California’s PicoJool Raises $39.5 Million for 200G AI Data Center VCSELs
PicoJool raises $39.5M to advance 200G VCSEL technology for hyperscale AI data centers, targeting 1.6T connectivity with energy-efficient optical interconn
Research Desk
PicoJool, an optical connectivity startup headquartered in Palo Alto, has closed a USD 27.5 million Series A round that brings its cumulative funding to USD 39.5 million, following a USD 12 million seed round previously led by Playground Global. Socratic Partners led the new round, with Hudson River Trading joining as a participant.
Where the Money Is Going
The company plans to direct the Series A capital toward expanding its workforce and physical operations across both the United States and Taiwan. Growth is targeted across research and development, operations, sales, and marketing functions. The immediate commercial priority is qualifying, manufacturing, and supporting products for hyperscalers and data center operators. Chip-level VCSEL products have already entered early sampling ahead of wider availability.
Chief executive and founder Al Yuen described the company's current position as one of operational execution following a technical milestone. "Reaching 200G per lane proved that the performance is there. Our focus now is execution," Yuen said. "This funding allows us to expand our teams in the U.S. and Taiwan, qualify products with customers and move a complete portfolio from chip-level VCSELs and microVCSELs into AOC and NPO modules."
What PicoJool Has Built
The company's core technology is based on Vertical Cavity Surface Emitting Lasers, known as VCSELs, which it is developing for optical connectivity products designed to serve large-scale AI systems. Its product portfolio spans high-speed VCSEL configurations at 100G and 200G per lane, alongside lower-power microVCSEL offerings at 50G NRZ and 64G NRZ/PAM4. The 200G VCSEL products deliver more than 37 gigahertz of bandwidth. The full lineup includes quad 100G, quad 200G, and 32x50G NRZ microVCSEL configurations.
PicoJool has said these options are intended to give customers multiple routes to 1.6 terabit connectivity, with the overall portfolio designed to scale from 800G through 1.6T, 3.2T, and beyond depending on bandwidth, power, and system needs.
The microVCSEL platform relies on what the company calls a massively parallel architecture, which is designed to increase aggregate bandwidth while reducing the energy consumed per bit of data transmitted. PicoJool has reported growing customer interest in specific microVCSEL configurations, including a 50G NRZ product aimed at optical compute interconnects and 64G NRZ/PAM4 variants targeted at PCIe 6.0-based near-package optics applications.
The Problem Driving Demand
The scaling of AI clusters has placed intensifying pressure on both interconnect performance and data center power consumption. As AI systems incorporate more processors and the volume of data moving between them rises, the need for faster and more energy-efficient optical connectivity has grown sharply. PicoJool is positioning its VCSEL-based product line as a solution to the bandwidth and power constraints that hyperscale AI deployments are generating.
PicoJool unveiled its 200G VCSEL and microVCSEL product family for scale-up AI data centers in June 2026, ahead of the Series A announcement. The company is currently engaged with system companies and hyperscalers to define next-generation solutions spanning pluggable optics, near-package optics, and co-packaged optics.
Production Infrastructure
To ready its 200G VCSELs for high-volume manufacturing, PicoJool is partnering with WIN Semiconductor alongside other gallium arsenide foundries. WIN Semiconductor brings substantial production experience to the arrangement, having shipped more than one billion chips over the past decade. The partnership is intended to support PicoJool as it moves its products through the qualification and production process.
How Investors See the Opportunity
Greg Waters, General Partner at Socratic Partners, tied the investment thesis directly to the economics of moving data at AI scale. "Moving data efficiently dictates how quickly and economically AI scales," Waters said. "PicoJool combines a proven team with a unique approach that brings VCSEL technology's cost and manufacturing advantages to AI-scale bandwidth and power needs. They are poised to become a vital connectivity platform for next-generation data centers."
Pat Gelsinger, a general partner at Playground Global, which led the seed round and also participated in the Series A, emphasized the combination of technical performance and manufacturing capability. "The network is the AI, and connectivity is becoming one of the defining constraints on performance," Gelsinger said. "PicoJool brings together breakthrough VCSEL performance, deep systems expertise and a manufacturing base capable of supporting tremendous scale."
Company Foundations
PicoJool was founded by Al Yuen, whom the company identifies as a photonics industry veteran. The company operates from its Palo Alto headquarters and maintains research, development, and operational activities across both the United States and Taiwan.
Its stated mission is to develop optical chips and modules that deliver high-bandwidth, low-cost connectivity for hyperscale AI data centers, with a scope that extends from semiconductor design through transceiver and system-level integration.
Tagged