Locator: 48710NVDA.
Locator: 48710CUDA.
Update
May 29, 2025: from Shay today. A very, very long note. This is the first little bit:
The real story from Nvidia earnings wasn’t the revenue beat. It wasn’t the guide. It wasn’t even the China write-down. It was the $4.96B in networking revenue.
That number -- blowing past the $3.45B estimate -- is the loudest signal we’ve gotten yet that the center of gravity in AI compute has shifted. Not away from Nvidia, but within it. From training to inference. From chip to system.
This is the regime change the market hasn’t fully priced: inference isn’t a tail-end workload anymore. It’s becoming the default mode of AI at scale. And the economics, the architecture, the physical infrastructure of that shift all tilt back toward one conclusion -- this isn’t about GPUs alone. This is about the pipes. The fabric. The connective tissue. And right now, Nvidia owns that, too.
What they’re building isn’t a product line. It’s a substrate. Blackwell isn’t just replacing Hopper -- it’s replacing the idea that AI progress happens in discrete steps. When 70% of data center compute revenue already comes from a chip architecture launched just months ago, that’s not a product cycle. That’s cadence. And the fact that $MSFT has already deployed tens of thousands of GB200s with plans to scale to hundreds of thousands means this isn’t theoretical demand. It’s committed capital. It’s forward-positioned infrastructure. It’s factory-scale deployments being rolled out to meet the explosion in real-world token usage -- 100 trillion tokens processed in a single quarter. A 5x YoY surge.
You don’t build for that with old assumptions. You build with bandwidth. With latency minimization. With full-stack integration from silicon to rack. Which is why Nvidia’s networking business didn’t just outperform -- it became the heartbeat of the AI factory model. Spectrum-X, NVLink, custom interconnects -- they’re no longer supplementary. They’re elemental.
And here’s where the impact spills out to the rest of the ecosystem. Because once the market wakes up to inference at scale, it doesn’t just re-rate Nvidia. It elevates every company within the stack.
Start with AVGO. You don’t move trillions of tokens per quarter without silicon photonics, custom switch ASICs, and hardened firmware stacks that Broadcom dominates. Their position inside META, MSFT, GOOGL & AMZN networking is no longer optional -- it’s foundational. The scale Nvidia just outlined will require not just horizontal expansion, but vertical reengineering of every signal pathway in the data center. And Broadcom is already embedded in those redesigns.
Original Post
CUDA: an acronym used in computing, closely associated with Nvidia. Nothing to do with the publicly traded company Barracuda whose stock symbol is CUDA.
Reminder, FUD: fear, uncertainty, doubt.
From Shay this morning, link here:
The market keeps circling the same question: will China be the thing
that derails Nvidia? Will Huawei catch up? Will export controls cripple
growth?
Here’s the truth -- Nvidia isn’t reliant on China. It’s moving past it.
Yes, the hit was real: an $8B drag from the H20 export ban and related
write-downs. But even with that, Nvidia guided to $28B for Q2 -- in line
with the sell-side, and $2B ahead of where the buyside was modeling.
And yet, the headlines keep coming. Huawei’s building custom chips. A
Blackwell killer is on the way. Nvidia’s lead is narrowing. But this was
never about specs on paper -- it’s about systems, architecture, and
above all, CUDA.
Let’s be real: GOOGL, MSFT, META, and AMZN -- with bottomless
R&D budgets -- haven’t replaced CUDA (see at very end of this post).
TSLA, AAPL, ORCL all tried
building in-house stacks. None succeeded.
But somehow, Huawei -- under
sanctions, locked out of TSMC and ASML -- is going to not only match
Nvidia’s hardware but rebuild its software ecosystem from scratch?
I call BS.
Even if they match on silicon -- and they haven’t -- they’re still a
generation behind on the compiler. On the optimization. On the ecosystem
that makes the system run. That’s why even the most vertically
integrated hyperscalers still buy tens of thousands of GB200 Superchips.
