Grishin Robotics

AI inference is moving from training scarcity to deployment economics. VSORA just added Ardian to its shareholder base through an undisclosed minority investment. The French fabless semiconductor company designs AI inference accelerators for data center workloads, with its Jotunn8 processor entering manufacturing after tape-out in 2025. The problem it is attacking is the memory wall. As models move into production, buyers care less about benchmark theater and more about latency, power draw and cost per query. VSORA says its architecture is built specifically for inference, aiming to lower infrastructure cost and energy consumption while keeping throughput high. The strategic signal is European AI infrastructure depth. Ardian is not backing an app layer company here; it is adding exposure to silicon that could sit under sovereign cloud, AI gigafactory and data center capacity plans. Quick facts👇 ● founders: Khaled Maalej; Julien Schmitt; Trung Dung Nguyen; Pierre-Emmanuel Bernard ● total capital raised: undisclosed ● HQ: Meudon-La-Forêt, France ● Investors: Ardian If inference becomes the dominant AI workload, the next infrastructure winners may be decided as much by power efficiency and memory architecture as by model quality.

kim.cc raised undisclosed institutional venture funding from Stellaris Venture Partners to scale AI-native customer support for e-commerce brands. The company sits between two models that keep disappointing merchants: fully automated chatbots that break on messy edge cases, and traditional BPO teams that struggle with speed, consistency and cost. kim.cc combines AI agents, workflow memory, quality checks and human oversight so brands can automate the routine work while keeping judgment in the loop. The traction is the interesting signal. kim.cc says it already serves 200+ Shopify merchants, manages large-scale support workflows, and delivers up to 40% lower support costs without giving up quality. That turns the round into more than a customer-support software story. It is a bet that service delivery itself becomes AI-native. Quick facts👇 ● founders: Sachin Jaiswal; Phani Yedavilli; Kaushik Barodiya ● total capital raised: undisclosed ● HQ: San Francisco, California ● Investors: Stellaris Venture Partners If AI support moves from ticket deflection to full operations, the winners may be the companies that understand workflows, QA and accountability better than chat interfaces.

���️ Episode 19 of Bits and Atoms is live! This episode, our host Elias Can sits down with Shane Parmar, Co-Founder and CEO of ONDA Plus | Wearable Mental Wellness, to explore the journey of building technology that tackles stress before it becomes a problem. Shane shares how one of the most difficult periods of his life inspired the creation of ONDA+, an AI-powered wearable platform designed for real-time stress detection and early intervention. The conversation dives into the realities of building a hardware startup, the challenges of product validation and fundraising, lessons learned from entrepreneurship, and the mindset required to stay resilient through the highs and lows of the founder journey. Full video link in the comments 👇 📣 Applications for our Autumn ’26 cohort are still OPEN! 👉 Apply now: https://lnkd.in/epxT4RBF

AI data centers are usually discussed through GPUs, networking and power. Stathera, Inc. is betting that precision timing belongs in that same infrastructure conversation. The Montreal fabless semiconductor company raised US$55M in Series B funding led by Maverick Silicon, with Celesta Capital, BDC Capital, MediaTek Innovation Fund, TXC Corporation and Ultratech Capital Partners participating. Total capital raised is now US$75M. Stathera, Inc. builds MEMS-based silicon timing to replace legacy quartz components. Its GEN2 portfolio is moving into mass production, while the new GEN3 platform is being designed for AI, communications, enterprise and data-center systems. The AI link is less flashy than another accelerator launch, but important: GPU clusters, switches and optical interconnects need precise synchronization as requirements move from microseconds toward nanoseconds. If timing becomes a constraint on cluster performance, independent silicon timing can become a strategic layer rather than a commodity component. Question for the next AI infrastructure cycle: how much value moves into the invisible control layers that keep dense compute systems coherent? Quick facts👇 ● founders: George Xereas; Ed Harvey; Charles Allan ● total capital raised: US$75M ● HQ: Montreal, Canada ● Investors: Maverick Silicon; Celesta Capital; BDC Capital; MediaTek Innovation Fund; TXC Corporation; Ultratech Capital Partners

Homebuilding is usually framed as land, labor and lumber. Higharc is betting that the next big lever is AI across the design-to-construction workflow. Higharc raised $95M in Series C funding led by Insight Partners, with Wellington Management, Fifth Wall, Spark Capital and Lux Capital among the participating investors. The round brings total funding to more than $170M. The Durham company turns homes into 3D spatial data, then uses that structured model to automate design, construction documents, real-time estimating and shoppable sales experiences. Its new AI Estimating product extends that workflow to building-materials distributors, with US LBM as the first partner. The interesting part is not another AI wrapper on top of CAD. It is the attempt to make the house itself a structured data object that builders, suppliers and buyers can all work from. If that layer becomes trusted, AI can move from design assistance into pricing, material planning and production coordination. The real prize may be the workflow layer that links design, estimating and production end to end. Quick facts👇 ● founders: Marc Minor; Michael Bergin; Peter Boyer; Thomas Holt ● total capital raised: >$170M ● HQ: Durham, North Carolina ● Investors: Insight Partners; Wellington Management; Fifth Wall; Spark Capital; Lux Capital

