Participated · Series C · Nov 2025

Founded in 2019, Santa Clara-based d-Matrix develops a full-stack inference platform that integrates Corsair™ in-memory inference accelerators, JetStream™ networking NICs, and the Aviator™ software stack. The architecture delivers up to 10× faster performance, 3× lower cost, and 3–5× better energy efficiency than contemporary GPU systems, enabling customers to serve models as large as 100 B parameters in a single rack and to generate roughly 30 K tokens per second on Llama-70B at 2 ms latency per token. By allowing one data center to handle the workload of ten, the platform directly tackles the rising energy and TCO challenges of AI deployment. d-Matrix is currently scaling global deployments for hyperscale, enterprise, and sovereign customers and has announced an open standards reference architecture, SquadRack™, with partners Arista, Broadcom, and Supermicro. A forthcoming roadmap includes 3-D memory stacking innovations aimed at further boosting throughput and energy efficiency. The company employs more than 250 people across offices in the U.S., Canada, Australia, India, and Serbia and has raised a cumulative $450 million in venture funding to date.

Participated · Equity · Oct 2024

DataPelago unveiled a Universal Data Processing Engine designed to accelerate any engine, including open-source ones, on any hardware and for any data type to support GenAI and analytics workloads. The engine is offered as an end-to-end solution or can integrate with Substrait-based open source frameworks to turbocharge Spark and Trino, delivering disruptive price/performance without changing applications or workflows. It integrates with existing data stores and lakehouse platforms to avoid data migration and vendor lock-in. Architecturally the engine comprises three layers—DataVM (a domain-specific ISA for data operators), DataOS (maps data operations to heterogeneous accelerated computing elements), and DataApp (a pluggable integration layer)—which together claim one to two orders of magnitude performance gains versus today’s query engines. Targeted use cases include analyzing billions of transactions while maintaining data freshness, AI-driven threat detection at wire-line speeds, and scalable deployment of training, fine-tuning, and RAG inference pipelines. Co-founder and CEO Rajan Goyal, who has 20+ years building accelerated computing solutions, has assembled a multidisciplinary team to develop the technology and bring it to market; the company launched from stealth with $47 million in funding.

Participated · Series B · May 2022

Point2 Technology develops RF signaling-based interconnect platforms—branded Active RF Cable and e-Tube—that use RF over plastic waveguides to connect accelerators and scale rack-level compute. Its technology aims to extend reach compared with copper, lower power and cost versus optical approaches, and deliver ultra-low latency without laser-based systems. The platform supports multiple form factors including pluggable cables, near-package modules, and co-packaged solutions for accelerator-to-accelerator communication. Point2 is moving from development toward commercialization and deployment and is expanding engineering and systems teams to support that transition. The company has garnered industry recognition such as a BloombergNEF Pioneers award and recently raised $76 million in a Series B extension to fund commercialization and continued product development.

Participated · Series C · Apr 2022

Headquartered in San Jose, CA, Ayar Labs builds co-packaged optics that move data between chips with light instead of copper, reducing power consumption while boosting bandwidth for AI infrastructure. Its flagship TeraPHY optical engine is manufactured in concert with leading semiconductor partners and is designed to slot directly into existing accelerator and switch designs. By replacing traditional electrical I/O with optical links, the technology delivers significant performance gains and cost savings in power-constrained environments. Following its latest financing, the company plans to ramp production and testing capacity, open and staff a new office in Hsinchu, Taiwan, and deepen ecosystem partnerships to speed customer deployments. Ayar Labs’ CPO solution is already qualified for industry-standard fab, manufacturing, and packaging flows, positioning it for broad adoption. The company has raised a cumulative $870 million to date, providing substantial capital to execute its global expansion and commercialization roadmap.

Participated · Series C · Mar 2022

Balbix offers the Balbix Security Cloud, a cybersecurity posture automation platform that uses AI and automation to help organizations inventory cloud and on-prem assets, run risk-based vulnerability management, and quantify cyber risk in monetary terms. Led by CEO and founder Gaurav Banga, the company plans to use the funds to build out go-to-market capabilities and continue engineering investments across unified asset inventory, software bill of materials (SBOM), vulnerability management, cyber risk quantification and security posture automation. The Series C follows growth in annual recurring revenue over the prior twelve months and four multi-million-dollar deals closed with Fortune 100 companies. Balbix launched an automated cyber risk quantification solution and reported more than 200% growth in new enterprise customers across the Americas, Asia, Europe, Middle East and Africa. The company has expanded its partner ecosystem with a new AWS integration and a partnership with global MSSP Microland. Balbix offers a predictive cybersecurity platform that uses artificial intelligence and machine learning to map an organization’s attack surface and surface prioritized vulnerabilities. The system integrates with existing operational security products to provide real-time visibility, reactive responses, and risk insights for stakeholders. Balbix’s approach measures a wide range of signals—from OS versions and weak passwords to poorly secured endpoints—so teams can decide and prioritize remediation based on data. The company does not charge by security event, aligning its incentives with breach prevention. Customers include enterprise organizations and global managed security service providers, and Balbix says it has international reach in sales. For now the company maintains a single office in San Jose but is weighing opening international offices to support growth. Balbix offers a predictive breach-risk platform that calculates and visualizes an enterprise’s cyber-breach risk and resilience across all devices, users and apps in an extended network. The system is designed to help CISOs and CIOs predict top breach scenarios, prioritize security mitigations and provide risk insights to prevent incidents before attacks occur. The company has just officially launched this platform and is using external funding to continue product development and expand operations. Balbix is led by founder and CEO Gaurav Banga. The company is based in San Jose, California.