“There is some concern that this extension of NVLink into the x86 domain could replace the traditional PCIe connection between the x86 CPU and NVIDIA GPU, which would impact ALAB given its meaningful PCIe content. However, we would hesitate to draw firm conclusions at this stage.”
But, people are getting ahead of themselves. The article points out that anything produced from this new deal are at least a couple years away from hitting the market.
For me, I still believe those that don’t want to also rely on Nvidia’s networking will stick with PCIe as long as the performance is there. I’m considering adding to my ALAB position.
On August 28th ALAB CEO Sanjay Gadendra spoke at the Deutsche Technology Conference. One of the things he touched on is how hyperscalers are supportive of open standards, like PCIe and the standard being put forward by UA Link consortium, of which CEO Gagendra is a member. At the Deutsche conference he said he believes that hyperscalers “don’t want to be held hostage” and “no one wants to be … restricted in some ways of how they do their business. I mean, today, if you think about, there are a couple of ways in which they’re sort of in a jail, right? There is the vertical integration that Nvidia is pursuing and then you have folks like Broadcom that are trying to provide a certain type of solution that combines with their custom ASICS and networking and all that.” Gadendra concludes by saying: “We truly believe that there is a third way of achieving this, which is to provide an open rack standard that services the nuances of various different use cases. And that’s clearly where we see the momentum right now and that’s what we’re focusing on.” In other words, Nvidia and Broadcom are putting forward proprietary protocols that, if adopted by a hyperscaler, may impinge on their freedom to select hardware by other vendors that is incompatible.
Later in that talk, Gadendra said that: “The Scorpio X series is something we’re super excited about. It’s going to completely transform the company because both from a business standpoint and also because these are anchor sockets, meaning if you are a system engineer at a HyperScaler, there are two devices you will care about during the first five minutes of starting a new design. One is the GPU, the other one is the scale-up switch, right? And then everything else is designed around it. For us, Scorpio X series makes us part of the conversation literally from the first meeting and right now we are having these conversations looking at not just the immediate generation but for the next two or three generations with our hyperscaler customers.”
I think both of these statements speak to the advantages ALAB has in pursuing ever greater “real-estate” in AI racks. But it also speaks to what competitors are doing to deprive ALAB, and other competitors, from getting a toe hold.
On a related topic, I posted on the NBIS thread about a series of questions that NBIS investor “Markos” is preparing for NBIS management to get a better picture of MSFT contract terms: https://x.com/MVanbrunsc47513/status/1968703739063320818. One of his questions: “2.Hidden Handcuffs → Why would Microsoft help a rival? Here I want to look at the asymmetry: what subtle constraints or “validation traps” might be in that make Nebius useful to Microsoft but keep it strategically boxed in.” You can follow that link to get full the picture, but suffice it to say he has concerns that MSFT will try to box in NBIS to prevent them from poaching MSFT business.
So, fasten your seat belts, there is a lot of maneuvering as competitors attempt to integrate vertically to claim more share in AI Racks and in managing AI Workloads.
My take here is that the hyper scalers want Astera to win out in this networking space because they are proposing an open standard through UALink. This allows the hyperscalers to customize chips how they want and not be locked in on another company’s platform. Broadcom on the other hand is proposing a closed standard, and NVIDIA has their own standards as well. What the hyperscalers are seeing is if Astera doesn’t succeed they may be caught between NVIDIA and Broadcom dictating pricing for every component of their systems.
What’s good for Astera is that while their COSMOS software does not have explicit lock-in, a lot of engineers are trained up on this system. Launching new products can connect seamlessly to the COSMOS platform. I’m also really liking how huge the vision the management team has, “AI infrastructure 2.0”, “The mothership of the data center”, “the nervous system of the data center”. Just stepping back for a second, just think how much should a company be worth that is the “nervous system” of the data center.
Another bit of good news for Astera investors is that the company is no longer reliant purely on PCIe technology for all their revenue. When they IPO’d it was 85% PCIe sales through Aries, and it was a real risk to be this concentrated on one product. Now they are saying the Switch Scorpio product is a bigger TAM than Aries. Additionally, Credo just confirmed they are not getting into Switches. Management for Astera keeps using the term “greenfield TAM”, and that’s really what it is here. Maybe Intel and NVIDIA will build a better standard in PCIe, but I wouldn’t be surprised if Astera is one or two steps ahead of them on some new products as well.
Interesting side note here, is that Morgan Stanley was the lead underwriter on their IPO, and Intel was a lead pre-IPO investor at 3.6%
Morgan Stanley has a pretty poor track record calling the shots on Astera. They were convinced back in 2024 that Blackwell would be a disaster for Astera and lowered their price target from $85 to $55. The stock subsequently went on a run up to nearly $150. It does seem like Morgan Stanley has a wide reach, when they publish these takes the stock does trade down heavily. Historically, after publishing their bear cases, Morgan Stanley had to raise their price targets after the stock rose.
