• Nvidia breaks tradition with NVLink Fusion, allows competitors’ chips to integrate with its AI infrastructure.
• Jensen Huang unveils GB300 systems for Q3 release, announces “Constellation” Taiwan headquarters.

Nvidia shattered its walled garden approach to AI infrastructure at Computex 2025, with CEO Jensen Huang unveiling NVLink Fusion – technology that lets customers integrate non-Nvidia chips into the company’s coveted AI systems.

The surprising pivot comes as Microsoft and Amazon ramp up in-house chip design efforts that threaten Nvidia’s data centre dominance. Speaking on Monday from Taiwan’s premier electronics forum, Huang abandoned his usual consumer GPU focus to instead reveal how Nvidia plans to remain indispensable even as competitors encroach on its AI territory.

Breaking down the walls

Nvidia’s NVLink Fusion marks a departure from the company’s historically closed approach. For the first time, Nvidia will let customers use processors and accelerators from other manufacturers alongside its own AI chips. The pragmatic concession acknowledges the looming reality that tech giants are building their specialised silicon.

“When new markets have to be created, they have to be created starting here, at the centre of the computer ecosystem,” Huang said, emphasising Taiwan’s pivotal role while tactfully avoiding mention of the defensive business strategy driving this open architecture approach.

The technology gives data centre customers the flexibility to integrate components from different vendors while ensuring Nvidia’s interconnect technology remains the crucial backbone. It’s less an act of generosity than a calculated move to prevent companies like Microsoft and Amazon from cutting Nvidia out of their AI infrastructure entirely.

Strategic alliances with potential rivals

Huang has wasted no time in lining up partners for the new interoperable ecosystem. MediaTek, Marvell, and Alchip will develop custom AI chips compatible with Nvidia processor-based systems, while Qualcomm and Fujitsu will create processors designed to work with Nvidia accelerators.

The partnerships validate the new approach and ensure Nvidia maintains its central role. According to Bloomberg, which covered the Computex keynote, the opening-up of Nvidia’s designs “gives Nvidia’s data centre customers more flexibility and allows a measure of competition while still keeping Nvidia technology at the centre.”

Industry analysts see this as Huang’s preemptive strike against commoditisation. By opening limited aspects of its architecture while keeping control of the critical interconnect technology, Nvidia creates the illusion of openness while cementing its position as the essential foundation of any serious AI deployment.

Next-generation hardware and AI factories

The company confirmed its next-generation GB300 systems are coming in the third quarter of 2025, upgrading the current Grace Blackwell systems being installed at present by cloud service providers. Huang also introduced a new RTX Pro Server system, claiming it offers four times better performance than Nvidia’s former flagship H100 AI system with DeepSeek workloads.

For smaller-scale implementations, Nvidia announced that its DGX Spark and DGX Station systems, revealed earlier this year, will be offered by a broader range of suppliers, including Acer, Gigabyte, Dell, and HP, starting this summer.

Huang repeatedly used the term “AI factory” rather than “datacentre” to describe these massive computing facilities. According to the presentation at TechPowerUp’s live blog, systems pull “100s of megawatts of power to offer compute power to hundreds of datacentres.”

Software infrastructure and robotics push

Beyond hardware, Nvidia unveiled DGX Cloud Lepton, a service designed to help cloud computing providers like CoreWeave and SoftBank automate the process of connecting AI developers with computing resources.

The move positions Nvidia to create what appears to be a global marketplace for AI computing. In a significant push toward the robotics frontier, Huang announced the Isaac Groot humanoid robot foundation model along with the Newton physics engine, designed to bridge physical robots with digital simulations.

“Everything that moves will be robotic,” Huang declared, positioning Nvidia to capitalise on what he framed as the next multi-trillion-dollar industry.

Taiwan’s commitment amid geopolitical tensions

Cementing Nvidia’s ties to Taiwan’s semiconductor ecosystem, Huang announced “Constellation”, a new company headquarters in Taipei’s northern suburbs. The timing comes just days after Huang joined President Donald Trump’s Middle East trade mission, where Nvidia secured lucrative deals with Saudi and UAE tech firms.

The Taiwan headquarters announcement serves as a careful balancing act, reassuring TSMC and other Taiwanese partners of Nvidia’s long-term commitment even as the company diversifies its global relationships amid escalating US-China trade tensions and increasingly complex semiconductor export restrictions.

Impact on enterprise IT

Huang predicted a shift in how enterprises approach computing, suggesting that “AI changed compute, storage, and networking for enterprises. Agentic AI is the future for businesses.”

He went further, suggesting “IT departments will become ‘HR for digital workers,'” a vision where AI agents become fundamental components of enterprise operations.

What’s next for Nvidia

As Nvidia’s share rally continues following its Middle Eastern deal-making trip, the company appears determined to shore up its central position in the AI boom. By embracing a more open ecosystem approach while still maintaining architectural control through technologies like NVLink Fusion, Nvidia demonstrates its strategic future.

The focus on Taiwan – from establishing a new headquarters to partnering with local manufacturers – underscores the importance of the region’s supply chain to Nvidia’s future. As Jensen put it when thanking the scores of suppliers from TSMC to Foxconn, these partnerships are essential to “build and distribute Nvidia’s tech around the world.”

For enterprises and developers watching Computex 2025, Nvidia’s announcements signal both continuity and evolution – maintaining its core focus on high-performance computing while adapting its business model to accommodate the changing dynamics of the AI industry.