This site is fictional demo content. It is not real news or affiliated with any real organization. Do not treat it as fact or professional advice.

NextPaper Logo
nextpaper.ccCYBER CHRONICLE
RSS
EN中文

Full article

FULL TEXT

View this issue
HeadlineTECH

Photonic Computing Chip LuxCore Goes Commercial: Compute Density Breaks 1000 TOPS, 50x More Energy-Efficient Than Silicon

Canadian startup Luminous Computing released its first commercial photonic computing chip LuxCore, achieving 1000 TOPS on specific AI inference tasks at just 20 watts — 50x more energy-efficient than equivalent silicon GPUs.

On April 2, 2028, Toronto-based startup Luminous Computing officially released its first commercial photonic computing chip, the LuxCore LP-1000. This chip uses optical signals instead of electrical signals for matrix operations, achieving a peak compute of 1000 TOPS on Transformer model inference benchmarks while consuming only 20 watts — approximately 50 times more energy-efficient than comparable silicon GPUs.

LuxCore's core innovation lies in its "on-chip optical interconnect" architecture. In traditional chips, data movement between compute units and memory consumes enormous energy and time — known as the "memory wall." LuxCore integrates waveguide arrays on a silicon photonic substrate, allowing optical signals to propagate at near-light speed within the chip, fundamentally eliminating the data movement bottleneck. The chip integrates over 12,000 micro-optical modulators, each measuring just 3 micrometers.

First customers include Google DeepMind and Amazon AWS. Google infrastructure VP Parthasarathy Ranganathan stated that photonic computing offers significant cost advantages in large-scale inference scenarios. Calculating the billions of AI inference requests Google Search handles daily, full adoption of photonic chips could reduce inference energy consumption by over 90%.

However, LuxCore is not a universal solution. Luminous Computing CEO Marcus Chen acknowledged that photonic chips don't yet match traditional GPUs in training scenarios, as gradient backpropagation during training requires high-precision floating-point arithmetic that current optical components cannot deliver. LuxCore is positioned for inference acceleration, particularly large-scale online inference services.

Pricing is set at $4,800 per LuxCore LP-1000 chip, with development boards at $12,000. Luminous Computing has completed a $280 million Series C funding round at a $1.8 billion valuation, planning to release its next-generation product in 2029 targeting 10x energy efficiency improvement for training workloads.