LeapXpert has raised $180M to push governed enterprise messaging from compliance archive into AI infrastructure. Riverwood Capital led the growth investment, with existing investor Portage also supporting the round. The product edge is clear: employees and customers already talk through WhatsApp, iMessage, Signal and WeChat, while regulated companies still need capture, governance, retention and auditability. The AI angle is not another chatbot. LeapXpert is trying to make off-channel conversations usable enterprise data. If a bank, government agency or Global 2000 company can govern those conversations in real time, it can also extract intent, risk signals and relationship intelligence from data that used to sit outside core systems. That is why this round feels like enterprise AI plumbing rather than communications software alone. The valuable layer may be the governed record of what customers and employees actually say. Quick facts👇 ● founders: Dima Gutzeit; Avi Pardo; Rina Charles ● total capital raised: >$222M ● HQ: New York, NY ● Investors: Riverwood Capital; Portage Can governed messaging become the system of record for customer conversations before the big collaboration suites get there?

Together AI just closed an $800M Series C at an $8.3B valuation, led by Aramco Ventures. This is not only another GPU-cloud round. It is a bet that enterprises will move more AI workloads toward open-source models, specialized inference stacks, and providers that can make tokens cheaper without forcing teams into one closed frontier model vendor. Together AI sits in that shift as an AI neocloud renting Nvidia GPU clusters and AI-specific infrastructure. The timing is sharp: the company says annual bookings passed $1.15B last quarter, while open-source model usage has tripled across the industry over the past year. The interesting question now is whether AI infrastructure winners are decided by who controls GPUs, who improves serving economics, or who can package both into reliable production workflows. Quick facts👇 ● founders: Vipul Ved Prakash; Ce Zhang; Chris Ré; Tri Dao; Percy Liang ● total capital raised: ≈$1.3B ● HQ: San Francisco, California ● Investors: Aramco Ventures; General Catalyst; NVIDIA

Amity Robotics is taking AI concierge robots from mall lobbies and hotel floors into a wider Asia rollout. The Thailand-based company raised $7M in seed financing, combining equity led by East Ventures and debt led by AlteriQ Global, with 500 Global also participating. Its core product, ARC Base, is a physical AI kiosk that answers questions, guides visitors and supports service teams across large properties. The company says ARC Base is already live in more than 30 properties across Singapore, the UAE, Saudi Arabia, Hong Kong, Thailand, Malaysia and Indonesia, with a first European deployment planned for July. The next step is ARC Move, a mobile robot designed to approach, help and guide customers around stores and large properties. This is a useful test for physical AI: service robots do not need to look futuristic to matter if they can reduce queues, answer in multiple languages and create measurable ROI in high-traffic spaces. Quick facts👇 ● founders: Korawad Chearavanont ● total capital raised: $7M ● HQ: Bangkok, Thailand ● Investors: East Ventures; AlteriQ Global; 500 Global

Oxmiq Labs has raised $35M in Series A funding to scale OxCore, its licensable GPU architecture for custom AI silicon. Fundomo and Samsung Catalyst Fund co-led the round, bringing total capital raised to $60M, with MediaTek, Pegatron Venture Capital, Morgan Creek Capital Management, LLC and other strategic investors participating. The product bet is specific: OxCore combines a CUDA-compatible GPU engine, tensor processing engine and orchestration engine into a single licensable core. OxQuilt then lets customers adapt chiplets, memory, interconnects and packaging across different supply chains. AI infrastructure buyers increasingly want their own compute roadmap, but a full custom chip program is expensive, slow and tied to narrow ecosystem choices. OXMIQ Labs is trying to make the Arm-style IP model work for inference-era GPUs, from physical AI devices to data centers. The contrarian angle: the next AI compute winners may not all be chip companies selling complete chips. Some may be IP and software companies that make custom silicon cheaper to build. Quick facts👇 ● founders: Raja Koduri ● total capital raised: $60M ● HQ: Campbell, California ● Investors: Fundomo; Samsung Catalyst Fund; MediaTek; Pegatron Venture Capital; Morgan Creek Capital Management, LLC

Factory AI wins on deployment reliability, not demo intelligence. CarbonSix raised $40M in Series A funding to scale physical AI for manufacturing, with DSC Investment and LB Investment co-leading the round. The company is building deploy-ready robotic intelligence software plus robotic hands and manipulators for real production lines. Its bet is practical: give factories automation tools they can use immediately, capture task-specific data from daily operations, then feed that data back into better robot skills. That feedback loop matters because high-mix, high-variance manufacturing is where traditional path-based automation often breaks. CarbonSix is trying to turn operator know-how into reusable robot intelligence for cable fastening, film peeling and other variable factory tasks. Will the winners in physical AI be the teams with the biggest models, or the teams that own the factory feedback loop? Quick facts👇 ● founders: Jehyeok Kim; HJ Terry Suh; Terry Moon; Hunmin Cho ● total capital raised: $44M+ ● HQ: San Francisco, California ● Investors: DSC Investment; LB Investment; IMM Investment Corp.; Korea Development Bank; SV Investment; Foothill Ventures; Storm Ventures