AI is on Fire—New Broadcom AI Fabric Aims to Stoke the Flames
The Futurum Group’s CEO Daniel Newman dives into the recent announcement coming out of Broadcom. The Broadcom Jericho3-AI, introduced this week, provides a high-performance ethernet for a 32,000 GPU cluster. Newman thinks this will hopefully improve workload balance and performance for AI operation.
Cite this

The News: Broadcom Inc. says it’s created the highest performance fabric for use with AI networks. The Broadcom Jericho3-AI, introduced this week, provides a high-performance ethernet for a 32,000 GPU cluster. This will hopefully improve workload balance and performance for AI operation. Read the full Press Release from Broadcom.
AI is on Fire—New Broadcom AI Fabric Aims to Stoke the Flames
Analyst Take: While the announcement of fabric supporting AI functions doesn’t necessarily get the fanfare that ChatGPT and other generative AI feats do, for those of us involved in tech, Broadcom’s announcement is an important one. Without the foundation to support the potential of AI, there is no room for AI to flourish. Make no mistake: what Broadcom claims about its solution–perfect load balancing of traffic throughout the fabric, ultra-high radix that scales connectivity to 32,000 GPUs, Zero-Impact Failover for sub-10ns automatic path convergence, and faster job completions (10 percent shorter overall)—is something that every AI developer, investor, and creator is excited about.
Setting the Stage for this Announcement
Despite the name, Jericho3-AI is not an AI-powered fabric. It simply enables the AI magic to happen more quickly and easily than other options already on the market (think NVIDIA’s Infiniband for AI). The fabric would be used in places like data centers and cloud environments where complex AI models get trained and pushed out. With global AI spending expected to hit $300 billion by 2026, that’s a good thing. Reports show Jericho3-AI features 26 petabits per second of ethernet bandwidth. That’s almost 4x the previous generation. It also features 40 percent less power per gigabit. Broadcom says it will reduce job completion times for just about any AI or ML job. Lower cost of ownership, faster training for AI models—this is the equation for AI success. And regardless where you stand in the tech market—end user or developer—you’re going to enjoy the fruits of Jericho3-AI’s labor.
A Broader Look
In many ways, Broadcom’s announcement isn’t just about Broadcom’s Jericho3-AI. It’s about the industry recognizing that there’s room to improve and grow when it comes to AI development. Again, not as sexy as a new generative AI app. But without these foundational technologies, generative AI itself wouldn’t likely exist.
Indeed, behind every OpenAI is other technology supporting those complex AI models. What I expect to see in coming months, just as we saw with the generative AI war happening now between Microsoft and Google, is for other—perhaps even more novel—solutions to pop up in this space. The race for AI is already on. The company that can move a developer ahead in that race faster, cheaper, and more efficiently is going to win it. As of now: Broadcom is holding the baton. My question: who will be next?
My co-host of The Six Five Webcast, Patrick Moorhead of Moor Insights and Strategy and I discussed Broadcom and AI in our latest episode. Check it out here and be sure to subscribe to The Six Five Webcast so you never miss an episode.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other insights from The Futurum Group:
Microsoft and Epic Partner for OpenAI EHR Integration
Subscribe to The Weekly
The week's market signals, one email, no filler. Unsubscribe any time.
More from Daniel Newman
The Foundry Bet: IonQ, SkyWater, and the Industrialization of Quantum
Futurum’s Daniel Newman shares his insights on why quantum’s next phase is an industrialization race, why the SkyWater acquisition puts IonQ in the top echelon, and why the bet still has to be proven.
The Token Cost Reckoning Arrives in the CFO’s Office
In our latest thought leadership brief, The Token Cost Reckoning Arrives in the CFO’s Office, completed in partnership with Google Cloud, Futurum Research examines how enterprises can track cost per successful agent run and cut token spend before their next budget review.
AI Decision Intelligence: The Category That Replaces Legacy Research
Futurum CEO, Daniel Newman, shares that Futurum Forward is the first platform purpose-built for the AI enterprise cycle. Replacing legacy research, it combines a proprietary data spine with a productized operator model for real-time intelligence.