Arm Took Both Server Sockets at Once — The Week Qualcomm Dragonfly and NVIDIA Vera Nailed Meta and Dell as Launch Customers in 72 Hours
x86's two-horse race is over — two Arm-based server CPUs landed launch customers in the same 72 hours, and only one company collects royalties on both
1. 72 Hours — Arm Took Both Server Sockets at Once
Between 2026-06-24 and 06-26, two Arm-based server CPUs locked in their first hyperscaler sockets in the same week. On June 24, Qualcomm formally launched its first data center CPU — the Dragonfly C1000 — with Meta as the launch customer. In the same window, NVIDIA's Vera CPU was unveiled as the base CPU inside Dell's PowerEdge XE8812 — a next-generation HPC/AI server delivering 144 GPUs per rack. The market read these as two separate trade catalysts ("good for QCOM," "good for NVDA"). But the fact that two Arm server CPUs nailed their anchor customers in the same 72 hours has to be read one level of abstraction up — at Arm Holdings itself. That is exactly what TD Cowen flagged on the same day, raising its ARM price target on the "AI CPU opportunity."
2. Qualcomm's Full-Stack Data Center Pivot
Qualcomm's announcement is not just another CPU. The Dragonfly platform bundles the C1000 CPU with the AI300 accelerator, networking silicon, and a modular-architecture acquisition — a full stack, not a single chip. Qualcomm guided to "billions in additional revenue" from this lineup and previewed a separate dedicated AI CPU for 2028. A China-specific, export-control-compliant AI accelerator lineup is reportedly in parallel development. Eighteen months ago Qualcomm was a modem + handset + automotive SoC company. Today it is repositioned as a data center vendor selling server CPU + AI accelerator + networking silicon as a unit — and Meta just issued the first certificate.
3. NVIDIA Vera CPU — Five Fronts in One Week
NVIDIA's Vera CPU pushed in five directions in the same week. (1) Dell PowerEdge XE8812 — 144 GPUs per rack, shipping 2027. (2) Super Micro design-in confirmed — both top-2 AI server OEMs are now locked into the Vera Rubin cycle. (3) Los Alamos National Laboratory deployment for scientific AI — a government reference site. (4) China hyperscaler push despite export controls — into direct competition with Hygon's C86. (5) Jensen Huang's annual shareholder meeting positioned Blackwell as the "king of inference" and Vera Rubin as the next-gen unified platform. NVIDIA's market cap crossed $5T immediately after the Vera Rubin reveal cycle. A single CPU architecture aligning OEM, government, overseas market, and CEO messaging in one week is unprecedented in NVIDIA's history.
4. x86 Is Alive — But the Duopoly Is Broken
In the same week, AMD hit an all-time high of $345. Goldman Sachs and Insider Monkey cited an x86 CPU re-acceleration (simultaneous server and client share gains) as the bull case. Korean media framed Intel's "CPU king" comeback as a centerpiece of the Trump administration's industrial policy. China's Hygon unveiled its next-gen C86 CPU, adding weight to Beijing's push for x86-compatible domestic server processors. x86 is alive — prices are strong. But what the 24–26 announcement cluster confirms is that five architectures (Intel Xeon, AMD EPYC, Hygon C86 — x86 / Qualcomm Dragonfly, NVIDIA Vera — Arm) now compete for the same hyperscaler socket in the same quarter. The 30-year x86 duopoly is now a five-way race, and the two newly inserted seats are both Arm.
If this analysis was helpful · ☕ Support Us · ✈️ Telegram