Nvidia, Snowflake, and Palo Alto: Decoding Strategic Wins and Challenges in Tech
2024.11.22 09:48
Our inaugural issue dives into three critical earnings reports that came out over the past week:
- NVIDIA Corporation (NASDAQ:) fueling a rally in Networking and Power Gen
- Snowflake Inc (NYSE:) turning the corner toward sustained growth
- Palo Alto Networks Inc (NASDAQ:) great but not good enough
Nvidia igniting a rally in Networking and Power Gen
Nvidia (NASDAQ:), reported solid results, with revenues at $35B, exceeding street expectations by $2B. The guide, however, was underwhelming at $37B +/- 2%. Here is what is behind it:
Conservative? It is almost impossible for Blackwell (the new chip) to contribute $ billions, as the company indicated, and Hopper (the current gen chip) to grow, also implied as a potential scenario during the call, and, at the same time, the company to deliver only $2-3B incremental revenues. What makes management more cautious?
The AI Factory. Nvidia has less visibility into this product intro because Blackwell (1) requires brand-new data centers and (2) has a much more complex supply chain. Offline, the company cited that, unlike the Hopper, Blackwell has many more networking components beyond traditional switches, e.g., optical cables, transceivers, etc., which is a significant positive for Nvidia and the value chain but makes forecasting more challenging.
But less visibility doesn’t mean that results will be worse. On the contrary, once Blackwell ramps, we expect to see a step change in revenues.
The GB200 NVL72 systems require >100KW of rack density, and only
“I call it an AI factory because that’s really as close to what it is. It’s unlike a data center of the past” – Jensen Huang on the F3Q25 call
These comments are significantly positive for Networking, Power Gen, and broader data center infrastructure. We covered this in detail in our most recent webinar: Powering the Next Data Center Cycle. Access the replay here.
Scaling laws keep scaling. The biggest debate in AI is that hardware is clearly becoming more powerful, but is there an actual use? While we have already started seeing applications emerge (productivity & sales tools, cyber products, autonomy), the key question remains: where will the incremental demand come from?
Nvidia management spent a meaningful time on the call explaining how model training is changing over time. In the future, as compute costs decline, companies will start introducing models that can reason and continuously learn.
“A new scaling law called test-time scaling has emerged” – Jensen Huang on the F3Q25 call.
You can read more about this in our Data Center Primer, where we covered the topic in more detail (Generative AI => Thinking AI).
Snowflake is finally turning the corner
Snowflake (NYSE:), a data infrastructure company, reported an outstanding report, driving the stock up 30%+. We are finally seeing signs of AI fundamentals broadening from hardware to software. Here are the details:
Inflection. Snowflake’s stock has been under pressure this year due to several factors, but in addition, the company dropped the ball on product innovation. However, new management is bringing a fresh focus on go-to-market, which is finally bearing fruit, with the company delivering 29% growth in product revenue in 3Q and guiding to 22% in 4Q. Some highlights:
- Added $71M of new business (among the highest in history
- RPO increased 55% Y/Y to $5.7B and included three >$50M TCV deals…good as we look out to ’25/’26.
- NRR stabilized at 127%, flat Q/Q compared to 100-400bp declines in the past 3 quarters…big change.
Visibility into ’25. The go-to-market (GTM) changes were made around incentivizing the salesforce to drive consumption, and the most important point that came from the call is that the sales organization identified a strong backlog of new work-loads which customers plan to move into production in 4Q25/FY26…another solid data point for next year
- New products: $200mm run-rate in Snowpark, Dynamic tables, Connectors, etc…finally gaining some scale.
- 30% of customers (>3,200) were using Snowflake AI/ML.
- Minimal storage headwinds from Iceberg…which was a big driver of the bear thesis.
Competition. Investors are always focused on competitive dynamics, especially since DataBricks has been making a strong marketing push. But, Snowflake and Databricks each have their own turfs, with Snowflake being a better fit for enterprises trying to build AI applications and Databricks better with AI (tech) companies. While there is some overlap, the market is so big and evergreen, making it a lot more about execution rather than competition.
Bottom in Software. The most important broader point is that software has been demolished this year. Between interest rates remaining higher for longer and earnings being dampened by tough macro, many companies that were trading >10x EV/Revenues with 30% growth are now trading significantly below with an opportunity to re-rate if they can accelerate growth. This was the case with Snowflake, Shopify (NYSE:) and we expect will be the case for several more.
Palo Alto – great but not good enough
Palo Alto (NASDAQ:), a cybersecurity leader, reported another very strong quarter with NGS ARR the bright spot and platformization showing progress.
NGS & Platformization Highlights:
- 39% YoY growth now ~$4.5B
- Cortex surpassed $1bn of ARR, 22% of total NGS ARR
- Prisma Access 20% YoY growth in active SASE customers with 40% SASE customers new to Palo Alto.
- Closed ~100 platform deals, compared to 90 and 65 in the last two quarters.
But the numbers are becoming a bit messy, making it hard to make a cohesive ’25 revenue projection.
Disappointing billings, disappointing cRPO, total new ARR down, tailwinds to FCF, which may not persist)…making it very hard to get comfortable with C’25 growth.
The big question: Is Palo Alto going to be able to maintain 20% growth going into next year, or is it more of a “teens” grower. The company is currently priced as a 20s% grower. In order to realize upside, growth would need to further accelerate. Similarly, growth in the teens would potentially mean a significantly lower multiple.
This is going to be the key question for the rest of the cybersecurity companies reporting over the next two weeks.