Because nothing else scales like Nvidia.
This isn’t about whether Nvidia can survive without China -- it’s about
how long the market will keep pretending China is even in the race.
***********************************
Original Post
The AI ramp -- the sixth industrial revolution -- I honestly don't think -- in terms of dollars -- in the private sector -- being spent in one sector -- anyone has ever seen this before -- except perhaps with tulips, February, 1637. Example, MSFT / NVDA, link here.
Disclaimer
Brief
Reminder
Briefly:
- I am
inappropriately exuberant about the Bakken and I am often well out front
of my headlights. I am often appropriately accused of hyperbole when it
comes to the Bakken.
- I am inappropriately exuberant about the US economy and the US market.
- I am also inappropriately exuberant about all things Apple.
- See disclaimer. This is not an investment site.
- Disclaimer:
this is not an investment site. Do not make any investment, financial,
job, career, travel, or relationship decisions based on what you read
here or think you may have read here. All my posts are done quickly:
there will be content and typographical errors. If something appears wrong, it probably is. Feel free to fact check everything.
- If anything on any of my
posts is important to you, go to the source. If/when I find
typographical / content errors, I will correct them.
- Reminder: I am inappropriately exuberant about the Bakken, US economy, and the US market.
- I am also inappropriately exuberant about all things Apple.
- And
now, Nvidia, also. I am also inappropriately exuberant about all things
Nvidia. Nvidia is a metonym for AI and/or the sixth industrial
revolution.
- I've now added Broadcom to the disclaimer. I am also inappropriately exuberant about all things Broadcom.
- I've now added Oracle to the disclaimer. I am also inappropriately exuberant about all things Oracle.
- Longer version here.
****************************************
Nvidia
App economy, link here:
Data center revenue, link here:
Pre-market:
- P/E down to 43
- market cap: $3.3 trillion
*****************************
Previously Posted
May 28, 2025, 4:02 p.m. CT: and now that this is over, I'm done with CNBC and following the market for another three months when AAPL and NVDA report again. Good luck to all. After hours,
- the big four after Nvidia's earnings:
- NVDA: $142.47, up $7.66, up 5.68%;
- ORCL: $166.50, up $2.65, up 1.62%;
- AVGO: $247.16, up $7.79; up 3.25%;
- AMD: $114.89, up $2.03; up 1.75%;
- others;
- ARM: $139.22, up $3.68; up 2.72%
- TSM: $200.40, up $.26; up 2.17%
- GEV: $499.00 -- all-time high; up $13.99; up 2.88%;
- MU: $98.00; up $1.87; up 1.94% (was slightly down during the day)
- AMD up a bit
- AAPL: flat -- not a good sign for AAPL
And here they are, earnings:
- ticker
up $2.00 after hours and after earnings have apparently been released;
now up over $3.00 / share -- is that good news? Maybe.
- seems to have settled: up 3.5% after earnings release.
- later, after "everyone" has had chance to see the numbers and heard Jensen's comments, NVID is up over $8 / share; up about 6%;
- CNBC :
- beat on top and bottom line
- 96 cents vs 93 cents EPS
- $44.06 billion vs $43.31 billion
- $4.5 billion charge vs $5.0 billion charge
- $39.1 billion from data center services revenue; slightly ight; vs 39.22 billion estimate -- the crawler on CNBC is wrong (it will probably be corrected).
- "ticker is now popping." But not by much. Up over 4% after hours.
- Investopedia. Link here.
- Business Insider. Link here.
