Uploaded July 2026 | Updated September 2026, 2 weeks ago
In this interview from theCUBE + NYSE Wired: Robotics & AI Infra Leaders 2026, Darren Chien, managing director of APAC at Positron AI, joins theCUBE's John Furrier to discuss why inference has become the defining economic battleground of the AI era. Chien breaks down the mechanics of prefill and decode, explaining how compute-bound and memory-bound workloads require different optimization strategies. He details how Positron AI is built to drive down the cost of tokens, competing on tokens per dollar and tokens per watt rather than raw throughput alone. Chien also shares that the company's first-generation Atlas product is on pace to close the year near $100 million in revenue, with the upcoming Asimov chip and Titan system designed to maximize memory capacity and bandwidth utilization for next-generation models.
The conversation also explores how Positron AI's air-cooled Titan systems let enterprises add inference capacity into existing data centers without costly retrofits, lowering the barrier for legacy IT teams to compete with hyperscale buildouts. Chien traces the company's origins to a team with deep semiconductor and cloud backgrounds, formed around 2023 to design chips for today's workloads rather than yesterday's. He details Positron AI's expanding footprint across Taiwan, Korea, Singapore and Japan, pointing to nine to 10 gigawatts of underutilized air-cooled space across APAC ready for monetization. Chien also weighs in on how the rise of open models and competitive pressure between NVIDIA and AMD is accelerating accessibility and innovation across the industry. From chip design to go-to-market partnerships with players like Oracle, he lays out how Positron AI is positioning itself at the center of the inference economy.
Find more SiliconANGLE news and analysis siliconangle.com
Follow theCUBE's wall-to-wall event coverage siliconangle.com/events
Learn about the latest theCUBE events thecube.net
00:00 - Advancements in AI: A Deep Dive into Innovation and Market Dynamics
02:45 - Advancing AI: Innovations in Hardware and Market Strategies
04:48 - Prefill, Decode, and Context: Enhancing AI Inference
07:37 - Positron AI: Origins, Strategies, and Market Integration
10:47 - Navigating Innovation and Monetization in the Evolving Tech Landscape
13:50 - Title: "Open Models and Their Impact on AI Competition in the APAC Region
#theCUBE #RoboticsAIInfraLeaders #theCUBEresearch #PositronAI #AIInference #EdgeComputing
In this interview from theCUBE + NYSE Wired: Robotics & AI Infra Leaders 2026, Darren Chien, managing director of APAC at Positron AI, joins theCUBE's John Furrier to discuss why inference has become the defining economic battleground of the AI era. Chien breaks down the mechanics of prefill and decode, explaining how compute-bound and memory-bound workloads require different optimization strategies. He details how Positron AI is built to drive down the cost of tokens, competing on tokens per dollar and tokens per watt rather than raw throughput alone. Chien also shares that the company's first-generation Atlas product is on pace to close the year near $100 million in revenue, with the upcoming Asimov chip and Titan system designed to maximize memory capacity and bandwidth utilization for next-generation models.
The conversation also explores how Positron AI's air-cooled Titan systems let enterprises add inference capacity into existing data centers without costly retrofits, lowering the barrier for legacy IT teams to compete with hyperscale buildouts. Chien traces the company's origins to a team with deep semiconductor and cloud backgrounds, formed around 2023 to design chips for today's workloads rather than yesterday's. He details Positron AI's expanding footprint across Taiwan, Korea, Singapore and Japan, pointing to nine to 10 gigawatts of underutilized air-cooled space across APAC ready for monetization. Chien also weighs in on how the rise of open models and competitive pressure between NVIDIA and AMD is accelerating accessibility and innovation across the industry. From chip design to go-to-market partnerships with players like Oracle, he lays out how Positron AI is positioning itself at the center of the inference economy.
Find more SiliconANGLE news and analysis siliconangle.com
Follow theCUBE's wall-to-wall event coverage siliconangle.com/events
Learn about the latest theCUBE events thecube.net
00:00 - Advancements in AI: A Deep Dive into Innovation and Market Dynamics
02:45 - Advancing AI: Innovations in Hardware and Market Strategies
04:48 - Prefill, Decode, and Context: Enhancing AI Inference
07:37 - Positron AI: Origins, Strategies, and Market Integration
10:47 - Navigating Innovation and Monetization in the Evolving Tech Landscape
13:50 - Title: "Open Models and Their Impact on AI Competition in the APAC Region
#theCUBE #RoboticsAIInfraLeaders #theCUBEresearch #PositronAI #AIInference #EdgeComputing
