The networking giant's new 7060XE7 series targets AI clusters requiring hundreds of thousands of processors, and the infrastructure spending wave has implications far beyond traditional tech.
Arista Networks just dropped what might be the most consequential networking hardware announcement of the year. The company unveiled its 7060XE7 Series of 1.6 terabit networking platforms on June 9, purpose-built for the kind of rack-scale AI infrastructure that the industry’s biggest players are racing to deploy.
The 7060XE7 Series represents a generational leap from the previous 800G standard. The initial models feature 64-port air-cooled switches designed for AI clusters that may require hundreds of thousands of XPUs, the broad category of accelerator chips that includes GPUs and custom AI silicon.
The first units are scheduled to ship in Q4 2026. Both Meta and Microsoft have effectively endorsed Arista’s approach, which centers on what the company calls its Etherlink AI fabric — a networking backbone specifically designed so that thousands of AI chips can talk to each other with minimal delay.
The larger port capacity also means higher potential revenue per unit. When you’re selling 64-port switches instead of lower-density alternatives, the math on average selling price gets more interesting for shareholders.
Arista isn’t making this bet from a position of weakness. As of March 31, 2026, the company reported approximately $2.79 billion in cash on hand and $9.56 billion in marketable securities. Total liabilities stood at roughly $8.17 billion.
The AI infrastructure buildout is the single largest driver of energy demand growth in the US right now. Every new AI data center that Meta or Microsoft stands up competes for the same power grid capacity that Bitcoin miners and proof-of-work operations depend on. That competition for power has already started reshaping where Bitcoin miners can operate profitably.
The semiconductor supply chain is another intersection point. The XPUs that Arista’s switches connect are manufactured by the same foundries, primarily TSMC, that produce chips for crypto mining hardware. Massive AI orders can create ripple effects on lead times and pricing for other chip customers.
The Q4 2026 ship date is the first milestone that matters. The relationship with Meta and Microsoft deserves close monitoring, as these hyperscalers represent an outsized share of potential demand for 1.6T networking gear. Competitive dynamics are also shifting, with Broadcom, Cisco, and several smaller players all pursuing the AI networking opportunity.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our