-
FIRST QUARTER RESULTS
Revenue $44.06 billion, +69% y/y, estimate $43.29 billion
- Data center revenue $39.1 billion, +73% y/y, estimate $39.22
billion
- Automotive revenue $567 million, +72% y/y, estimate $579.4
million
- Adjusted gross margin 61% vs. 78.9% y/y, estimate 71%
-R&D expenses $3.99 billion, +47% y/y, estimate $4.07 billion
- Adjusted operating expenses $3.58 billion, +43% y/y, estimate
$3.63 billion
- Adjusted operating income $23.28 billion, +29% y/y, estimate
$27.15 billion
- Adjusted EPS 81c, estimate 93c
- Free cash flow $26.14 billion, +75% y/y
Comments
Bad:
- most worrisome: margins, big miss, and significantly lower than previous
- operating income missed estimate but still up 29% y/y
Nice:
- operating expenses, though significantly higher than y/y, came in lower than estimate
Analysts:
- able to keep up with supply chain; meeting buyers' demands.
- remember: Nvidia has been locked out of China and yet they made their numbers.
Will be interesting to see the App Economy graphic.
Update
May 28, 2025: Beth says Nvidia will generate $1 trillion in cumulative revenue through fiscal 2029.
May 28, 2025: First time in months that I've tuned into CNBC. Anticipating Nvidia earnings today.
- Observations:
- analysts are over-thinking this; LOL;
- today is sort of like the Stanley Cup of trading; NBA championships of trading; nothing more, nothing less;
- wow, talk about endless talking about nothing that matters;
May 28, 2025: the
Trump administration orders US companies that offer software used to
design semiconductors to stop selling their services to Chinese groups.
We'll
start seeing stories everywhere. The big question is whether these
announcements are simply things that Trump has come up with or if there
is some discussion among his inner group? If the former, Trump will run
circles around bureaucracies and the judicial system. Again, can the
president simply issue edicts like this? It will take days / weeks /
months / years for the legislative and the judicial branches of the
government to consider these issues.
May 28, 2025: Palantir goes deeper into the US bureaucracy -- has now teamed up with Fannie Mae to fight mortgage fraud.
May 27, 2025: earnings out tomorrow, Wednesday, May 28, 2025, after market closes.
May 24, 2025: link here.
May 24, 2025: analysts' expectations, Investopedia.
Analysts on average expect Nvidia to report:
- quarterly revenue of $43.38 billion, 66% higher year-over-year; and,
- adjusted net income of $21.29 billion, or 87 cents per share, up from $15.24 billion, or 61 cents per share, a year earlier.
Original Post
****************************************
Previously Posted
Locator: 48478NVDA.
All about tech, right now.
Most important: AAPL, NVID, CRM.
Apple (AAPL) news here. (pending)
Over on twitter: Evan, Giovanni Staunovo, and EconomyApp.
Start here.
First two charts, Nvidia, link here:
- 2Q25: look at those incredible margins and other data points, from left to right:
- largest revenue center, "Data Center": $26.3 billion; grew 16% q/q
- revenue: $30.08 billion; grew 15% q/q
- gross profit: $22.6 billion; grew 3pp q/q; 75% margin;
- operating profit: $18.6billion; grew 3pp q/q; 62% margin;
- net profit: $16.6 billion; grew 2pp q/q; 55% margin;
From yesterday:
- 1Q25: look at those incredible margins and other data points, from left to right:
- largest revenue center, "Data Center": $22.6 billion; grew 23% q/q
- revenue: $26.08 billion; grew 18% q/q
- gross profit: $20.4 billion; grew 2pp q/q; 78% margin;
- operating profit: $16.9 billion; grew 3pp q/q; 65% margin;
- net profit: $14.9 billion; grew 2pp q/q; 75% margin;
Salesforce, link here:
- 2Q25: compare CRM margins below with NVDA margins above (spoiler alert: there's simply no comparison) and other data points, from left to right:
- subscription and support: $8.8 billion; grew 9% y/y;
- five revenue centers: each $1 billion to $2 billion; each grew about 10% y/y;
- revenue: $9.3 billion; grew 8% q/q
- gross profit: $7.2 billion; grew 1pp q/q; 77% margin;
- operating profit: $1.88 billion; grew 2pp q/q; 19% margin;
- net profit: $1.4 billion; grew 1pp q/q; 15% margin;
Now, conventional comparison, CRM vs NVDA:
CRM, p/e = 46.54
NVDA, forward p/e = 40; current p/e = 74:
NVDA vs CRM:
Summary of Nvidia's Earnings Call
Link here.
