Key Highlights
- Nvidia shares advanced 1.6% during Wednesday’s premarket session, hitting $177.97
- Arm Holdings unveiled its debut data center CPU called Arm AGI CPU, with projections for $15B yearly revenue by 2031
- The new Arm chip doesn’t directly challenge Nvidia’s GPU lineup — though potential overlap exists with Nvidia’s standalone Vera CPU products
- Amazon Web Services committed to purchasing 1 million Nvidia GPUs for AI inference tasks, a transaction valued above $50 billion
- Nvidia has restarted H200 chip manufacturing and is developing China-compliant Groq 3 variants, representing approximately $32B in yearly revenue
Nvidia shares posted gains in Wednesday’s early session, brushing aside concerns about Arm Holdings’ entrance into the AI chip sector. The advance accompanied two additional significant revenue catalysts that had received minimal attention from investors.
Arm revealed its inaugural data center CPU on Tuesday evening — dubbed the Arm AGI CPU — counting Meta Platforms and OpenAI among its initial client base. During an after-hours investor presentation, Arm outlined bold projections, estimating the CPU division will deliver approximately $15 billion in yearly revenue by 2031, contributing to an overall $25 billion revenue target.
Arm shares surged 12% in premarket activity following the disclosure.
However, industry analysts quickly clarified that this new processor doesn’t directly compete with Nvidia’s primary GPU offerings.
Benchmark Research’s Cody Acree noted that Arm’s strategy is “less about catching up to the accelerator wave and more about inserting itself deeper into the architecture that governs how AI infrastructure actually runs.”
Nvidia’s CEO Jensen Huang even appeared in Arm’s launch documentation, describing their nearly twenty-year collaboration as fundamental to creating “one seamless platform, from cloud to edge to AI factories.”
The competitive dynamics become more intriguing concerning Nvidia’s standalone Vera CPUs, which debuted during last week’s developer conference. J.P. Morgan’s Harlan Sur highlighted potential competitive overlap with Arm’s offering. He further noted Meta has already secured a deal with Nvidia for its Arm-architecture CPUs — adding complexity to the competitive landscape.
Amazon Web Services Commits to Massive Nvidia Purchase
Separately, Amazon Web Services revealed plans to acquire 1 million Nvidia GPUs dedicated to AI inference workloads. The announcement caught many off guard — AWS had recently positioned itself as hosting “the largest cluster of non-Nvidia chips in the world” following its Indiana facility launch in October 2025.
The agreement encompasses a “broad mix” of six additional Nvidia chip variants, including the recently announced Groq 3 inference processors, alongside Nvidia networking equipment. Industry estimates place the complete package value above $50 billion, with completion targeted for late 2027.
This singular agreement accounts for approximately 25% of Nvidia’s complete 2025 yearly revenue.
Chinese Market Revenue Returns
CEO Jensen Huang confirmed last week that Nvidia is resuming H200 chip manufacturing — engineered to satisfy U.S. export control requirements — for Chinese customers. Industry sources additionally suggest a China-compliant Groq 3 chip variant is under development.
Nvidia had projected zero Chinese data center revenue in its Q1 forward guidance. Throughout 2025, those sales averaged approximately $8 billion quarterly — totaling around $32 billion annually, representing roughly 15% of complete 2025 revenue.
Together, the AWS agreement and China market resumption account for more than $82 billion in revenue absent from Nvidia’s current financial projections.
Nvidia shares traded up 1.6% at $177.97 during Wednesday’s premarket session, bouncing back from a 0.3% decline in the previous trading day.
