Participated · Series A · May 2026

GridCARE develops the Energize platform, which uses physics-based AI models to evaluate grid conditions—congestion, outages, weather, and demand fluctuations—in real time to locate underutilized power capacity. The company positions this approach as a way to shorten interconnection timelines for large data centers and AI facilities from years to months by unlocking existing grid capacity. GridCARE reports active projects across more than a dozen markets representing over 2 gigawatts of prospective AI compute capacity and cites a Portland General Electric collaboration that identified a pathway to unlock more than 400 megawatts with 80 megawatts expected online in 2026. Co-founder and CTO Ram Rajagopal is a Stanford professor on leave, and the company emphasizes working collaboratively with utilities. GridCARE frames the market opportunity around a “Time-to-Energize Crisis,” arguing power availability is a growing bottleneck for AI expansion. Financially, the company recently raised a $64 million Series A in an oversubscribed round that materially increased its valuation versus a round less than a year earlier.

Participated · Series A · Apr 2026

OpenLight develops a heterogeneous III-V silicon photonics process and a foundry-validated process design kit (PDK) that integrates active and passive components—such as integrated lasers, modulators, amplifiers, and detectors—into custom PASICs. Its PDK is validated at Tower Semiconductor and is used by more than 25 companies to design and fabricate production-grade PASICs across datacom, telecom, AI data centers, sensing, and quantum computing. The company holds over 410 patents related to its PDK and manufacturing of heterogeneously integrated III-V photonics. OpenLight is advancing leading-edge components including 400G modulators and indium phosphide on-chip lasers and amplifiers, and plans to develop standards-based reference PICs at 1.6T and 3.2T. The firm aims to expand its PDK library, worldwide sales, and technical support as it scales; it is headquartered in Santa Barbara, California, with offices in Silicon Valley and has raised $84 million in total funding.

Participated · Series A · Feb 2026

Gruve.ai operates an Inference Infrastructure Fabric that pairs modular, high-density rack-scale pods with a low-latency edge network to run production-grade AI inference. The Redwood City company claims to have unlocked more than 500 MW of distributed inference capacity across the United States and is bringing 30 MW online immediately at four domestic sites. Its full-stack offering includes 24×7 AI-powered security operations, network services, and cluster management, targeting neocloud providers, enterprises running real-time agents, and AI-native startups moving from prototype to production. Gruve.ai argues that most current inference workloads sit on systems optimized for training, creating poor unit economics that its fabric can remedy. The firm intends to bypass multi-year data-center build cycles by quickly deploying capacity in power-available locations and plans near-term geographic expansion into Japan and Western Europe. Since inception it has raised a total of $87.5 million, including the recently announced $50 million follow-on Series A, to scale deployments, broaden strategic partnerships, and grow its agentic services layer.

Participated · Seed · Nov 2025

Hammerhead AI is a San Francisco-based startup developing the orchestrated RL control agents (ORCA) platform, which deploys reinforcement-learning agents to deliver power-aware orchestration for AI factory operators. By transforming power constraints into actionable strategies, the platform boosts token-processing capacity and, in turn, revenue and shareholder value without waiting for grid expansions. Led by CEO Rahul Kar and CTO Rajeev Singh, the company positions itself as a critical layer between infrastructure OEMs and large-scale AI compute facilities. Hammerhead AI plans to use its new capital to accelerate product development, expand global deployments, and deepen collaborations with infrastructure providers and AI factory customers. The $10 million seed round marks the company’s first disclosed financing and provides the resources to scale both engineering and go-to-market operations.

Participated · Series A · Nov 2025

Amperesand is building an integrated Medium Voltage Solid-State Transformer platform that unifies hardware, software, and intelligent controls into a compact product able to cut electrical footprints by more than 80% while delivering faster, more reliable power to high-density compute environments. The company’s technology targets hyperscale AI data centers, defense installations, and megawatt charging applications, where traditional power infrastructure is facing long lead times and mounting grid-stability challenges. Backed by eight years of SST R&D at Singapore’s Nanyang Technological University, Amperesand was spun out in 2023 and now operates engineering and manufacturing hubs in San Francisco, Reno, and Singapore. Commercial momentum includes contracts to deliver 30 MW of systems in 2026, beginning with a mission-critical charging pilot for PSA International at the Port of Singapore and multiple additional hyperscaler pilots. The MV SST platform is designed for secure, regionally sourced supply chains and incorporates AI-native software controls and advanced cybersecurity features. The company plans to exceed 100 employees globally by the end of 2025 to support scaled production. Fresh capital is earmarked for expanding engineering capacity and manufacturing lines to meet surging global demand for critical power solutions.