$NVDA scientist on why he isn't worried about competition from inference ASICs, basically you want the flexibility of the GPU to deal with model changes going forward. We're running out of new data to train on, and so innovation will have to come from model changes. Only a
This is a credible long term bear case I've come across for $NVDA. The hyperscalers - from the large public clouds to the big internet giants - have an extremely strong position and the means to move to internal silicon over time. This will take time i.e. years and years, but as
Some thoughts on $NVDA's CUDA as a moat and what the actual moats for the company are:
A lot of analysts have the impression that AI coding is happening in CUDA, which would give Nvidia a strong moat on the software side, but this is unfortunately not the case. Practically all
Bloomberg has been trying $AMD and $INTC GPUs but is sticking with $NVDA for 95% of their workloads. Their H100 demand is not fulfilled, they are still trying to get the H100s they need. Expert made available via Tegus:
Some interesting analysis from TechInsights on SMIC's N+2 node (processor in Huawei's latest phone):
One of our salespeople in Asia was able to procure a Huawei Mate Pro 60 and hand carried it back to the lab in Ottawa.
When we analyzed the previous N+1 device we found
$NVDA's Jensen confirms the $CDNS and $SNPS bull case:
"But in the area of these tools, Cadence and others, they're going to build their own copilots, and they will rent them out as as engineers. I think they're sitting on a gold mine."
A key question for $NVDA is how long the current AI GPU boom will last. The company is currently shipping datacenter GPUs at a rate of around 3.5 million per annum. This translates into a penetration rate of only 13% of global cloud servers having one GPU. Ultimate demand will
How I'm currently thinking about investing in AI. Most investors are focused on the box on the left, the infrastructure buildout, however, long term the biggest value creators will be placed in the box on the right i.e. AI software. Similar to how the long term winners from the
$NVDA's CFO was interviewed by JP Morgan, very bullish outlook, highlights:
- PCs are now going to be a big part of that AI era in front of us. We are seeing a need to enable the developers that are working on AI. And NVIDIA RTX already is ready for AI for PCs. We have more than
$ASML CEO: "The semiconductor industry is currently working through the bottom of the cycle and our customers expect the inflection point to be visible by the end of this year. Customers continue to be uncertain about the shape of the demand recovery in the industry. We therefore
$AMD has successfully taken a share of the AI accelerator market for supercomputers (data from the Top500 list). With the launch of the MI300 in the second half of this year, this puts the company in a good position to take further share in the coming years. One example here is
A key reason why $MSFT is winning market share vs $GOOGL and $AMZN in the cloud - it's just cheaper. For example, to rent a Nvidia A100 40GB server in the cloud:
There's a lot in this - who I'd back in the HBM market, my thoughts on why AI clouds should be great investments, AMD's large opportunity no one is talking about (hint: it's not the MI300 or MI350), and KLA's coming semicap ramp!
$MU $AMD $NVDA $KLAC
My current portfolio.
Historically, the more hate a certain position gets from colleagues, the better it usually performs afterwards. The best ideas tend to be the contrarian ones where you have high conviction.
20% cash will be allocated based on research over the coming
The $SNPS + $ANSS combination will be one of the most attractive assets to own in the entire semi value chain (together with CDNS):
In foundry, we'll be looking at a battle between TSMC, INTC and Samsung at 2 and 3nm, and while it is likely that TSMC maintains a fairly dominant
$ASML is seeing a very strong cycle for memory orders, on a rolling two quarter basis this is easily illustrated, EUR 6.5 billion in orders over the last two quarters alone, highest ever basically in two quarters:
$META's engineering team disclosed details on their latest two AI training clusters, both make use of close to 25k $NVDA H100s. It's clear that there is a lot of both hardcore software and network engineering involved here, and that's not simply buying a bunch of H100s and
Why I expect $AMD to be able to take share over time in AI training vs $NVDA's current market share of >75%👇
PyTorch has become the dominant library to code AI models. This library basically provides Python objects and functions so that you can write all your needed AI code in
Both $SNOW and Databricks are winning share from traditional datawarehousing, but it's also clear a battle is brewing between the two. Databricks' advantage is that it is really built around data science and machine learning workloads, whereas SNOW is designed for traditional
Record quarterly bookings for $ASML driven by record memory bookings, logic also bouncing back, shares up 7%.
