Participated · Series C · Jul 2026

Etched.ai develops rack-scale inference systems and broader frontier inference infrastructure. Since founding, the company has grown to more than 400 employees, hiring over 200 people after its Series B late last year. It has secured more than $1 billion in customer orders and established deep manufacturing and supply-chain partnerships, including opening a factory in Taiwan. Etched is currently validating its first rack-scale inference product with customers. The company’s near-term plans emphasize production, manufacturing scale-up, and deployment of its inference systems, supported by cumulative funding of $800 million.

Participated · Series H · May 2026

Anthropic is a premier artificial intelligence safety and research company renowned for developing Claude, a family of highly capable, steerable, and safe large language models. Founded by former OpenAI researchers, Anthropic pioneers techniques like "Constitutional AI" to build systems that are helpful, honest, and harmless. Beyond its core research, the company has rapidly scaled into a dominant market force, experiencing exponential revenue growth driven by widespread enterprise adoption. Anthropic has attracted unprecedented investor attention, recently raising approximately $30 billion at a $380 billion valuation, with ongoing negotiations potentially boosting its valuation to roughly $800 billion. As a leading competitor in the generative AI space, the company is reportedly exploring an IPO later this year.

Participated · Equity · Nov 2025

Fabric8Labs, founded in 2015 and headquartered in San Diego, has created a patented Electrochemical Additive Manufacturing (ECAM) platform that builds metal parts atom-by-atom at room temperature, eliminating costly post-processing while achieving exceptional feature resolution and surface finish. The technology enables production of next-generation components such as direct-to-chip cooling structures, 3D antennas printed on PCBs, and high-current interconnects for electric vehicles. With ECAM, customers can move seamlessly from prototyping to high-volume output on a single manufacturing line. Backed by an ISO9001-certified, ITAR-registered U.S. facility, Fabric8Labs plans to expand production capacity from 5 million to 22 million components annually. The company is hiring across manufacturing, design, quality, and process engineering to meet demand from AI/HPC, wireless communications, and power electronics sectors. Its U.S. manufacturing presence also addresses supply-chain resiliency requirements for advanced electronics customers.

Participated · Series B · May 2025

Avicena, based in Sunnyvale, CA, develops LightBundle, a microLED-based ultra-low power, high-density optical interconnect architecture. Its technology uses arrays of GaN microLEDs integrated onto CMOS ICs, coupled into multi-fiber cables and matching silicon detectors to deliver high bandwidth density, energy efficiency, reliability and lower cost versus laser-based solutions. The company targets AI datacenters, disaggregated memory fabrics, AI/ML, HPC, sensors, 5G and aerospace applications and says the approach leverages the large display ecosystem. Avicena positions its product to enable disaggregation of compute and memory and to grow system throughput while reducing energy impact. It will use new funding to scale its team and ramp its first product into production. Financially, Avicena raised $65M in a Series B, bringing total capital raised to $120M. AvicenaTech develops ultra-low energy optical links based on microLEDs, marketed as its LightBundle I/O solution. LightBundle is built from arrays of GaN micro-emitters that can be integrated onto high-performance CMOS ICs and leverages the microLED display ecosystem for high-volume, low-cost production. The technology is designed to eliminate copper-link bottlenecks in power, latency, reach, and bandwidth density and to enable higher-performing architectures in HPC, AI/ML, cloud computing, next-generation cellular radio, remote sensing, and aerospace. Avicena positions its products for chip-to-chip interconnects, memory disaggregation, and medium-reach links across datacenter and wireless use cases. The company is privately held and located in Mountain View, CA, and is led by founder and CEO Bardia Pezeshki. Avicena says it will use the $25M Series A to scale its team and build initial products for partners and customers.

Participated · Series A · Dec 2022

NeuReality develops purpose‑built AI inference system architecture, hardware and software under its NR1 AI Inference Solution, including the NR1 NAPU™ 7nm server‑on‑a‑chip. Its NR1‑M™ and NR1‑S™ PCIe cards integrate into server racks and route AI tasks independently of CPUs to improve utilization and energy efficiency. The company says its systems drive 100% AI accelerator utilization and connect directly to Ethernet to manage large AI query pipelines. NeuReality has delivered NAPUs from TSMC and demonstrated compatible server configurations with partners including AMD, IBM, Lenovo, Qualcomm and Supermicro. It is running early deployments with select cloud and enterprise customers across financial services, business services and government and aims to accelerate broader deployment into additional markets and generative AI use cases. The company’s additional capital will be used to shift from early deployment to growth and expand customer adoption. NeuReality develops the Network Attached Processing Unit (NAPU) family — including the FPGA‑based NR1, the NR1‑M PCIe module and the multi‑module NR1‑S — combined with a software stack (SDK, deployment manager, monitoring) to simplify AI inference deployment. The company targets cloud data centers, on‑premises edge systems and enterprises running computer vision, NLP and recommendation workloads. NeuReality says its hardware and software optimize data movement, scheduling and virtualization to improve inference efficiency, though it has not published independent benchmarking to validate some performance claims. The startup has partnered with Xilinx for production and with IBM as a design partner; it has been shipping prototypes since May 2021. NeuReality was co‑founded in 2019 and currently has about 40 employees, with plans to hire 20 more over the next two fiscal quarters. To date the company has raised $48 million in venture capital and will use its latest funding to finalize chip design and begin customer shipments. NeuReality is developing a novel AI inferencing platform that removes the CPU-centric bottleneck to allow hyperscale clouds and data centers to offload ML models. The team applies techniques from networking and storage to create a bottom-up architecture it describes as the next evolution of AI computer engines. The company claims its system can deliver about 15x performance per dollar for basic deep learning offloading and larger gains when the entire pipeline is offloaded. NeuReality has working prototypes implemented on a Xilinx FPGA and plans to offer fully custom hardware early next year. Target customers include hyperscale cloud providers, data center owners, software solutions providers such as WWT, and OEMs/ODMs. The founding team includes CEO Moshe Tanach, VP of operations Tzvika Shmueli and VP for very large-scale integration Yossi Kasus, and Naveen Rao (former Nervana CEO and Intel AI Products Group GM) is joining the board.