Led · Series F · Jun 2026

Groq has repositioned itself from a semiconductor maker into a provider of AI inference compute infrastructure following the sale of most of its chip assets to Nvidia in late 2025. After that transaction Groq's previous leadership and a majority of its engineering team moved to Nvidia. The company currently operates a network of 13 data centers and plans to expand total capacity to roughly 200 megawatts by the end of 2027. Groq raised $650 million in the most recent financing led by Disruptive and Infinitum to fund that expansion. In September 2025 it previously raised $750 million at a $6.9 billion valuation with investors that included Samsung, Cisco and BlackRock. Groq is positioning itself to address growing demand for reliable, cost-efficient inference services for deployed AI models.

Participated · Series B · Sep 2023

Enfabrica builds high-performance networking silicon and software for AI and accelerated computing, centered on its Accelerated Compute Fabric (ACF) SuperNIC. The company’s ACF SuperNIC delivers multi-port 800‑Gigabit‑Ethernet connectivity to GPU servers and claims four times the bandwidth and multipath resiliency of other GPU-attached NIC products. Enfabrica announced a 3.2 Terabit/sec ACF SuperNIC chip and said initial quantities and pilot systems will be available in Q1 2025 and are commercially orderable. To support that rollout, Enfabrica raised $115 million in an oversubscribed Series C equity financing led by Spark Capital. The company plans to use the funds to drive the volume production ramp of the ACF SuperNIC, scale its global R&D team, and expand next-generation product development. Investors in the round include strategic and corporate backers such as Arm, Cisco Investments, Maverick Silicon, Samsung Catalyst Fund and VentureTech Alliance alongside existing venture investors. Enfabrica develops networking silicon and systems aimed at addressing I/O, memory movement and networking bottlenecks for parallel AI workloads. Its flagship product, the Accelerated Compute Fabric Switch (ACF-S), delivers multi-terabit-per-second data movement between GPUs, CPUs and accelerators and uses standards-based interfaces to scale to tens of thousands of nodes. Enfabrica says the ACF-S can reduce GPU compute for large language models (along the lines of Llama 2) by around 50% at the same performance point and eliminates the need for disparate server I/O and networking chips. The company plans to deploy the new funding to support R&D, production, operations and to expand engineering, sales, marketing and go-to-market efforts. Enfabrica began its journey in 2020, emerged from stealth in 2023, and is based in Mountain View with just over 100 employees across North America, Europe and India. Leadership includes co-founder and CEO Rochan Sankar and chief development officer Shrijeet Mukherjee, and the product is positioned as agnostic to processor vendor to avoid proprietary lock-in.