In a move that could reshape the data centre landscape for years to come, Meta and NVIDIA have announced a sweeping multi-year, multi-generational strategic partnership to build what they describe as the world's most advanced AI infrastructure.
The centrepiece of the announcement: Meta has become the first hyperscaler to deploy NVIDIA's Grace CPUs as standalone processors in production data centres - a landmark moment that signals NVIDIA's ambitions extend far beyond GPUs.
The Scale of the Deal
The partnership encompasses millions of NVIDIA processors, including Grace CPUs deployed now, Blackwell and Rubin GPUs, and future Vera CPUs expected in 2027. The financial scope is estimated in the tens of billions of dollars.
To put this in perspective, Meta has committed to spending $115 to $135 billion on capital expenditure in 2026 alone, with this NVIDIA partnership representing a major portion of that investment.
"No one deploys AI at Meta's scale - integrating frontier research with industrial-scale infrastructure to power the world's largest personalisation and recommendation systems for billions of users." - Jensen Huang, NVIDIA CEO
Why Grace CPU Matters
The NVIDIA Grace CPU is not just another chip. Built on 72 Arm Neoverse V2 cores running at 3.35 GHz, it delivers specifications that challenge the established server CPU market:
- Up to 480 GB of memory per standalone processor (960 GB in Superchip configuration)
- Up to 1 TB/s memory bandwidth between two dies
- Up to 2x performance-per-watt improvement on backend data centre workloads
That last point - performance per watt - is arguably the most significant. In an era where data centre power consumption is a primary operational concern, achieving twice the workload throughput for the same energy cost fundamentally changes the economics of hosting infrastructure.
Going Against the Grain
What makes Meta's decision particularly noteworthy is that it runs counter to the prevailing industry trend. Amazon has its Graviton processors, Google has Axion, and Microsoft is developing its own custom Arm chips. These hyperscalers have invested heavily in designing proprietary silicon.
Meta is taking the opposite approach - choosing to buy NVIDIA's off-the-shelf Arm CPUs rather than designing their own. This vote of confidence validates NVIDIA's strategy of competing not just in GPUs but across the entire data centre compute stack.
What's Next: Vera and Beyond
The partnership doesn't stop at Grace. Meta and NVIDIA are already planning for the next generation:
- Vera CPU - NVIDIA's next-generation processor featuring 88 custom Arm cores, expected for large-scale deployment in 2027
- Spectrum-X networking - NVIDIA's Ethernet networking switches integrated into Meta's open switching platform
- Confidential Computing - Enhanced privacy features for WhatsApp's private processing infrastructure
- GB300-based GPU systems - Next-generation AI training and inference clusters
Mark Zuckerberg himself framed the ambition: "We're excited to expand our partnership with NVIDIA to build leading-edge clusters using their Vera Rubin platform to deliver personal superintelligence to everyone."
What This Means for the Hosting Industry
For the broader hosting and infrastructure industry, this partnership signals several trends:
- Power efficiency is the new battleground - With energy costs rising and data centre capacity constrained, performance-per-watt is becoming more important than raw performance
- Arm architecture is going mainstream - Whether through custom chips or NVIDIA's Grace, Arm-based servers are no longer niche alternatives to x86
- AI workloads are driving infrastructure decisions - Companies are building their entire compute strategy around AI requirements
- Scale continues to accelerate - When a single company's annual capex exceeds $100 billion, the ripple effects on supply chains, power grids, and the hosting market are enormous
Whether you're running a small email hosting operation or managing enterprise infrastructure, the underlying message is clear: the data centre of 2027 will look very different from the one of 2024, and the companies investing now will define that future.
