaller stakes (under $4 million):
🗣️ SoundHound AI (SOUN): Investment in voice recognition technology.
🚚 TuSimple (TSPH): Autonomous transportation.
🔬 Nano-X Imaging (NNOX): Medical imaging development.
These strategic bets illustrate Nvidia's expansive approach to AI, positioning the company at the forefront of multiple industries.
☁️ Microsoft is developing a networking card
Last quarter, Microsoft announced custom AI chips, the Azure Maia AI Accelerator, and the Azure Cobalt CPU. All cloud providers have been focused on lowering costs and improving their self-reliance.
Now, according to The Information, Microsoft is developing a networking card.
🛠️ Self Reliance: Developing a networking card would lessen Microsoft’s reliance on NVIDIA to enhance server chip performance at a lower cost.
🤝 Strategic Moves: The project is led by Pradeep Sindhu (co-founder of Juniper Networks) following Microsoft’s acquisition of Fungible.
🌐 Head-to-head with NVIDIA: The project mirrors NVIDIA's ConnectX-7 card. Jensen Huang said networking exceeds a $10 billion annual revenue run rate.
🤖 Supporting OpenAI: This project could benefit OpenAI, with Microsoft's infrastructure aiding AI model training.
The next few years look bright for NVIDIA’s Data Center revenue. But the desire for self-reliance across all hyperscalers—NVIDIA’s largest customers—could challenge the company's dominance in AI accelerators.
3. Key quotes from the earnings call
CFO Colette Kress:
On Data Center:
“Data center growth was driven by both training and inference of generative AI and large language models across a broad set of industries, use cases and regions.”
She provided updates on the three major customer categories:
☁️ Cloud service providers (CSPs) contributed roughly half of Data Center revenue in previous quarters. All hyperscalers (Amazon, Microsoft, Google) are NVIDIA customers. Kress highlighted Microsoft’s success with Github Copilot (an AI pair programmer).
💬 Consumer Internet companies have been the second largest category this year. Our review of Meta’s latest earnings covered the company’s massive investment in 350,000 H100s from NVIDIA (over $10 billion).
🗄️ Enterprise demand for AI and accelerated computing touches all sectors. Leading software companies like Adobe, Databricks, and Snowflake are adding AI copilots to their platforms. Adoption is far-reaching across automotive, healthcare, and finance.
On the impact of export restrictions:
“Our data center revenue [in China] is ~20% to 25% of any one of our quarters. China represented a mid-single-digit percentage of our data center revenue in Q4 and we expect it to stay in a similar range in the Q1 FY25.”
The US regulations affect the highest performance levels. New regulation-compliant products will become available in the coming months.
On software:
“We also made great progress with our software and services offerings, which reached an annualized revenue run rate of $1 billion in Q4.”
NVIDIA DGX Cloud added AWS as a new partner, joining other CSPs.
CEO Jensen Huang:
On inference vs. training:
“The amount of inference that we do is just off the charts now […] The inference part of our business has grown tremendously. We estimate about 40%. The amount of training is continuing, because these models are getting larger and larger, the amount of inference is increasing.”
When you interact with ChatGPT, that’s an example of an inference handled by NVIDIA’s GPU platform. Last quarter, we discussed that NVIDIA’s focus on inference performance and cost-effectiveness is essential. Inference is expected to represent an increasing share of AI workloads.
On diversification of the customer set:
"We are diversifying into new industries. Auto, health, robotics, financial services [...] a collection of multibillion-dollar industries are embracing our generative AI."
This will be increasingly important for NVIDIA to sustain the demand for its Data Center solutions over many years.
On Sovereign AI:
“We're seeing sovereign AI infrastructure is being built in Japan, in Canada, in France, so many other regions. And so my expectation is that what is being experienced here in the United States, in the West, will surely be replicated around the world.”
Local governments and vertical industries want to protect their data and train and transform their own models. This implies an expansion of the Generative AI market. Jensen Huang previously described a ‘generative AI wave’ from one category to the next:
→ Startups and CSPs.
→ Consumer Internet.
→ Software platforms.
→ Enterprise and government.
In short: Everyone is getting an LLM.
On demand visibility:
Wall Street focuses on when the current demand surge will taper off. Asked whether the Data Center would peak in 2024, Huang said:
“The conditions are excellent for continued growth calendar '24, to calendar '25 and beyond.”
Huang continues to see three massive tailwinds:
Transition from general-purpose to accelerated computing
Generative AI.
A whole new industry (think ChatGPT, Midjourney, or Gemini).
On the opportunity in Enterprise Software:
“Every enterprise in the world, every software enterprise company that [is] deploying software, in all the clouds and private clouds and on-prem will run on NVIDIA AI enterprise.”
