Primary is the premier early stage venture firm working alongside founders to build unicorns like Alloy, Chief, Electric, K Health, Slice, Vestwell, and many more.  The firm takes a low-volume, high-conviction approach—and devotes unreasonable resources toward making sure founders can succeed. With an Impact team that outnumbers investors 2:1, operational experts spend 100% of their time making hires, signing customers, and enabling future fundraising to supercharge growth. This partnership with top-tier founders makes companies in Primary’s portfolio 2x more likely to raise a Series A and 19x more likely to reach a billion-dollar-plus valuation—a third of Fund I companies are unicorns, less than a decade out.  We’re always excited to meet 99th-percentile founders building in B2B SaaS, built world, fintech, healthcare, infrastructure, and supply chain. And while we have a home court advantage in NYC, we invest globally, at pre-seed through A.

Primary Venture Partners

We invest in pre-seed and seed stage companies in AI/ML infra, open-source, cybersecurity and Vertical AI/SaaS.

Firestreak Ventures

Positive Sum is a venture capital and private equity firm that helps teams build extraordinary businesses.

Positive Sum

Fontinalis Partners, with offices in Detroit and Boston, is a venture capital firm strategically focused on next generation mobility. Fontinalis’ mission is to leverage the team’s considerable management experience, market access, strategic relationships, international expertise, and background in transportation innovation to scale companies providing next-generation mobility solutions. Fontinalis invests as a strategic partner across all facets of the world’s mobility systems on a stage- and size-agnostic basis. Fontinalis is not affiliated with Ford Motor Company.

Fontinalis Partners

Two Sigma Ventures is an early-stage venture capital fund based in New York City investing capital and expertise in entrepreneurs leveraging data and computing to create the future. In partnership with our parent company, Two Sigma Investments, we provide entrepreneurs access to a proprietary network of experts and startup resources to give them a competitive edge.

Two Sigma Ventures

Hummingbird Ventures is a venture capital partner for outlier founders building generational companies. Hummingbird teamed up with many transformative companies, including Kraken, Peak Games, Deliveroo, BillionToOne, Showpad, OneCard, Pristyn Care, Gram Games and many others.

Hummingbird Ventures

Industry-transforming companies & strategic capital partners

Dolik Ventures

Skybox Datacenters is a leading mission critical data center owner, developer and operator with a focus on wholesale enterprise facilities. Our team’s experience comprises 50 million square feet of real estate development and over $435 million worth of mission critical transactions throughout North America.

Skybox Datacenters

Oceans is a venture capital firm focusing on giving startups the hands-on support they need to succeed. They are a team that builds teams, a culture that shapes cultures, and a group of mentors who unlock skills and talent to build a future of technology. They help founders build the foundations that unlock the power of people. Oceans was founded in 2018 and is headquartered in New York, USA.

Oceans

Earthshot Ventures is an early-stage venture capital firm that supports entrepreneurs developing creative solutions to address climate change. Earthshot harnesses the power of the global network our team has built over the last two decades to accelerate climate technology adoption and create iconic brands. The fund is led by Mike Jackson who has spent 15 years as an entrepreneur and VC in climate tech, and Dawn Lippert, founder and CEO of Elemental Impact.

Earthshot Ventures

Fundomo

Lightscape Partners is a venture capital firm.

Lightscape Partners

Etched.ai

Etched.ai is the developer of Sohu, an AI chip optimized for transformer models. By embedding the transformer architecture into its chips, Etched.ai is pioneering the creation of the  servers for transformer inference.

Etched.ai is a chip startup founded by Harvard dropouts Gavin Uberti and Chris Zhu, focusing on developing specialized artificial intelligence (AI) accelerator chips dedicated to large language model (LLM) acceleration. Their aim is to dramatically enhance the performance and reduce the cost of running AI models like GPT-4 by designing purpose-built chips optimized for the transformer architecture used in LLMs.

