Ayar Labs
Ayar Labs enables high-bandwidth chip interconnects using optical I/O technology.
Ayar Labs develops TeraPHY, an industry-first optical I/O chiplet that replaces traditional electrical interconnects in semiconductor systems using silicon photonics. The solution delivers up to 1000x bandwidth density improvements and 1/10th the power consumption compared to conventional electrical I/O, enabling 8 Tbps of bandwidth at 10-nanosecond latencies. Built for AI accelerators, data centers, and distributed computing systems, TeraPHY eliminates GPU communication bottlenecks and solves the memory wall problem in high-performance computing.
Problem solved
Electrical I/O creates communication bottlenecks in distributed AI systems, limiting bandwidth density, increasing power consumption, and introducing latency that constrains GPU-to-GPU and intra-GPU data transfer.
Target customer
Semiconductor manufacturers, hyperscale data center operators, system integrators, and AI accelerator/GPU makers building next-generation compute infrastructure
Founders
M
Mark Wade
CEO & Co-Founder
Pioneer in photonics who led the team designing optics for the world's first light-based processor; MIT and UC Berkeley research from 2010-2015.
C
Chen Sun
VP Silicon Engineering & Co-Founder
Led electronics and optical device driver design for the first light-communicating processor; holds PhD and MS in EECS from MIT and BS from UC Berkeley.
V
Vladimir Stojanovic
CTO & Co-Founder
Led MIT's Physical Optics and Electronics Group; former ARPA-E Program Director at US Department of Energy (2010-2012) and Director of MIT Center for Integrated Photonics (2004-2012).
M
Milos Popovic
Co-Founder
Led University of Colorado Boulder lab that developed optical devices for first light-communicating processor; author/co-author of 25+ patents and 150+ publications.
Funding history
Grant
Unknown
January 2016
Led by OSTI
· Unknown
Series A
Unknown
November 2018
Led by Intel Education Accelerator
· Unknown
Series B
$35M
November 2020
Led by Downing Ventures, BlueSky Capital
· Applied Ventures, Castor Ventures, SGInnovate, Intel Capital, Lockheed Martin Ventures, Global Foundries, Playground Global
Series C
$130M
April 2022
Led by Unknown
· Hewlett Packard Enterprise, NVIDIA
Series D
$155M
December 2024
Led by Advent Global Opportunities, Light Street Capital
· AMD Ventures, Intel Capital, NVIDIA
Series E
$500M
March 2026
Led by Neuberger Berman
· Unknown
Total raised:
$870M
Industries
Pricing
Not publicly disclosed. Licensing and direct sales model to semiconductor manufacturers, data center operators, and system integrators.
Notable customers
Intel/Altera, select data center customers, undisclosed AI accelerator and switch manufacturers
Integrations
Intel Agilex FPGA, UCIe standard (in-package and off-package interconnect standard compliance)
Tech stack
LazySizes (JavaScript libraries)
jQuery Migrate (JavaScript libraries)
jQuery (JavaScript libraries)
core-js (JavaScript libraries)
RSS
Open Graph
WordPress (Blogs)
Site Kit (Analytics)
HubSpot Analytics (Analytics)
Matomo Analytics (Analytics)
Linkedin Insight Tag (Analytics)
Hotjar (Analytics)
Google Analytics (Analytics)
Facebook Pixel (Analytics)
Google Font API (Font scripts)
Nginx (Reverse proxies)
PHP (Programming languages)
Apple iCloud Mail (Webmail)
Google Workspace (Email)
MySQL (Databases)
AdRoll (Advertising)
Google Tag Manager (Tag managers)
Divi (Page builders)
Yoast SEO (SEO)
SiteGround (Hosting)
AdRoll CMP System (Cookie compliance)
The Events Calendar (WordPress plugins)
HubSpot WordPress plugin (WordPress plugins)
Website
Competitors
Lightmatter
Also uses photonics for AI/computing but focuses on optical processors; raised $400M Series D at $4.4B valuation in October 2024.
Acacia
Optical networking company with different target markets and technology approach.
Spectra7
High-speed interconnect provider with alternative electrical and optical solutions.
