John Abbott is a principal research analyst in the Applied Infrastructure & DevOps channel of 451 Research, a technology research group within S&P Global Market Intelligence. He covers semiconductor, systems and software infrastructure. This includes new cloud and datacenter system architectures (scale-up and scale-out), accelerated computing and AI infrastructure, semiconductor ecosystem and supply chain issues, and legacy-to-modernization platform evolution. John arrived at S&P Global Market Intelligence through its 2019 acquisition of 451 Research. He was a cofounder of The 451 Group in 2000, running analyst operations from the San Francisco office until 2007. He has been the principal author of many special reports on areas such as storage virtualization, blade servers, software- defined infrastructure and virtual desktop infrastructure. John began covering the technology industry in 1984, building on work as a user of mainframes, early PCs and Unix machines, and then becoming a technical author. As a freelancer in the 1980s, he contributed to BBC World Service, Computing, Computer Weekly, the Financial Times and the Times. He worked on the pioneering daily IT news service Computergram International and became its editor in 1997. John holds music and English degrees from Keele University and a master’s degree in modern English literature from the University of London.
The Stargate Project intends to invest over $500 billion over the next four years to build new AI datacenters in the US for OpenAI. SoftBank takes the financial lead and OpenAI will have operational responsibility.
The company straddles many technical and geopolitical borders: consumer and datacenter/enterprise markets, original design manufacturer and branded hardware products, and its operations in China and the rest of the world. With worldwide PC sales falling, Lenovo is pinning its hopes on AI acting as an engine for a new refresh cycle.
The target has been developing a software stack that serves models and accelerates AI inferencing on CPUs and GPUs in the cloud, on-premises or at the edge. It has also taken the lead on commercializing the open-source virtual large language model project from the University of California, Berkeley.
The two long-term rivals want to boost performance and compatibility across the x86 ecosystem. Billed as intended to "shape the future of the world's most widely used computing architecture," the alliance is indicative of the competitive pressure that the x86 ecosystem now faces. This collaboration between Intel and AMD should simplify development and speed up innovation.
High-end AI infrastructure is complex to deploy. With the pickup of ZT Systems, AMD is obtaining a 1,000-strong team of server design engineers to help its customers deploy AI at the system, rack and datacenter layers. The target is a 30-year-old systems specialist whose largest clients are the hyperscalers.
Building multiple generations of AI acceleration silicon is an expensive business, and UK-based Graphcore was running out of financing options. SoftBank intends to operate the target as a wholly owned subsidiary, but might also use its technology internally.
New Alert Set
"My Alert"
Failed to Set Alert
"My Alert"
Here are a few tips for navigating our Research.
Quickly access any of our research and data products.
Browse the research & data that aligns with your specific interest.
Discover new and more flexible ways of finding relevant content.
Set notifications for the topics & companies you care about most.
Access any page in the dashboard from here.
Hand picked content for you and your team.
