Day 1 of Computex Taipei 2025 showed how the balance of power in global computing is starting to shift.
NVIDIA opened up to hyperscalers, Qualcomm returned to the data center and Arm strengthened its position in AI, clear signs that these companies are thinking beyond products and focusing on long-term strategies.

NVIDIA opens up to hyperscalers

The most significant announcement from NVIDIA was around NVLink Fusion, where the company is opening up its industry-leading NVLink networking IP to help hyperscalers integrate their semi-custom chips into NVIDIA’s partner ecosystem.

“It addresses a major challenge for hyperscalers looking to scale up and scale out using semi-custom infrastructure alongside NVIDIA’s systems,” observes Neil Shah, VP of Research at Counterpoint. “It also enables NVIDIA to monetize and scale its NVLink IP, positioning it as the industry’s gold standard.”

The company also announced NVIDIA Constellation, a much larger office in Taiwan acting as a second HQ to provide proximity to key partners and access to talent. It reflects NVIDIA’s continued ecosystem building and strengthens collaboration efforts with the HPC and PC supply chain, as well as with growing EMS players like Foxconn in automotive.

Besides, NVIDIA’s partnership with TSMC, Foxconn and Taiwan’s government to build the country’s largest quantum supercomputer provides a scalable template for global deployment. Similar efforts are underway in Japan, where NVIDIA is working with AIST to launch the world’s largest research supercomputer for quantum computing (ABCI-Q), reinforcing its push across the HPC spectrum.

NVIDIA also introduced the RTX Pro 6000 Blackwell Server Edition GPUs to help enterprises build ‘AI factories’ supporting diverse workloads – from multimedia and simulations through to AI agents and digital twins.

This system includes up to eight RTX Pro 6000 GPUs, Bluefield-3, CX-8 Super NICs and NVIDIA’s AI Enterprise Software platform. With a wide ecosystem of partners (like Dell, Lenovo, HPE, Asus, Foxconn and Gigabyte) and software providers (like Ansys, Cadence, CrowdStrike and Red Hat), this end-to-end offering can transform trillion-dollar enterprise IT systems for the AI era. Notable customers already deploying AI factories include Foxconn, Lockheed Martin, Aramco, HUMAIN and Telenor.

Source: Counterpoint Research

Source: Counterpoint Research

Qualcomm’s expansion across PCs, data centers

A slew of announcements came from Qualcomm, first and foremost around the Snapdragon X Series powering more than 85 PCs. Also, the company reinforced its dual approach to AI – on-device and cloud-based, especially across the enterprise. This was further emphasized with agentic AI use case examples for productivity and creativity.
The previous week saw the company announcing its re-entry into the data center CPU market with custom, AI-centric processors connecting to NVIDIA’s chips and aimed at high performance and low power usage.

Finally, the Snapdragon X2 platform was teased. Scheduled for a September launch, the Snapdragon X2 is expected to include 18 (up from 12) third-generation Oryon V3 cores, support for up to 48GB of LPDDR5X RAM and integration of a 1TB NVMe SSD into the chip package.

“Qualcomm continues to demonstrate its commitment to advancing AI integration across the broad computing ecosystem,” says David Naranjo, Associate Director at Counterpoint. “The announcement around its data center strategy highlights its intent to leverage NUVIA CPU IP and increase enterprise visibility.”

Source: Counterpoint Research

Arm powering AI Everywhere

Arm reiterated its critical role in enabling AI, closely collaborating with Taiwan’s industrial ecosystem across SoC design, wafer manufacturing, packaging, and server/motherboard vendors.
During its keynote, Arm unveiled its next-generation CPU, Travis, which includes Scalable Matrix Extension (SME) to deliver double-digit IPC gains and support for AI acceleration at the CPU level. The company also highlighted next-generation GPU Drage, which offers advanced graphics capabilities to meet growing demand in immersive gaming.

“This reflects Arm’s continued investment in AI software ecosystems, making it easier and faster for developers to create AI-powered applications,” says William Li, Senior Analyst at Counterpoint.

In the data center segment, Arm is gaining momentum through its deep partnerships with hyperscalers like AWS, Google and Microsoft. The company noted that 50% of new server chip solutions for hyperscale data centers are Arm-based, reinforcing its leadership in addressing the rising demand for inferencing and scalable AI compute.
In edge and consumer devices, where Arm has long dominated, compute architecture is driving the evolution of AI PCs and tablets. With Apple’s M series and Qualcomm’s Snapdragon X series pushing Windows on Arm, Arm forecasts that over 40% of PC and tablet shipments will be Arm-based in 2025, a projection that aligns with Counterpoint’s Arm shipment numbers forecasting a significant ramp-up in Arm-based laptops next year.

Source: Arm