Quintessent
Competing optical interconnect technology with different architectural approach.
Celestial AI
Alternative photonics-based chiplet solution for AI accelerators.
Lightelligence
Optical computing company with different market focus.
Why this matters: Ayar Labs is the clear market leader in optical I/O chiplets with $870M in funding and a $3.75B valuation, backed by semiconductor giants Intel, NVIDIA, AMD, and HPE. As AI compute demands explode, the company's TeraPHY technology solves a fundamental physics problem—electrical interconnects are hitting bandwidth and power limits—making optical I/O increasingly essential for next-generation data centers and AI infrastructure.
Best for: Semiconductor manufacturers and data center operators designing next-generation AI accelerators, GPUs, and high-performance compute systems that require extreme bandwidth density and power efficiency.
Use cases
GPU-to-GPU Communication in AI Training Clusters
Ayar Labs' optical I/O eliminates the electrical interconnect bottleneck between GPUs in distributed AI training, enabling 8 Tbps bandwidth with minimal latency. This is critical for training large models across hundreds or thousands of GPUs where GPU-to-GPU communication is often the limiting factor rather than GPU compute capacity.
Data Center Switch and AI Accelerator Integration
Hyperscale data centers integrate TeraPHY optical chiplets into next-generation switches and custom AI accelerators to achieve higher throughput between compute and network layers. The solution enables scaling beyond current electrical I/O limits while reducing power consumption in power-constrained data center environments.
In-Package Multi-Chiplet Systems
System designers use TeraPHY for millimeter-scale intra-package communication between dies (memory, logic, accelerators) in a single package, solving the memory wall problem by enabling much higher bandwidth between CPU/GPU and high-bandwidth memory (HBM) compared to traditional electrical interconnects.
Alternatives
Electrical I/O (Status Quo)
Current standard interconnect; limited by physical constraints to lower bandwidth density and higher power consumption; no latency advantage.
Lightmatter Optical Processors
Focuses on end-to-end optical computing rather than I/O chiplets; different architectural philosophy and target market positioning.
HBM and Advanced Memory Stacking
Addresses memory bandwidth bottleneck through 3D stacking rather than optical interconnects; complementary but different approach with different cost/complexity tradeoffs.
FAQ
What does Ayar Labs do? +
Ayar Labs designs and manufactures TeraPHY, an optical I/O chiplet technology that replaces traditional electrical interconnects in semiconductors. Using silicon photonics integrated into standard CMOS processes, it enables dramatically higher bandwidth density (up to 1000x), lower power consumption (1/10th), and lower latency (10 nanoseconds) for communication between chips in data centers, AI accelerators, and high-performance computing systems.
How much does Ayar Labs cost? +
Pricing is not publicly disclosed. The company uses a licensing and direct sales model tailored to semiconductor manufacturers, data center operators, and system integrators. Customers should contact Ayar Labs directly for pricing information.
What are alternatives to Ayar Labs? +
Alternatives include Lightmatter (optical processors), Celestial AI (optical chiplets), Spectra7 (high-speed interconnects), and traditional HBM/memory stacking solutions. For electrical interconnects, companies may continue using existing solutions like PCIe or proprietary electrical I/O, though these have inherent bandwidth and power limitations.
Who uses Ayar Labs? +
Target customers include semiconductor manufacturers (Intel/Altera confirmed), hyperscale data center operators, system integrators, and AI accelerator makers. The company has begun shipping components to data centers and is engaged with select customers on integrating optical interconnects into next-generation AI accelerators and switches.
How does Ayar Labs compare to Lightmatter? +
Both use photonics but with different approaches: Ayar Labs focuses on optical I/O chiplets for interconnection between existing semiconductor components, while Lightmatter builds end-to-end optical processors. Ayar Labs targets the interconnect layer, while Lightmatter targets optical computing at the processor level. Lightmatter has higher valuation ($4.4B) but Ayar Labs has raised more total capital ($870M vs. $400M Series D).
Tags
optical interconnects
silicon photonics
chiplets
semiconductor
AI infrastructure
GPU communication
high-bandwidth memory
data centers