ASML CEO: "The semiconductor industry continues to work through the bottom of the cycle. Although our customers are still not certain about the shape of the
$LRCX has developed a tool that can disrupt Tokyo Electron's coating business and also lowers the need for EUV. CEO at Bernstein:
"Both in the foundry logic space and the DRAM space, I think many people are aware that the lithography is changing to EUV to be able to print
Fewer and fewer companies have been able to compete at the leading edge in semis over time - currently there are only two-three left: TSMC, Samsung and Intel (sort of). This consolidated landscape has resulted in tremendous margins for the dominant player i.e. $TSM.
Will
While both $CRWD and $PANW are the strong players in cloud security, CRWD has a competitive advantage as they are running everything from their single Falcon platform, whereas PANW has been stitching together its solutions in three platforms really. CRWD's CEO on the recent call:
TSMC price increases (5% for N3, 10-20% for CoWoS) are great for the entire semi value chain: Nvidia will pass these on to consumers - capacity is fully booked till '26 so the demand outlook is strong - while TSMC gets more gross margin oxygen to embark on N2 and CoWoS investment
Thoughts on tech valuations here:
1. A lot of quality software names are reasonably valued, and investors should be able to pick up long term compounders here. I've been adding to a number of them over the last few months.
2. A handful or so of high-growth software names are
Software is weak cause enterprises are focused on AI. No one wants to get disrupted so the key focus right now is to get AI right. So naturally there's much less focus on other software projects, e.g. to get new workloads up and running on a particular SaaS. As a result,
I actually used to be a $PYPL shareholder from 2016 to 2021, however it is hard to see what all the fuss is about now — classic value trap in my opinion and might make for a solid short.
$ADYEN's multiple has halved, is it time to go in? Find out 👇
$AMD unveiled their strategy to take on $NVDA in the AI data center, highlights:
First, Lisa Su went over AMD's investments in software to easily deploy AI models on their GPUs. This is one of the key areas where the company has been lagging due to NVDA's decades long
Some interesting comments from $VRT's CEO on AI datacenters:
- He sees AI penetration in the data center higher than 20% by 2028. "A lot of the data centers are today being designed for being hybrid and future ready."
- BOA analyst mentioned that $VRT's AI project pipeline
You already know that $NET is an interesting asset with their distributed network of edge servers from where they can run: a/ web app protection, b/ SASE and c/ developer services, i.e. a mini-Amazon AWS. However, this is all reflected in valuation and this isn't a cheap name..
$MSFT's Head of Cloud AI on AI demand:
"We look at demand in 3 different aspects. So the first is customer demand. Both from those that are coming in in a pilot phase and then is that pilot translating into an at-scale deployment. So they're not just experimenting but they're
$ARM will likely be IPO'd at around $53 a share.
- based on the latest reports from the likely market cap, i.e. $55bn, and the number of shares in the prospectus, 1.02 billion.
What do I think the shares are worth?
Maybe $16.5 if I'm generous.
I'll likely publish something
Interesting $NVDA analysis from Bernstein on the inferencing market:
Basically the cost of each ChatGPT query is about $0.04. If volume grows to one-tenth the size of Google search, that GPU deployment will cost $48 billion, with another $16 billion to maintain operations.
So
When looking at the semicap universe, a basket of $ASML, $AMAT, $LRCX and Tokyo Electron should be a good way to get quality investment exposure. Each of these holds strong market shares in key steps of the semi manufacturing process (chart below). The high-end, let's say below
An overview of the industrial analog semi space, $TXN vs $ADI, the advantages and risks for both these players, thoughts on China building up capacity, the analog cycle, plus a detailed valuation with some potentially attractive long-term IRRs.
$AVGO laying out the huge AI bull case - they're talking with customers about building one million GPU clusters.
