PicoJool, a startup supported by former Intel CEO Pat Gelsinger, has raised $27.5 million in Series A funding to accelerate deployment of innovative optical laser technology designed to dramatically increase data center bandwidth and energy efficiency for AI workloads.
- Raised $27.5M Series A led by Socratic Partners with backing from Pat Gelsinger’s Playground Global
- Develops VCSEL optical interconnects delivering 100G to 200G speeds per lane, targeting AI data center clusters
- Focus on commercializing technology with partners including WIN Semiconductors for mass production
Market signal
The rapid expansion of AI processing workloads is driving an urgent need for higher bandwidth and more energy-efficient data center interconnect technology. Traditional copper wiring solutions are increasingly inadequate for the massive data exchange between large GPU and accelerator clusters powering advanced AI models. PicoJool’s substantial early-stage fundraising reflects market recognition of optical VCSELs as a promising alternative to meet these escalating demands.
The participation of notable investors such as Socratic Partners and Hudson River Trading, alongside founder Pat Gelsinger’s Playground Global, signals strong confidence in PicoJool’s ability to address a core infrastructural hurdle. The startup’s focus on delivering scalable VCSEL solutions with speeds up to 3.2 terabits per second positions it to capture critical adoption among data center operators pursuing next-generation AI capabilities.
Operator impact
Data center operators running large-scale AI clusters face operational and cost challenges stemming from bandwidth bottlenecks and power inefficiencies of legacy interconnects. PicoJool’s VCSEL-based optical chips and modular solutions offer a pathway to both increase data throughput and reduce energy consumption relative to copper wiring, potentially lowering total cost of ownership while enabling faster AI model training and inference.
The startup’s architecture employs massively parallel MicroVCSEL designs that aggregate bandwidth across multiple lanes, delivering flexible configurations from 50G up to 200G speeds. This modularity will allow data center operators to tailor connectivity solutions specifically optimized for AI cluster size and workload type, supporting scalability as AI infrastructure demands continue to grow.
What to watch next
Commercialization progress will depend on PicoJool’s ability to finalize product qualification with early customers and scale semiconductor fabrication through collaborators such as WIN Semiconductors. Monitoring announcements around customer partnerships and product certifications will be key indicators of technology adoption momentum.
Further developments in VCSEL technology performance, including additional speed grades and energy efficiency improvements, will also influence market traction. Industry uptake may accelerate as end users seek to alleviate growing connectivity constraints and control rising power costs associated with expansive AI clusters.