Semiconductor intellectual property (IP) leader Arm’s CEO Rene Haas expressed heartfelt sentiments at Computex in Taipei on the 2nd, stating that Taiwan’s ecosystem has produced approximately 250 billion Arm-based chips to date. From the very first Arm chip being packaged and tested in Taiwan to today’s most cutting-edge artificial intelligence (AI) and robotics applications, Taiwan has been an indispensable core partner. He stated bluntly: “Without Taiwan, there would be no Arm.”

Haas’s keynote focused on the computational foundation for the era of Agentic AI and unveiled, for the first time, Arm’s multi-generation CPU blueprint specifically designed for Artificial General Intelligence (AGI). He confirmed that the first-generation Arm AGI CPU, manufactured by TSMC (2330.TW), has entered the production phase. The second-generation product is currently under development, while preliminary planning for the third generation has already commenced, with a future trajectory aimed at more cores, higher performance, and greater energy efficiency.

Notably, NVIDIA CEO Jensen Huang made a surprise appearance on stage at Haas’s keynote after concluding his global media Q&A session, showing his support. Upon taking the stage, Huang joked, “The stock prices of companies that partner with me tend to surge,” drawing roaring laughter from the audience and underscoring the increasingly close collaboration between the two companies within the AI computing ecosystem.

A Taiwan Journey: From One Chip to 250 Billion

During his speech, Haas reflected on the three-decade-long partnership between Arm and Taiwan. He noted that Arm was founded in 1990 and established connections with Taiwan shortly thereafter. Around 1993, the first Arm-based chip designs were closely tied to Taiwan’s ecosystem. At that time, the concept of System-on-Chip (SoC) was not yet mature, and design tools and Electronic Design Automation (EDA) support were still in their early stages. However, Taiwanese partners helped Arm test its IP and design methodologies, forming a crucial foundation for Arm’s subsequent global expansion.

Haas specifically mentioned that the very first Arm chip was packaged and tested in Taiwan. Today, through its partners, approximately 250 billion Arm-based chips have been produced in Taiwan, a scale higher than any other region globally.

He further elaborated that whether it is the electronic products he uses almost daily—from major global brand smartphones, Apple MacBook laptops, and Meta’s Ray-Ban smart glasses—or chips inside humanoid robots and Tesla electric vehicles in the realm of physical AI, or high-end processors in cloud AI racks, these products are all realized within Taiwan’s industrial ecosystem.

AGI CPU Multi-Generation Roadmap and Technical Details

On the technical front, Haas disclosed the specific configurations of the Arm AGI CPU. Two system solutions are currently offered: the first is an air-cooled rack solution with a power consumption of approximately 36kW, providing around 8,000 cores; the second is a liquid-cooled rack solution with a power consumption of about 200kW, delivering over 25,000 cores. Compared to equivalent x86 systems, the Arm AGI CPU can provide roughly twice the rack performance under the same power conditions, enabling customers to achieve higher computing capabilities within the same power budget.

Haas emphasized that Arm has evolved from a traditional IP supplier into an end-to-end solution provider. In addition to the Arm AGI CPU currently in mass production, Arm also offers Compute Subsystems (CSS) and standalone IP products, providing flexible collaboration models tailored to different customer needs. Major players including Google, Amazon, NVIDIA, and various hyperscale cloud service providers can adopt different levels of solutions based on their requirements.

All products are built upon the CSS platform strategy, and the compute subsystems provided to partners also serve as the technical foundation for Arm’s own products, showcasing Arm’s strategic transformation from a licensing model toward deeper system integration.

Jensen Huang’s Surprise Appearance and RTX Spark Collaboration

Jensen Huang’s appearance was the biggest highlight of the keynote. Midway through Haas’s speech, Huang unexpectedly took the stage, instantly electrifying the atmosphere. Huang opened by joking that “the stock prices of companies that partner with me tend to surge,” a remark that not only echoed the market reaction from the previous day—where partner MediaTek (2454.TW) saw its stock price spike to a new high and Arm’s stock price also surged to a record peak following NVIDIA’s announcement at GTC Taipei of its return to the PC market with the RTX Spark super chip—but was also jokingly dubbed by netizens as “the strongest live-stream sales pitch.”

Haas displayed an old photo from 2012 with Huang on stage, humorously acknowledging their long-standing friendship. Huang mentioned having worked with Haas in the past and expressed excitement about the opportunity to collaborate again in the AI era. The two jointly promoted the RTX Spark on stage and shared their latest views on future AI development, highlighting the deep cooperation between Arm and NVIDIA in the AI computing ecosystem.

AI-Generated Night Market Short Film and Taiwanese Imagery

Another captivating segment of the keynote was a short film generated by artificial intelligence played by Arm. The video featured Haas strolling through Taiwanese night markets, signing autographs for locals, riding a motorcycle through alleyways, ascending Taipei’s landmark 101 building, and finally parachuting safely into downtown Taipei. This short film, rich with local Taiwanese elements, not only demonstrated the technical prowess of AI-generated content but also conveyed, through a soft approach, Arm’s regard for and emotional connection to the Taiwanese market.

Taiwan as the Core Hub of Global Semiconductors

Haas repeatedly emphasized Taiwan’s critical position in the global semiconductor industry during his speech. He pointed out that whether for edge computing, Physical AI, humanoid robots, or cloud AI, Taiwan is the world’s most important core hub. From AI server CPUs, NVIDIA racks, and AWS Graviton to various cloud AI infrastructures, everything is highly interconnected with Taiwan’s manufacturing, design, and system integration capabilities.

Haas expressed Arm’s deep gratitude to its Taiwanese partners, talent, and industrial ecosystem. Taiwan is not only an important partner for Arm but also a key force underpinning Arm’s ascent to a core position in global computing. Arm will continue to deepen its cooperation with Taiwanese partners to jointly drive the next generation of AI computing development.

This year’s Computex once again affirmed Taiwan’s irreplaceable position in the global technology supply chain, while the Arm CEO’s emotional confession and Jensen Huang’s supportive appearance added both topicality and strategic significance to this annual tech extravaganza.