"If we go back to 2 years ago, a cluster that was state-of-the-art at the time had 4,096 XPUs, each XPU was a couple of hundred watts, and to interconnect 4,000 of
I've seen a number of accounts lose money shorting tech last year, here's my methodology which should help you going forward:
1. Short companies operating in competitive industries or with a questionable product, and which are burning lots of cash and will run out of it within
Current valuations in semi land - next year's EPS growth vs PE next-twelve-months.
Naturally cycles matter, the market thinks for some of these names we're heading towards peak eps e.g. $NVDA, whereas for others the market reckons we're going through a temporary lull e.g. $ASML
Good presentation from $INTC's head of Foundry at the Goldman conference. This is the most interesting story for Intel, as the company has been bleeding share in the datacenter and on the client side. The success of the company will really depend on the success of this business,
Bullish $ASML conference call:
- They think they might too little EUV capacity due to the strong orders coming in from DRAM and HBM.
- Total backlog is already 1.44x consensus '24 sales. And much higher once you strip out services & consumables revenues.
- High-NA to ramp,
My thoughts on $INTC:
The outlook for Intel is definitely much better than it was a few years ago. Now that the company is competitive again on process technology, there is potential for the datacenter business to improve, both on the CPU and GPU sides, as well as in networking.
Fintwit accounts that are desperately trying to find accounting shenanigans in certain $NVDA contracts are missing the forest for the trees - we are in the midst of a major GPU capex boom where everyone has been scrambling to get their hands on NVDA GPUs. It's very easy to do
Why $TSM's conf call was bullish for $ASML and semicap, multi-year perspective:
- Due to AI, more demand is moving to the leading edge node. Historically $AAPL was the main customer at the leading node, but in the future AI accelerators should bring in a large chunk of demand as
$TSM vs $INTC - TSMC CEO saying they will outperform Intel's 18A with their N3P already: "Our internal assessment shows our N3P demonstrated comparable PPA to 18A competitors' technology but with an earlier time to market and much better cost. Our 2-nanometer technology without
Semicap is currently running on a China sugar high as the Chinese are rushing to get as many tools into the country as humanly possible..
$ASML $AMAT $LRCX $KLAC
Looking at the market for semi IP licenses last year, $SNPS took the top spot, beating ARM. Also interesting that $AWE (UK) is now close to a 5% market share, catching up with $CDNS.
ARM still leads in terms of royalty collections with a 61% share, although they lost a 7% market
$ANET remains very bullish on ethernet vs $NVDA's infiniband:
"In a recent blog from one of our large Cloud and AI Titan customers, Arista was highlighted for building a 24,000 node GPU cluster based on our flagship 7800 AI Spine. This cluster tackles complex AI training tasks
Some thoughts on a $SNPS - $ANSS combination:
$ANSS used to be mostly a simulation software company for the aerospace and automotive sectors, however, quite a large part of their business is now High-Tech, i.e. leading edge semiconductors. Their CEO has been talking up the role
Some of $ASML's secret weapons:
- Owned Cymer manufactures the EUV laser
- Jointly owned Carl Zeiss SMT and owned Berliner Glass provide the optics for the EUV tools
- Holistic lithography: enables a continuous feedback loop of metrology measurements into the litho tools to
Good presentation from $AMD's head of datacenter at UBS. Some highlights:
On their current market share in datacenter GPUs, the UBS analyst mentioned: "The numbers we keep hearing is that you're roughly 10% unit wise of what $NVDA is." This would be a very promising launch of
One analyst on the $AMD call: "A lot of folks that we talk to think that the compute game is shifting completely from CPUs to GPUs.."
This is a complete misunderstanding of basic computer science - for the running of all conventional code, the CPU will remain king here. Only for
$NVDA's head of healthcare gave a detailed overview of the coming biological computation market at the JP Morgan conference, and it sounded like a big opportunity, both for NVDA and the biological CAD/ simulation software industry:
"So now for the first time, we can represent
Why I'm still long $NVDA:
1. TSMC said their revenues from AI accelerators will more than double in '24
2.