Empowering Tomorrow's Innovations with Generative AI Infrastructure

Generative AI Infrastructure

Transformer model chipmaker Etched.ai raises $120M to challenge Nvidia's market dominance

Etched.ai is a chip startup founded by Harvard dropouts Gavin Uberti and Chris Zhu, focusing on developing specialized artificial intelligence (AI) accelerator chips dedicated to large language model (LLM) acceleration. Their aim is to dramatically enhance the performance and reduce the cost of running AI models like GPT-4 by designing purpose-built chips optimized for the transformer architecture used in LLMs.Etched.ai's first chip, codenamed Sohu, is designed to have substantial memory and support large batch sizes, enabling it to achieve 140 times higher throughput per dollar compared to an Nvidia H100 PCIe card when processing GPT-3 tokens. The company claims that this significant performance uplift is primarily driven by impressive throughput capabilities rather than a substantial cost differential.By specializing in LLM acceleration, Etched.ai aims to provide a more efficient and cost-effective solution for organizations seeking to leverage the power of large language models.

Founded in Israel, NeuReality specializes in providing hardware infrastructure to streamline computation processes to accelerate the development of AI applications. Its technology assists AI inferencing deployment by offering a comprehensive system solution that seamlessly supports a vast array of deep learning models.NeuRealitys’ infrastructure includes an architecture called the Network Addressable Processing Unit (NAPU) to maximize the capabilities of Deep Learning Accelerators (DLAs). This design facilitates inference through purpose-built NAPUs that are specifically crafted for AI inference tasks, enhancing performance and efficiency.The NR1 AI Inference Solution by NeuReality is a comprehensive system comprising architecture, hardware, and software platforms. Launched in November 2023, it intends to simplify the installation, utilization, and management of AI, making it accessible and manageable for users.The hardware offerings of the solution include,1. NR1 Network Addressable Processing Unit : A network addressable inference Server-on-a-Chip with an embedded Neural Network Engine, which is a workflow-optimized hardware device with specialized processing units, native network capabilities, and virtualization capabilities2. NR1-M™ AI Inference Module : A full-height, double-wide PCIe card containing one NR1 Network Addressable Processing Unit (NAPU) system-on-chip and a network addressable Inference Server that can connect to an external Deep Learning Accelerator (DLA)3. NR1-S AI Inference Appliance : An optimized design for an inference server that contains NR1-M modules with the NR1 NAPU, which enables disaggregated AI service in a scalable and efficient architectureThe software component of the solution is designed to be compatible with any trained model from any development environment. It includes tools that handle the entire AI pipeline and enable seamless connection of AI workflows to any environment.NeuReality differs from traditional AI chip connectivity in replacing CPU-centric support with an AI-centric platform. This allows components to communicate directly with AI accelerators, eliminating the need for commands to pass through a central processing unit. Additionally, the technology enables AI accelerators to connect with each other, enhancing efficiency and performance in AI applications. It also claims its AI-centric approach reduces complexity, cost, and power usage. The firm claims it can reduce inference costs by 90%.Key customers and partnershipsIn November 2021, NeuReality teamed up with IBM to develop advanced AI inference technology by enhancing its NR1 chip and system for use in IBM's Hybrid Cloud.In September 2021, the company partnered with Xilinx to develop AI-centric platforms that improve real-world AI applications. This collaboration was intended to focus on NeuReality's AI-centric inference platform NR1-P and address existing issues with AI deployment including cost and complexity.

NeuReality

<div>
 
 Accubits is an AI and BLOCKCHAIN focused development and solutions company based in Washington DC with its development offices in India and Dubai. Pioneering as a custom Blockchain solutions provider delivering solutions ever since 2012, Accubits is one of the oldest and experienced Blockchain dev house in the market and is featured as one among the top 13 Blockchain companies in India by INC42. We are a big family of engineers, innovators, inventors and many others continually working to help our customers in leveraging their business and adding value to their lives. Over the past 6 years, We have delivered more than 600 projects, for over 200 clients worldwide. Several of Fortune 500 companies in sectors from Retail to Health are being powered by our solutions to streamline and leverage their growth.Our portfolio for AI solutions includes intelligent conversational tools, RPAs, CRMs, SIPs, real-time video and image analytics systems, state of the art ALPR systems and many more being used enterprises, startups across North America, Canada, Middle East, Hong Kong, Singapore, Indonesia, India &amp; Sweden. Our portfolio of Blockchain development includes custom DApps and smart contracts, crypto token and exchange development, integrated blockchain services and ICO platforms.
 
