Leveraging Trusted Execution Environment (TEE) technology for secure and scalable AI computation globally, the Exabits team is bringing the specialized expertise, knowledge, and resources required to manage this complex architecture

Exabits (https://www.exabits.ai/), a compute-based layer platform transforming high-performance GPU (graphic processing units) clusters into accessible investment assets, announces the integration of 4,000 NVIDIA H200 GPUs into its network. These GPUs are equipped with advanced TEE capabilities, providing secure and verifiable AI computations while ensuring data integrity and confidentiality.

As AI models and blockchain systems take on increasingly complex and sensitive tasks, maintaining trust and verifiability across distributed networks has become a challenge. In decentralized environments, where computation often spans multiple nodes and systems, ensuring the integrity and authenticity of data and processes is particularly difficult. With AI now supporting sectors like financial transactions and healthcare data (https://www.forbes.com/sites/hessiejones/2024/09/30/advancing-healthcare-with-data-the-critical-juncture-between-progress-and-privacy/) management, any security breach undermines its reliability and accuracy.

By incorporating TEE technology, Exabits offers high-level security, ensuring computations remain tamper-proof and that sensitive or proprietary data is handled with the utmost care. Exhibits’ new GPU infrastructure has been strategically deployed across the U.S. and Europe, interconnected by 3.2T InfiniBand architecture, to deliver the ultra-fast communication necessary for large-scale AI model pre-training. This configuration positions Exabits as the only player in the crypto space with in-house expertise to manage and scale a 4,000 GPU H200 architecture.

Following a successful $15 million seed funding round, Exabits has already achieved $10 million in annual recurring revenue (ARR) and holds a valuation of $150 million. Since its launch, the firm has been incubated by several reputable institutions, such as Stanford Blockchain Accelerator (https://sba.sites.stanford.edu/), Berkeley Xcelerator (https://xcelerator.berkeley.edu/), and is the first and only blockchain project supported by the Harvard Innovation Labs (https://innovationlabs.harvard.edu/). Recently, Exabits announced its collaboration with Seoul National University (https://en.snu.ac.kr/index.html), offering computing and processing power to its Vehicle Intelligence Lab.

Exabit’s integration of 4,000 NVIDIA H200 GPUs is part of its commitment to expand its offerings to Web2 and Web3 AI companies. These AI-ready GPUs allow Exabits to serve some of the most demanding enterprise clients, including those in autonomous driving, healthcare, and financial modeling sectors. In addition to commercial applications, Exabits plans to allocate a portion of its resources to accelerate advancements in science and technology, particularly in DeSci and AI Agent communities.

“We are thrilled to be at the forefront of the integration between AI and blockchain, as we recognize the immense potential of these two sectors," says Dr. Hoansoo Lee, Co-Founder of Exabits. "This advancement in our compute infrastructure marks not only a milestone for Exabits but for the broader AI community. With the power of TEE and our global GPU network, we are enhancing trust and security in AI computations, driving innovations in decentralized science (DeSci), AI Agent ecosystems, and beyond."

About Exabits: Established in 2021, Exabits is a revolutionary compute-base layer platform transforming high-end GPU clusters into accessible digital investment assets. With proprietary hardware and software, Exabits enables users to invest in GPU infrastructure, generating yield through tokenized compute assets. The company serves both Web2 enterprises and decentralized Web3 platforms, powering innovation through its scalable and secure infrastructure. To learn more, please visit https://exabits.ai/

Media Contact Ari Karp ReBlonde for Exabits ari@reblonde.com