@elonmusk
confirms - chart below
3. Stock is on 33x forward PE
$MDB's new vector database had the highest net promoter score in a survey from Retool among 1,500 tech workers, and was the second most widely used behind Pinecone.
Vector databases are going mainstream as they allow for current high-dimensional data to be fed to a trained LLM,
$NVDA's detailed datacenter roadmap below. With their move to a one year innovation cadence, it will be difficult for competition to get a foot in the door in the coming years. Next year we have the H200 and GH200, followed by the B100 and GB200 shortly thereafter. With a new
$KVYO is a typical piece of business software that's just executing extremely well. It reminds me a lot of $HUBS in '16, which turned out to be a 10-bagger.
And on a lot of metrics, the company is actually showing better potential than HUBS. One of the most attractive features
I'm liking a number of quality software names here, even if AI reduces the number of seats, these are companies with strong pricing power with plenty of scope to raise pricing over time. Additionally, quite a number of them have been talking about building in automated workflows
A full model on the $SNPS - $ANSS merger, working all the way back to how this will impact EPS for SNPS shareholders. IRR of 22% working from Wall Street consensus estimates, which I think should be do-able and beatable, and then modelling in some guided synergies.
Currently,
It's clear that $SNOW should be heading into a strong cycle, if they don't, I can only conclude that they're losing share to the likes of Databricks, Microsoft Fabric, Google Bigquery etc. Here's why the cycle should be strong:
$MSFT's Head of Cloud AI today: "Customers are
The $CDNS new emulation tool sounds seriously cool:
"So Z3 itself, we designed this advanced TSMC chip ourselves. And this is one of the biggest chips that TSMC makes. And 1 rack will have like more than 100 of these chips and then we can connect like up to 16 racks together. So
Chinese semicap players now have a near 50% market share with the Chinese foundries. Orders will be coming from the mature nodes, but nevertheless an interesting datapoint that they're making progress.
$AMAT $LRCX
My conclusion after having a fresh look at $U:
Overall, we have a new CEO and one with a good track record, who seems motivated to turn this into a successful business like he did with Red Hat Linux. I think it is likely that the board will offer him a permanent CEO role at the
Fortunately $TSM's fabs are on the west side of Taiwan and can resist shocks of magnitude 7. The impact of the shock at the fabs had already decreased to magnitude 4-5, so the fabs should be fine but the issue is that the wafers can be damaged. The processing time for wafers is
Relevant for $NVDA $AMD, Baidu says they're mixing GPUs in a training cluster:
"Leveraging our technical expertise, we can now
integrate GPUs from various vendors into a unified computing cluster to train an LLM. Our platform has demonstrated very high efficiency with this setup
So $ASML and $AMAT are telling us they're seeing a boom in lagging capacity in China whereas SMIC has been buying a bunch of tools to do quadruple patterning at 7nm. Well done
$ARM is telling investors they can grow double digits again, cause datacenter + AI, and also with price increases.
1. ARM doesn't have much weighting to leading edge CPUs (less than 10% of royalties), so if you're looking for exposure to AI inferencing (which comes after the
Decent orders quarter for $ASML at EUR 5.6 billion, mainly driven by Logic. Will write a post later on whether numbers like these are sufficient.
CEO's comments:
"Now I also want to share with you some important update about our EUV technology. First of all, 0.33 NA systems.
You already know that $NET is an interesting asset with their distributed network of edge servers from where they can run: a/ web app protection, b/ SASE and c/ developer services, i.e. a mini-Amazon AWS. However, this is all reflected in valuation and this isn't a cheap name..
All sorts of $NVDA GPUs are still being smuggled into China. Very interesting article how agents are buying up RTX 4090s across Taiwan, Singapore, Vietnam etc to resell them into China. However, even H100s are still available in Hong Kong on the black market, so it shouldn't be
Semi and AI news:
- $NVDA is testing Samsung's 12-layer HBM3E, SK hynix will ship samples later this month
- SK hynix raises DRAM prices again with 15 to 20%
- $IFX will supply Xiaomi's EVs with SiC power semiconductors and MCUs
- $MSFT is training an in-house LLM to compete