Transcript
0:01 So the accelerator model is one of our oldest products actually. You know the first thing I worked on when I joined analysis in short what it does is it really track all the shipments of nearly all of the AIO it is in the world definitely all the main ones. So from Nvidia and AMD GPUs to all the main AC programs like TPU tranium open AI and you know Microsoft there's a lot of chip startups that are coming up. we have several people distributed across Asia.
0:30 I'm based in Japan. I come from more of a financial analyst background. We have several people in Taiwan who are very close to the supply chain people in Singapore. We have a whole team now in Soul Korea. Some of them have been working with the supply chain directly such as we have next and Samsung employees working with us. So for each accelerator we track units, the ASP from the vendor, the specs, how many flops, power, what the silicon looks like, what's the HPM content, and then we flow that to front end logic wafers, packaging demand like co wafers, how many bits and cubes of HVM, and then we also bring it all the way down to the number of systems or racks that are coming out of the ODMs and the OEMs.
1:16 What we detail in our notes is basically anything material that we think changes the landscape. One example recently is just how updates are on what's happened with Ruben Ultra's specs. especially the downgrades in HPM content compared to people's previous expectations. So if you're a memory company, you're thinking, oh okay, if Nvidia needs less HPM, this impacts our forecast of bit demand. So financial clients will first use this to drive their predictions of what is Nvidia revenue, what is a boardcom's ASIC revenue. For instance, our most popular feature is our unit and ASP forecast. A unit time ASP gets you revenue.