PicoJool, which is developing AI data center interconnects based on vertical cavity surface emitting lasers, raised a $27.5M Series A led by Socratic Partners
A startup called PicoJool that's backed by former Intel Corp. Chief Executive Pat Gelsinger says it's ready to help ease the strain …
Context & Ripple Effects
PicoJool enters a funded field already populated by Celestial AI, which raised capital for photonic links between processors and memory, and Celestial AI's 2024 optical-connectivity financing framed those links as an AI-workload scaling problem.
The competitive set broadened again in 2025 when OpenLight funded custom silicon-photonic chips for data-center interconnects. Pat Gelsinger argued that processor-to-processor data movement is becoming a performance, power and cost constraint as AI systems scale, giving the funding round a clear infrastructure rationale.
First-order effects
- PicoJool gains $27.5 million of Series A capital, led by Socratic Partners, to advance its VCSEL-based optical connectivity for AI data centers.
- Socratic Partners becomes the lead financial backer of a company targeting the data-movement bottleneck identified by Gelsinger.
Second-order effects
- Celestial AI and OpenLight face another funded contender for customer validation and capital in AI data-center optical interconnects.
- AI infrastructure buyers gain an additional prospective approach to reducing the performance, power and cost burden of moving data among processors.
Third-order effects
- If funding continues to flow to optical links and related data-center components, AI infrastructure investment will extend beyond compute chips into the specialized systems that determine how efficiently compute is connected.
- The parallel funding of optical connectivity and Epic Microsystems' power-delivery architecture points to a broader competition around the physical constraints of AI data centers, not processor performance alone.
The trend: AI infrastructure capital is moving deeper into optical interconnects and power systems as data movement becomes a core constraint on scaled AI computing.