Playground Global Returns to Taiwan: AI Hardware and Innovation

On his return to Taipei on October 27, 2026 with Playground Global general partner Peter Barrett, former Intel CEO Pat Gelsinger made a pointed statement by attending the second annual Taiwan Deep Tech Summit. This highlights that future bottlenecks in AI will be in the physical domain, and Taiwan will be the place to overcome these bottlenecks.

Playground Global, with $1.7B under management, was founded on investing in the fundamentals of next-generation computing, transformational industry and engineered biology. While the company has invested in many companies including Agility Robotics and PsiQuantum, they are increasingly making their mark in Taiwan by frequently visiting. Gelsinger commented: ‘Taiwan is where the most ambitious technologies in the world get turned into real products at scale.’ Barrett added: ‘AI, quantum and cryogenics are changing the stack completely. Taiwan is where we turn compute into hardware.’

At the summit many of the companies within their portfolio showcased the advancements they’ve made. Nagare Labs, who were founded by researchers from Georgia Tech, Suman Datta and Shimeng Yu is a new venture launching from Taiwan. The startup has developed a new memory architecture that has high density, fast speeds and is compatible with standard silicon processes. They’ve chosen to launch in Taiwan because their technology can be manufactured using the processes already in place.

Syenta demonstrated their Localized Electrochemical Manufacturing process for the first time on a full wafer. They have developed a way to create copper interconnects in minutes without using any lithography. This will allow for simpler and faster advanced packaging solutions. The platform will also scale to large panels from wafers. This will utilise the packaging capabilities in Taiwan.

Key highlights include:

• Nagare Laboratories: Advanced AI memory technology.
• Power lattice: Cutting-edge power management chiplets for AI processors.
• Ayar Laboratories: Optical connectivity for quicker and more effective AI data centers.
• Syenta: Novel semiconductor packaging technology.
• Snowcap: Energy-saving superconducting computing.
• PicoJool: Quick optical connectivity for AI infrastructure.

Optics was a key theme. Ayar Labs, a Playground unicorn valued at $3.75 billion, announced the opening of a Hsinchu office and confirmed it had raised $650 million this year, which included investment from Wiwynn and MediaTek in Taiwan. They’re developing co-packaged optical solutions to replace electrical connections in order to alleviate bandwidth and power issues that large AI clusters face. PicoJool, who are focussed on developing microVCSELs for short reach data centre interconnects, raised $27.5 million Series A led by Socratic Partners, bringing its total funding to $39.5 million. The company is scaling up their operations in the US and Taiwan.

Powerlattice demonstrated what they claim is the first monolithic integrated voltage regulator chiplet in the industry. Their technology brings the final stage of voltage regulation closer to the processor. This means they can reduce the total amount of power needed for compute by over 50%. While this is ambitious, they’ll need to prove these claims against real workloads.

Recommended for You:

• How AI Is Redefining Human-Machine Interaction
• The Ultimate Guide to Using Artificial Intelligence for Lead Generation
• NVIDIA Plans $3 Billion Investment to Expand Meta Partnership on Muse AI Models

However, they are tackling the issue of ‘the power wall’ within AI. Snowcap are developing superconducting circuits that have no resistance. They have demonstrated working devices that have been manufactured on 300mm wafers using foundry CMOS processes. Their goal is to reduce the power requirements for large AI from gigawatts to megawatts.

Overall these companies are highlighting that within the next decade, AI performance will be limited by how much we can move data, supply power and package. Due to Taiwan’s strength in advanced packaging, precision manufacturing and high volume semiconductor manufacturing, they will naturally be the perfect complement to companies that have solved these problems within the lab. This has been highlighted by the Hsinchu office, strategic investment from Taiwan in Ayar Labs and the launch of Nagare Labs.

For Taiwanese manufacturers and system builders, if these technologies are successfully validated then there will be increased demand for components and packaging. Playground have had long-standing relationships with Foxconn and this has been highlighted by the co-hosted dinner.

Gelsinger has highlighted that during his first visit to Taiwan over 4 decades ago it kickstarted the motherboard and PC compatible industry, which has grown into servers. This is his second visit with Playground, having visited in November 2025. Due to the power and interconnect requirements of AI infrastructure, the company have made it an annual event to highlight that Taiwan is now a vital part of the deep tech portfolio.

None of the technologies will necessarily hit high volume manufacturing. The claims of reducing power by 50% and reducing gigawatts to megawatts will need to be qualified by customers. However, it’s clear that as AI systems increase in size and power requirements, whoever can combine innovative architectures with the highest concentration of semiconductor manufacturing in the world will benefit. Playground Global returning to Taiwan shows they recognise this.