Eastlink Capital

Recently, Eastlink Capital’s Portfolio Company, StreamNative, formed a partnership with Snowflake, in Real-Time Data Streaming. Following its successful collaboration with Databricks, StreamNative has now integrated with Snowflake Open Catalog, further solidifying its leadership in real-time data streaming and AI-driven analytics. Chris Child, VP of Product Management at Snowflake, emphasized the impact of this partnership: "We are thrilled to partner with StreamNative to bring seamless, cost-effective streaming data ingestion into Apache Iceberg through Snowflake Open Catalog. This integration will help customers with interoperability needs make real-time data AI-ready while ensuring governance across their data ecosystem. Together, we’re enabling organizations to apply open standards and unlock new levels of efficiency and value from their streaming data with Snowflake's data and AI platform." With the rapid adoption of generative AI,  the demand for low-latency, high-throughput real-time data continues to accelerate. However, enterprises face challenges in efficiently ingesting, managing, and utilizing real-time data at scale. StreamNative addresses this gap by simplifying and optimizing data ingestion, enabling businesses to leverage real-time insights without incurring high operational costs. This partnership represents a significant step forward in bridging real-time data pipelines with lakehouse storage, setting the stage for more efficient, scalable, and intelligent data management solutions across industries. Eastlink Capital is thrilled to be supporting StreamNative in this journey and will remain committed to supporting cutting-edge AI infrastructure developments and fostering key collaborations that drive the future of data and AI infrastructure.

How DeepSeek Shapes AI Investment Landscape – Trends in AI’s Next Frontier Over the past few weeks, DeepSeek has dominated conversations across the AI community. Its daily traffic now surpasses Claude, Perplexity, and even Gemini. Unlike resource-heavy, proprietary systems from LLM pioneers, DeepSeek-R1 is both cost-efficient and open-source—training costs were reported as $6 million, and inference expenses run ~93% lower than O1’s API. This shift is about more than just new models—it’s about redefining the entire AI infrastructure.  It doesn’t surprise us. As a VC fund focused on AI and Data infrastructure, we anticipated some of these developments and began deploying our AI investment strategies as early as 2021. Below are the key transformations we see shaping the AI landscape, especially catalyzed by the recent development from Deepseek: Open-source momentum: We anticipate an ever-growing ecosystem of accessible, high-performing models following DeepSeek’s lead. As strong believers in open-source, we started pursuing Databricks (Apache Spark) before it launched its enterprise products in 2015 and invested in StreamNative (Apache Pulsa) at its Series A round back in 2021—investments that underscore our conviction that collaborative, open platforms accelerate technology breakthroughs for everyone. Fine-tuning and customization:  As LLMs become more affordable, we see a future of specialized, highly customizable solutions enabled by lower training costs and greater efficiency. Our investment in Vellum(AI development platform) in 2023 underscores this vision: both leverage fine-tuned AI to address demands and fuel the next wave of domain-specific innovation. Smaller models can be just as great but Scaling Laws could still hold. Regardless, AI Inference matters more than training/fine tuning: Open-sourced LLMs have empowered a new set of distilled, specialized, and smaller models to be developed, and we believe these compact, efficient models are the future. Greater emphasis will be placed on post-training, particularly reinforcement learning, for reasoning and aligning model outputs with human preferences. Over time, LLMs will be commoditized. Our 2023 investments in Modal, which provides serverless AI inference infrastructure at large scale, exemplified such a conviction. Model Quantization & Local AI: By quantizing AI models, we can serve a broader audience, ensure data privacy, and harness local computational power. This philosophy guided our investments in Seraphic Security (privacy-first, localized AI) and MotherDuck (serverless analytics data warehouse with a single node). AI applications will blossom more and 2025 is the Year of Agents: With the costs of training and running LLMs going down, we will see a big wave of proliferation of AI applications, particularly AI agents for various verticals such as code generation, customer support, finance, legal, and productivity, etc. By Yiqiu (Rachel) L. Jun Wang and the Eastlink Team     