It seems odd to me that their lead underwriter from two years ago is the main institutional promoter of bear cases against Astera. My bet is with the Astera management team who see a clear path to success in 2027, versus the analysts over at Morgan Stanley who have been wrong consistently regarding Astera.
ALAB is trading >20% below where it was a week ago. The most obvious reason is that it got ahead of itself and that this is a normal reaction. However, I wonder if there is concern that NVDA will steer business toward it’s own proprietary NVLink gear as part of it’s $100B investment in Open.Ai, and away from outfits like ALAB who sell gear based on PCIe and UA Link that are open architecure and thus nonproprietary.
Based upon the ALAB expanding TAM, I purchased a little more ALAB to replace my CRDO which I sold this week up 200%. CRDO is fine, I just like the ALAB new TAM and broadening product line and positioning over CRDO. I believe this AI build out is at least another 1-2 years before maturing. And ALAB is the most perfectly placed. One thing that lets me take on this risk is that I am up 400% on my early ALAB purchase and up 200% on my later purchase. So I am riding a lot of buffer on this high risk. There is a lot of competition coming from NVDIA (already here), Marvel, and probably Broadcom. So my eyes are open on the next ER. Here is the published 2028 TAM.
PCIe Retimers $1.5B
Ethernet Retimers $1.5B
Leo CXL Memory Controllers $4B
Scorpio P and X Fabric Switches $5B
Astera Labs initially estimated the total addressable market (TAM) for its entire Scorpio portfolio to be about $5 billion, split 50/50 between the P-Series and X-Series. Currently, Scorpio’s TAM is over eight times the company’s second-quarter trailing 12-month (“TTM”) revenue of $605.55 million, representing a potential massive needle-moving growth driver over the next several years if it becomes a dominant player in the Smart Fabric Switches market. I am a big believer in Scorpio moving the needle.
Astera Lab CEO Q2 Transcript
The second step that we hit was to expand our PCIe retimer and Ethernet retimer business to go after custom ASICs. So this transition happened in Q3 of last year. Now where we are is our third step in that growth journey where we are ramped up our Scorpio P-Series, PCIe-based switch products, along with our Aries 6 retimers. So that’s going on all the third-party NVIDIA-based GPU platforms that are ramping.
The fourth step that we are highlighting as part of the call today is the Scorpio X-Series which is designed for scale-up networking, and that transition is currently underway in the sense that we are still in preproduction. And like we highlighted, throughout 2026, we expect that wave to transition to high-volume production, providing us a new baseline for revenue. And these are, of course, higher value sockets, meaning the dollar content with the Scorpio X-Series switches are significantly higher than what we have done so far. So you could expect that to play into the overall revenue projections that we would have as we get towards 2026.
Regarding the selloff in ALAB over the last week, a few stories are attributing to Morgan Stanley "cautious language,” saying that the Nvidia:Open.Ai transaction “could have major implications” for ALAB. This seems like Deja Vu All Over Again, as Yogi Berra once famously said. Morgan Stanley seems to have an axe to grind. But a broken clock shows the correct time twice a day, so maybe they’ll get it right this time. Sorry for all the mixed metaphors!
I came across something from the last earnings call for ALAB that I missed,
We joined AMD on stage during their Advancing AI 2025 keynote presentation as a trusted partner to showcase UALink, which is the only truly open memory semantic-based scale-up fabric purpose-built for AI workloads.
AMD recently fully endorsed UALink as it’s scale-up networking architecture of choice for all future generations of it’s rack-scale solutions, and we know of at least 1 other ASIC XPU vendor that’s going to be moving to UALink as well.
This confirms that AMD is committed to using UALink exclusively and maintaining the open standard. I do not think this implies 100% of UALink products for AMD will be Astera, but this sure sounds like they are teaming up in a big way. This also sounds like the most realistic path AMD has in hopes of catching up to Nvidia on some aspects.
Also, Astera mentioned on the last call the market for ASICs is bigger than NVIDIA related products for Astera. Assuming a worst case scenario, where NVIDIA + Intel build a dominant product, the ASIC market is still wide open for Astera.
And the fifth step that we called out as part of the communication is the UALink, that is going to be a growth story in 2027, and that is a greenfield application for us with a much broader deployment of scale-up networking along with a variety of other products that we intend to develop for UALink. And that is going to be the fifth step that we are executing towards.
Sorry, I did not include the 5th step identified by the Astera President and COO Sanjay Gajendra. I did not see it as material to my current position in ALAB. Note the growth story for UALink is 2027. That simply is not in my time horizon for owning the stock now. ALAB is the leader of the consortium standard and there is a lot of promise, but who knows.