Positives: -
Record Revenue Growth: NVIDIA reported a record revenue of $30 billion
for Q2, a 15% sequential increase and a remarkable 122% year-on-year
growth, significantly surpassing the expected $28 billion. - Strong Data
Center Performance: Data Center revenue reached $26.3 billion, up 16%
sequentially and 154% year-on-year, driven by robust demand for NVIDIA's
Hopper architecture and networking platforms, particularly in
generative AI applications. - Innovative Product Pipeline: The
anticipation for the upcoming Blackwell platform is high, with
expectations of several billion dollars in revenue from Blackwell in Q4.
The platform promises significant improvements in performance and
efficiency. - Diverse Customer Base: NVIDIA's customer base is
expanding, with strong demand from cloud service providers, consumer
internet companies, and enterprises across various sectors, including
healthcare and automotive. - Growing Software Revenue: NVIDIA's AI
Enterprise platform is gaining traction, with expectations to approach a
$2 billion annual run rate by year-end, indicating strong growth in
software and services.
Negatives: -
Gross Margin Pressure: Gross margins decreased slightly due to a higher
mix of new products and inventory provisions for low-yielding Blackwell
materials, with expectations of continued pressure as new products ramp
up. - Competitive Market in China: While Data Center revenue in China
grew, the market remains competitive, and NVIDIA faces challenges due to
export controls affecting its operations in the region. - Increased
Operating Expenses: Operating expenses are expected to grow in the mid-
to upper 40% range as NVIDIA invests in developing next-generation
products, which may impact profitability in the short term. - Transition
Challenges: The shift from Hopper to Blackwell may present challenges
in terms of customer readiness and integration, particularly with the
adoption of liquid cooling technologies.
Overall: NVIDIA's
earnings call highlighted strong financial performance and growth
prospects driven by innovation in AI and accelerated computing, while
also acknowledging challenges related to margins and competitive
pressures.
****************************
CUDA
AI Overview
In the context of investing,
CUDA is strongly associated with NVIDIA Corporation (NVDA). It stands for Compute Unified Device Architecture.
CUDA is a parallel computing platform and application programming interface (API) model developed by NVIDIA. It enables developers to use NVIDIA's graphics processing units (GPUs) for general-purpose processing, also known as GPGPU (General-Purpose computing on Graphics Processing Units).
Its importance in investing stems from the fact that it has become the go-to platform for GPU acceleration, particularly in deep learning and AI applications. This is because:
-
Ease of use: CUDA provides a software layer that allows developers to easily access and utilize the parallel processing power of GPUs, making it easier for them to create complex programs and speed up compute-intensive applications.
-
Strong ecosystem: Over the years, a large developer community has built up around CUDA, leading to a wide range of software and applications that rely on it. This makes it difficult for users to switch to alternative GPU platforms, effectively creating a strong competitive advantage for NVIDIA.
-
Dominance in AI: CUDA's synergy with NVIDIA's GPUs has solidified the company's position as a dominant player in the AI industry, as it serves as the backbone for AI development and deployment. This makes NVIDIA an attractive investment in the growing AI market.
In summary, CUDA is a key factor behind NVIDIA's success and market dominance, particularly in the fields of AI and accelerated computing, which makes it an important consideration for investors looking into the technology sector.
Important Note: While CUDA is primarily associated with NVIDIA, Barracuda Networks, Inc. also uses the ticker symbol CUDA on the NYSE. Barracuda Networks is a company that provides security and data protection solutions. It is important to distinguish between the two entities when researching or discussing investments.