</div>

Accubits

<div>
 
 LeewayHertz is a leading software development company delivering tailor-made digital solutions to businesses worldwide. Our team of 250+ full-stack developers, designers and innovators has designed and developed 100+ digital solutions across industry verticals. As a close-knit team of AI and web3 experts, designers and full-stack developers, we undertake collaborative research and development to create next-gen applications and solutions perfectly suited to the evolving tech space. As a technology pioneer with deep knowledge and expertise in AI, blockchain and other advanced technologies like IoT, LeewayHertz believes in helping companies overcome their most complex tech challenges and drive business growth. Our motto, &lsquo;You Dream It: We Build It,&rsquo; reflects our business philosophy. From putting flesh on your idea to delivering the expected end product, we do whatever it takes to make your project successful. Having worked with 30+ top fortune 500 companies and created various on-demand solutions, we boast an impressive work portfolio demonstrating our expertise.
 
</div>

LeewayHertz

Databricks, a data and AI company established by the original developers behind prominent Big Data solutions such as Delta Lake, Apache Spark, and MLflow, and provides organizations with AI and data analytics tools for efficient management of their data.&nbsp;Databricks' primary product is the Databricks Lakehouse platform, a unified data intelligence platform that merges a data warehouse and data lake to provide developers with efficient access to data and enables enterprise customers to create their AI models—similar to a ChatGPT-style AI system—using only their internal data.&nbsp;Databricks Unity Catalog, part of the Databricks Lakehouse Platform, provides unified governance for structured and unstructured data, machine-learning models, notebooks, dashboards, and files across various environments, offering automated monitoring, proactive alerts, and seamless data sharing through Delta Sharing integration. Furthermore, Databricks offers an open architecture that streamlines the AI development process, allowing the integration of various LLMs such as Llama 2, Azure OpenAI, and Anthropic, enabling customization with enterprise data for tailored model creation. In&nbsp;<a href="https://sp-edge.com/updates/28436">April 2024</a>, Databricks introduced its Data Intelligence platform for energy to use AI technology to process large data for energy companies. Following this, the company added synthetic data capabilities for AI agent evaluation in&nbsp;<a href="https://sp-edge.com/updates/37008">December 2024</a>.Notably, Lakehouse leverages NVIDIA technology for accelerated data processing and AI model generation. Databricks also offers industry-specific solutions like Databricks Lakehouse for Healthcare, Financial Services, Retail, and Manufacturing.Databricks has expanded its offerings through multiple acquisitions, including MosaicML (<a href="https://sp-edge.com/updates/19683">July 2023</a>), Arcion (<a href="https://sp-edge.com/updates/23041">October 2023</a>), Okera (<a href="https://sp-edge.com/updates/18197">May 2023</a>), and Datajoy (October 2022).&nbsp; Key customers and partnershipsAs of November 2024, Databricks served a global customer base of more than 10,000 organisations, including Comcast, Conde, DuPont, Honeywell, Rolls-Royce, Shell, and Tata Steel.Databricks collaborates with over 1,200 global cloud, ISV, and consulting partners to deliver data, analytics, and AI solutions, supporting customers in scaling their projects.Notable partnerships include collaboration with AWS in&nbsp;<a href="https://sp-edge.com/updates/34884">October 2024</a> to leverage AWS Trainium chips as the preferred AI chip to power Databricks' Mosaic AI model training and serving capabilities.&nbsp;Databricks also partnered with Karini AI in&nbsp;<a href="https://sp-edge.com/updates/33687">September 2024</a> to accelerate GenAI deployment for enterprises.&nbsp;