AI amplifies the demand for software solutions. NVIDIA is offering a sandbox on DGX Cloud either on-prem or via CSPs. It then provides support and charges a licensing fee. Customers will run their custom AI models on NVIDIA AI Enterprise, paying $4,500 per GPU per year.
4. What to watch looking forward
Buybacks
NVIDIA repurchased $2.7 billion worth of its own stock, bringing the total to $9.5 billion for the year. It was less than in Q3 ($3.8 billion) but still on an elevated pace. As a reminder, management initiated a new $25 billion repurchase program, hinting at more buybacks in FY25.
For context, management bought back $10 billion of NVDA shares in FY23, so the company maintained a similar pace in FY24. Time will tell if it’s the best use of the current influx of cash. But with clear demand visibility, management is once again sending a message to Wall Street that the stock is attractive enough.
Valuation
Still on the buy list? NVIDIA's stock valuation remains controversial. While Altimeter and Stan Druckenmiller continued to buy NVDA in Q4, the stock was mostly absent from high-profile hedge funds’ top picks in the latest round of 13F filings. Since then? The stock has surged nearly 60% so far in 2024. NVIDIA remained a top position for many money managers in December, from Coatue to Light Street.
NVIDIA's forward PE ratio is roughly 32, according to data from YCharts. For context, that’s slightly less than Microsoft, trading at 35 times forward earnings while growing revenue 12% Y/Y in constant currency, excluding the Activision acquisition. Of course, we are comparing apples to oranges here, and the sustainability of the current demand for NVIDIA’s GPUs is still unclear. A lot of future demand is already factored in NVIDIA’s forward earnings estimates, and the semiconductor industry is known for its cyclicality.
The next Cisco?
For so many analysts, the parallel is too tempting.
In the late 1990s, Cisco (CSCO) manufactured vital equipment necessary to expand the internet. Today, NVIDIA powers the AI supercycle.
Could future earnings expectations be too optimistic beyond the next two years?
CSCO had a trailing PE ratio of 240 times at its peak
NVDA’s trailing PE is roughly 66 as of this writing.
Even assuming NVDA is the next CSCO, we are still far from the valuation exuberance of the tech bubble.
I keep seeing chart crimes on social media, aligning the latest stock chart of NVDA to CSCO in the 1990s, and concluding the stock is about to crater. If you are going to compare these two companies, let’s focus on the fundamentals, shall we? NVIDIA is growing revenue and earnings much faster than Cisco ever did. The stock chart is meaningless in a vacuum.
NVIDIA has two key competitive moats:
Switching costs: Proprietary software like Cuda for AI tools.
Network effects: From ecosystem to compatibility, the more people use its products, the better they get.
Ultimately, NVIDIA will rise and fall based on the capital expenditures of its largest customers. If we enter a downturn in the economic cycle and the CSPs decide to cut their spending on Data Center, NVIDIA’s earnings (and valuation) will fizzle, at least temporarily.
Many things could go wrong:
⚙️ Existing competitors: AMD and Qualcomm are pushing to do better.
☁️ Large customers shifting in-house: Cloud providers like Microsoft, Alphabet, and Amazon are building solutions internally to be self-reliant.
🤯 New entrants: Bloomberg reported Softbank founder and CEO Masayoshi Son has a $100 billion plan to boost the global supply of AI-focused processors. OpenAI founder Sam Altman seeks up to $7 trillion (with a T) for a new AI chip project.
🗺️ Macro factors: Economic downturns, export controls, and supply chain disruption could challenge the growth story.
No matter what, NVDA isn’t an investment for the faint of heart, just like most semis stocks. History tells us that highly profitable industries tend to attract more competition, leading to mean reversion for the best performers.
Remember, one of the main questions defining NVIDIA’s future is how sustainable the demand for generative AI will be.
OpenAI recently unveiled Sora, its latest AI video model capable of turning text prompts into realistic videos up to a minute long. Our latest video explains the magic behind this technology. If you haven’t seen Sora in action yet, get ready for a ‘oh shit!’ moment. 🤯
The progress in generative AI in the past year has only reinforced the narrative that entire industries will be upended.
And a lot of companies are going to need a lot of GPUs.
That’s it for today!
Stay healthy and invest on!
Apply to Sponsor Us
We're offering new sponsorship opportunities for B2B and B2C brands to get in front of our audience of investors and business leaders. Click here to learn more.
Get Your Business a Custom Chart
Interested in custom charts for your organization or brand? Complete the form here, and we'll get in touch.
Disclosure: I am long AAPL, AMD, AMZN, GOOG, NVDA, and SNOW in the App Economy Portfolio. I share my ratings (BUY, SELL, or HOLD) with App Economy Portfolio members.
Author's Note (Bertrand here 👋🏼): The views and opinions expressed in this newsletter are solely my own and should not be considered financial advice or any other organization's views.
a turning point, perhaps.