Bending Metal: The Future of Taiwan’s Electronics Complex
In all the focus on the big processors and the “AI race”, we often overlook the far less glamorous business of actually building compute systems. We have explored different parts […]
Category Archives: Enterprise Electronics
Bending Metal: Combinatorics
Buyers want an infinite variety of servers, but no one is willing to pay for infinite designs. The costs of designing servers is a major barrier to entry.
What do the Hyperscalers Want?
In the semis landscape, when we talk about the hyperscalers we tend to focus on chip performance and power. Instead we should look at what problem they are really trying to solve with silicon.
Bending Metal: The Changing Server Business
Processors only make up ~20% of the cost of a server. We take a look at the business behind the rest of the bill of materials.
HPE + Juniper = ?
HPE is acquiring Juniper in a deal that is hard to parse. Juniper has been stuck on a treadmill for so long, and its business model look fated to resemble HPE’s more and more. Better to jump then get pushed.
Flying Above the Clouds
Many semis start-ups pin their hopes on non-Cloud customers. This is a large market, but also a conservative one that is reluctant to adopt new chips. The move to AI, data sovereignty and a growing awareness of cloud costs may crack that door open, a bit
Data Center Market Math
The data center silicon market is massive, but also challenging for incumbents let alone new entrants. We do some math to back this up.
Nokia and Ericsson: Adrift
What role is left for Nokia and Ericsson? Are they systems integrators? Specialty equipment vendors? Enterprise mobility solution provider? None are great options.
The Sixth Sense – We See Data
Moore’s Law brings benefits beyond semis. The price and accuracy of sensors are both moving in the right direction allowing for fine-grained pervasive sensor networks, just as advances in AI mean we can make real use of all that data.
Has Nvidia won the market for AI Training?
It seems very likely that most AI Training will be run on Nvidia GPUs. The software environments for training are too numerous and changing rapidly, which favors the soft, warm familiarity of CUDA over speculative gains offered by new entrants.