Databricks

Labelbox offers a data-centric AI platform for enterprises, enabling them to develop, optimize, and use AI to solve complex problems and power new products and services.&nbsp;&nbsp;Labelbox’s platform offers capabilities for building, optimizing, training, fine-tuning LLMs, and AI-enabled data labeling automation.&nbsp;Its products include&nbsp;1) Catalog: A platform for building AI applications with features such as advanced search capabilities and optimizing data for labeling budgets2) Annotate: A data labeling solution designed to streamline and accelerate labeling operations using automation tools and pre-labeling techniques(3) Model: Offers tools for data enrichment, model-assisted labeling, and custom Foundry Apps to automate data management tasks and improve AI model performance&nbsp;&nbsp;(4) Boost: Offers fully managed tailored solutions such as on-demand data labeling services to enhance AI development speed and efficiency.&nbsp;As such, Labelbox enables companies to streamline the fine-tuning and alignment of LLMs and enables the validation of outputs through techniques such as reinforcement learning with human feedback (RLHF), reinforcement learning from AI feedback (RELIEF), and evaluation. This incorporation of human expertise ensures more accurate, contextually relevant, trustworthy, and secure results from LLM systems.The platform is offered through a SaaS subscription model with packages ranging from free subscriptions with limited usage and custom pricing for enterprises.Key customers and partnershipsThe firm's solutions have use cases across diverse industries, including retail, consumer internet, healthcare, manufacturing, and financial services. Labelbox has a large portfolio of customers, including NASA's Jet Propulsion Laboratory and Blue River Technology, a smart AgriTech company that used Labelbox's model-assisted labeling suite to enhance its See &amp; Spray technology in smart agriculture, resulting in herbicide cost savings of 70%–80. Labelbox's customers also include several Fortune 500 companies, such as Walmart, P&amp;G, Audi Adobe, Airbnb, Bayer, and Bosch.Key partnerships include a long-term collaboration with <a href="https://sp-edge.com/updates/21770">Google Cloud </a>to facilitate digital transformations in retail, consumer internet, financial services, and manufacturing and logistics.&nbsp;

Labelbox

Integrated solutions that handle various processes such as model training, fine-tuning, model deployment, monitoring, and maintenance in working with and deploying LLMs

Platforms that provide hardware such as high-performance CPUs, GPUs, or TPUs, including cloud-based solutions that are commonly used to train and run GenAI models efficiently

Platforms that assist with fine-tuning pre-trained GenAI models or custom models tailored to specific use cases. Includes services that train models from scratch or use existing models

Platforms that store and retrieve training data in vector databases for access during the model training phase

Services to deploy GenAI models (on-premises or cloud-based) into production environments and integration tools to support existing systems and workflows

Ongoing monitoring of GenAI models to ensure performance. Includes regular model retraining, performance optimization, and troubleshooting

Platforms that assist in designing and refining prompts or instructions to GenAI models to get the desired responses or outcomes

Platforms that enable safe, responsible, and ethical GenAI deployment, ensuring compliance with regulations and/or protection against GenAI-based cybersecurity threats.

All services offered by these companies provide support to AI/ML

NLP is a key component of generative AI which is the core focus of this hub.

All companies within this space are offering infrastructure for generative AI

Generative AI infrastructure hub covers companies that deploy and run these foundation models.

ANALYST_QUICK_TAKE

<div>
 
 Etched has secured USD 120 million in a Series A investment round, with notable contributions from ex-PayPal CEO Peter Thiel and Replit's CEO Amjad Masad. This funding will facilitate the development of Etched's specialty chip, optimally designed for transformer models popular among AI firms, to enhance the content generation and response capacity of AI models, also known as inference.
 Etched has also teamed up with Taiwan Semiconductor Manufacturing Co. (TSMC) for chip production. The investment is projected to offset the costs of transferring its designs to TSMC and chip manufacturing, a process known as "taping out" a chip.
 
</div> Analyst QuickTake: 
 Etched is challenging NVIDIA in the AI chip market with its new ASIC, which reportedly offers 20x faster performance than NVIDIA's H100 GPUs. This advanced chip aims to provide more efficient and powerful solutions for AI workloads, positioning Etched as a strong competitor in the AI hardware space​.

Etched.ai

About EDGE

Contact Us

Resources

Area of Focus

Technology Research Made Effortless