Hi All,
Welcome to our brief overview of portfolio news from the past few days.
a. OpenAI’s Custom Silicon Strategy Takes Shape with Broadcom and Arm
The AI arms race continues to accelerate, with OpenAI announcing a strategic collaboration with Broadcom to deploy 10 gigawatts of custom AI accelerators. This multi-year plan targets initial deployments in the second half of 2026.
This partnership marks a major strategic evolution for OpenAI, moving it from a pure consumer of AI chips to a designer of its own custom silicon. By designing its own hardware, OpenAI can optimize performance for its specific models, potentially unlocking new capabilities.
Furthering this strategy, reports by The Information, indicate that OpenAI is also in talks with Arm Holdings to use a custom Arm-designed CPU for its server chips. This move suggests two key motivations. First, it shows a clear intent to reduce long-term reliance from Nvidia and AMD, giving OpenAI more control over its hardware roadmap and supply chain. Second, given the massive compute deals already announced, it also suggests that the existing market capacity may not be sufficient to meet OpenAI’s vast needs.
These moves highlight the industry’s increasing reliance on OpenAI’s strategic decisions, creating a potential vulnerability where the direction of the entire AI ecosystem is heavily influenced by the actions of one company. It also reinforces the central question: how will all of this compute capacity be financed?
b. Google Launches Gemini Enterprise to Compete with Microsoft
Google Cloud has officially launched Gemini Enterprise, its comprehensive AI platform for the workplace. The core value proposition is to offer a single, unified platform that acts as a “front door for AI” for every employee, connecting to a company’s data, workflows, and applications. As Google Cloud’s CEO Thomas Kurian puts it, while some competitors offer “the pieces, not the platform,” Gemini Enterprise is designed as a complete, end-to-end solution. This launch places Google in direct competition with Microsoft’s Co-pilot for the enterprise AI market.
c. AMD
Deepens AI Moat with Oracle Partnership
AMD continues its strong momentum in the AI data center market, announcing an expanded partnership with Oracle. Oracle Cloud Infrastructure (OCI) will be the launch partner for the first publicly available AI supercluster powered by AMD’s upcoming Instinct™ MI450 Series GPUs. The initial deployment will consist of 50,000 GPUs starting in the third quarter of 2026.
This deal is another win that strengthens AMD’s competitive moat. By securing large, multi-year commitments from key cloud players, AMD is locking in future revenue streams and validating its technology.
“Helios” Open AI Platform
At the OCP Global Summit 2025, AMD unveiled “Helios”, a rack-scale AI reference platform built on the Open Rack Wide (ORW) standard contributed by Meta. The design extends AMD’s open hardware philosophy from silicon to system to rack, combining Instinct™ GPUs, EPYC™ CPUs, and Pensando™ networking with open standards such as OCP DC-MHS, UALink, and UEC architectures.
Showcased as a reference design, Helios aims to help OEMs, ODMs, and hyperscalers adopt and customize open AI systems more efficiently. By promoting interoperability and scalability over proprietary alternatives, AMD positions Helios as a strategic step toward an open, high-performance ecosystem for next-generation AI infrastructure.
d. Other Industry & Portfolio Updates
Alphabet’s Data Center Investments: Google plans to invest $15 billion over five years to establish an AI data center hub in India and $9 billion through 2027 to expand its data center presence in South Carolina.
Amazon’s Strong Holiday Season Outlook: Amazon announced it is hiring 250,000 workers in the US for the upcoming holiday season. To put this in perspective, this single seasonal hiring push in the US represents ~16% of its global full and part-time workforce, which stood at roughly 1.5 million as of Q2 2025. This indicates for a healthy holiday shopping season.
That's a wrap. See you soon.