How DeepSeek Shapes AI Investment Landscape – Trends in AI’s Next Frontier Over the past few weeks, DeepSeek has dominated conversations across the AI community. Its daily traffic now surpasses Claude, Perplexity, and even Gemini. Unlike resource-heavy, proprietary systems from LLM pioneers, DeepSeek-R1 is both cost-efficient and open-source—training costs were reported as $6 million, and inference expenses run ~93% lower than O1’s API. This shift is about more than just new models—it’s about redefining the entire AI infrastructure.  It doesn’t surprise us. As a VC fund focused on AI and Data infrastructure, we anticipated some of these developments and began deploying our AI investment strategies as early as 2021. Below are the key transformations we see shaping the AI landscape, especially catalyzed by the recent development from Deepseek: Open-source momentum: We anticipate an ever-growing ecosystem of accessible, high-performing models following DeepSeek’s lead. As strong believers in open-source, we started pursuing Databricks (Apache Spark) before it launched its enterprise products in 2015 and invested in StreamNative (Apache Pulsa) at its Series A round back in 2021—investments that underscore our conviction that collaborative, open platforms accelerate technology breakthroughs for everyone. Fine-tuning and customization:  As LLMs become more affordable, we see a future of specialized, highly customizable solutions enabled by lower training costs and greater efficiency. Our investment in Vellum(AI development platform) in 2023 underscores this vision: both leverage fine-tuned AI to address demands and fuel the next wave of domain-specific innovation. Smaller models can be just as great but Scaling Laws could still hold. Regardless, AI Inference matters more than training/fine tuning: Open-sourced LLMs have empowered a new set of distilled, specialized, and smaller models to be developed, and we believe these compact, efficient models are the future. Greater emphasis will be placed on post-training, particularly reinforcement learning, for reasoning and aligning model outputs with human preferences. Over time, LLMs will be commoditized. Our 2023 investments in Modal, which provides serverless AI inference infrastructure at large scale, exemplified such a conviction. Model Quantization & Local AI: By quantizing AI models, we can serve a broader audience, ensure data privacy, and harness local computational power. This philosophy guided our investments in Seraphic Security (privacy-first, localized AI) and MotherDuck (serverless analytics data warehouse with a single node). AI applications will blossom more and 2025 is the Year of Agents: With the costs of training and running LLMs going down, we will see a big wave of proliferation of AI applications, particularly AI agents for various verticals such as code generation, customer support, finance, legal, and productivity, etc. By Yiqiu (Rachel) L. Jun Wang and the Eastlink Team     

We at Eastlink Capital are incredibly excited to announce that we're doubling down on our investment in Seraphic Security as part of its $29M Series A funding round! With enterprise browsers increasingly becoming the last unsecured edge, Seraphic is at the forefront of protecting enterprises, big or small. Their innovative, patented technology is already driving massive impact, with increasing adoption in the Fortune 500 and ambitious plans for market expansion. We’re proud to support Ilan Yeshua Avihay Cohen Suresh Batchu the exceptional team at Seraphic Security as they redefine browser security. More details on Seraphic and its vision here:  https://lnkd.in/eyJCCMvz #SeriesA #Funding #Cybersecurity #BrowserSecurity

We were thrilled to host the Eastlink Capital  2024 Annual Meeting on January 17, 2025. We celebrated the 10th anniversary of our firm with our LPs, portfolio founders, and industry leaders who have been pivotal in shaping our journey. At the event, our partners Steven Xi and Eric Ye shared the exciting momentum of our fund, reaffirming our continued vision for AI infrastructure. We kicked off the event with an exciting fireside chat with Francisco Alvarez-Demalde from Riverwood Capital. We were honored to have Ion Stoica, Co-Founder and Executive Chairman of Databricks to join us for a chat with Steven. Additionally, we had esteemed professors Kai Li(Princeton University) and Yiran Chen (Duke University), as well as Jamin Ball, Partner at Altimeter Capital, to join us for an insightful AI infrastructure panel discussion. A big shout-out to our portfolio companies for their presentations: Jordan Tigani, (CEO, MotherDuck), Kamal Gupta, (CEO, Omnistrate), Erik Bernhardsson, (CEO, Modal), Suresh Batchu, (COO, Seraphic Security), Rajeev Shrivastava, (CEO, TigerGraph). A heartfelt thank you to all our LPs,  portfolio founders, panelists, advisors, and guests for your trust and support—it drives us forward every day. A huge thank you to our incredible team—Jingxu Jia, Yiqiu (Rachel) L., and Tommy Ma—and, of course, our partners, Steven Xi and Eric Ye, for their hard work and leadership in organizing such a fantastic event. Stay tuned for more updates and events from Eastlink Capital! #AIInfrastructure #VC #AnnualMeeting #10yearsanniversary #founders  #AIInnovation

Big thanks to Modal for the holiday gifts: 'GPU' to keep our minds sharp and socks to keep us warm! Happy Holidays to everyone!

🎉 A Memorable Evening with AI and Data Infra Leaders! On Wednesday night, Eastlink hosted our Holiday Dinner, bringing together industry leaders in the AI and Data Infra space and founders from our portfolio companies. A special thank you to Rajeev Shrivastava, CEO of TigerGraph, Suresh Batchu, COO of Seraphic Security and Xiangyi Li, CEO of Benchmarkthing, for sharing their insights during the event. Wishing everyone a wonderful holiday season and looking forward to an exciting 2025 ahead! 🎉

Genmo has introduced Mochi 1, an open-source text-to-video generation model that is setting a new standard for realistic motion and video generation. The company also raised $28.4 million in a Series A round, led by New Enterprise Associates (NEA), with participation from The House Fund, Gold House Ventures, WndrCo, Eastlink Capital, and Essence VC. Kyt Dotson | SiliconANGLE & theCUBE https://lnkd.in/dGzwuJDy #AI #ArtificalIntelligence