People have asked how they can support our work. Our monthly Nanalyze Premium subscription is our version of Patreon ;) www.nanalyze.com/subscription-plans/
IT guy here. Yes, longer cables are always welcome. Cables are used to connect storage arrays to compute arrays, which are then connected to switches, and then to other devices downstream, like backup appliances and such. Sometimes these devices are in the same rack, but more often then not, they're not. Usually we need cables up to 20 feet, occasionally longer. Storage and compute hardware has a 3 to 5 year life cycle. When those devices get replaced, the cables get replaced as well, and they're usually supplied by the reseller. And that's where the real money comes in. As an example, Cisco charges TENS OF THOUSANDS for their Cisco branded cables. Even though the generic versions of these cables can be bought for pennies online, Cisco will only support you if you buy Cisco branded cables. If Astera Labs wants customers to use their cables over the inevitable knock offs, they'll need to create the need.
Astera have a great team of top engineers from Marvell and Arista and did churn out some very interesting products so far for datacenter connectivity. When I was working on these topics in Huawei, our main competitors were Marvell and then Arista, so I assume these guys have the smarts and the product market fit to drive some good business going forward. I like what the gentlemen at Kandou bus are doing as well.
@@NanalyzeI am not a datacenter architect, but I do have one comment on the cable length issue . As most modern setups are TOR (with switch on top of rack), the cables have to reach all the way to top with same datarate. And that is where passive coaxial high-speed ethernet cables fail, and where much more expensive and power consuming (active) cables come in. If one manages to make double the length passive and robust ethernet cables, or replace active optical cables with active copper cables, given the volume, such a company hits jackpot.
So Astera Labs competes with NVDA Infiniband for connectivity solving the data throughput problem for other GPU producers? Or are they complimentary with NVDA networking? I will look at it. I only recently did enough research to understand how Infiniband helped NVDA leap ahead of its competitors in the GPU data center business. Before I found that info, I could not understand how NVDA leaped so far ahead.
Good question. They don't cite NVIDIA as a competitor in their S-1, instead talking about "collaboration with data center infrastructure suppliers such as NVIDIA."
Short answer, no. You can always torture the data and make it say whatever you want. We did an academic paper on our Quantigence strategy and it spanked the S&P 500. We'll be releasing videos on this.
As a stock investor in Taiwan, it's crucial to have a deep understanding of the characteristics and operating models of the Taiwanese electronics industry to make informed investment decisions. In terms of Taiwan's electronics distribution agents, they typically do not represent multiple brands of homogenous products, especially when it comes to key components. In other words, brand manufacturers do not allow their distributors to sell products from competing brands.
@@Nanalyze some initial questions: - why forward face and be accountable to shareholders? - is this a FOMO grab for an overpriced IPO? - is capital raises too expensive to their bottom line? - is the IPO a way to fund the development of alternate glass product options? - are they one big contract pivot away from severe losses? - looking at their literature, it looks really easy to update their documentation with the word AI.
Well, plenty of people are asking about that. These companies are probably cramping PANW's style: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-tOX3fs7mHPI.html
@@Nanalyze They have been using chips like that for ethernet cables for years, It seems more like a commodity product and won't have much staying power.
