Technology And Domain Experts
Exabits offers enterprise-grade compute resources to companies like Viggle, Lepton.ai, FastGPU, IO.NET, Aethir, EMC, Akash, MyShell, Nebula Block, Game Killer, and many more. Our deep understanding of the compute industry and our ability to meet users' expectations generate a 300% compound quarterly revenue growth. This financial trajectory underscores the effectiveness of our approach and the robust demand for our services.
Our team's domain expertise in building and monetizing extensive clusters of enterprise-grade GPU infrastructure ensure that we provide a cost-effective, high-performance AI compute financialization platform. With proprietary software and hardware, Exabits significantly reduces operating costs and boosts compute output, maximizing revenue for all participants in our ecosystem.
Unique access to high-value, scarce GPUs
GPU clouds are very different from traditional CPU clouds. The supply chain is tight, and GPUs are hard to obtain, especially popular cards like the A100, H100, and even the RTX 4090. See Appendix A for a detailed explanation of the “Procurement Challenges of High-End GPUs.”
These high-end GPUs are oversubscribed, making it nearly impossible to acquire idle resources. In fact, idle GPU resources are generally low-end, outdated graphics cards with no market value. Regardless of their quantity (often advocated by crypto GPU platforms), there is no demand for them.
Exabits has unique access to the coveted enterprise-grade H100 and A100 GPUs (and soon the new B200s) due to its decades of experience in the data center business, further democratizing AI development. Members of our team are tier 1 resellers of Nvidia, AMD, and SuperMicro, who have completed deals worth over $2,000,000,000 for data centers and mining facilities, and have experience inside organizations such as Bitmain and mobile operators. Because of our experience in building data centers, we have kept our rolodex warm and have grown it throughout the years, allowing us to procure our initial partners.
Highly available services that satisfies users
Unlike traditional CPU clouds, GPUs lack the versatility to serve multiple users. Idle CPUs can be easily redirected to run other workloads or serve other customers, but GPUs cannot. This means that directly allocating an idle GPU to a user in an Airbnb-like fashion is unlikely to work, which is one of the reasons why many decentralized GPU platforms fail to attract real users.
Exabits has built a world-leading GPU operation and management platform, featuring multi-tenant automatic resource management, GPU process-level deep monitoring, adaptive resource allocation and scheduling, and real-time automatic resource scaling. Users can launch and use Exabits' GPU resources in one second, meeting their business needs in real-time.
Full-stack capability to build high-performance GPU Clusters
The barriers to entry for a GPU cluster go beyond the GPUs themselves and include the data centers that can host the GPU cloud. Unlike traditional CPU clusters, GPU clusters require data centers with extremely high specifications. For example, H100 servers must be hosted in at least T3 data centers that meet specific specifications. Currently, there are fewer than 10 data centers in the U.S. and fewer than 5 in Southeast Asia capable of hosting over 1,000 H100 GPUs. These data centers are very selective and have high qualification requirements for partners, usually requiring an audit of the partner's operational capabilities and track record. They are not likely to provide idle capacity for web3 projects.
Exabits is the only team in the crypto field capable of securing these high-end data centers and deploying 1,000 H100 GPUs within them. Therefore, several crypto GPU projects are in discussions with Exabits to collaborate, with Exabits addressing the data center requirements, building the H100 service clusters, and leasing them to these projects.
Capability to operate large-scale GPU clusters and serve enterprise-level customers
The deployment and operation of GPU clouds are very complex. Because these jobs span thousands of GPUs, you need to ensure you have reliable infrastructure, and then you need a backend network to communicate faster and sustain operations for extended periods. It is not simply a matter of buying a bunch of GPUs, hooking them together, and expecting them to work seamlessly. Achieving optimal performance requires significant system-level optimization and extensive experience across multiple generations.
Exabits is the only team in the crypto field with the capability to interconnect over 1,000 H100s. It is also the only team in the Crypto+AI industry capable of serving large-scale enterprise AI customers. For example, Exabits is currently the only team in the Crypto+AI industry that has successfully supported LLM pre-training:
Urgently needed onboarding solutions
LLMs and GPU clouds are new technologies, and it is common for customers to rent GPUs but not know how to use them. This has become a widespread pain point in the industry. Many enterprises using AI are not cost-driven. It's not that compute is too expensive; it's about figuring out how to create business value. Globally, there's a growing realization that necessary onboarding solutions are crucial for new tech adoption. Usually, it's not "the cheapest wins" but "the most implementable wins profits." Right now, the implementation layer is missing from GPU marketplaces.
AI infrastructure needs to hone its high-precision and cutting-edge capabilities. Price is just one factor; the ability to achieve low cost, high speed, and high concurrency—or the classic "car salesman problem" of fast, cheap, and good—is a core competency.
Exabits can achieve this, but other players in the crypto field cannot. As a result, Exabits now has $400k MRR and continues to grow rapidly, while other crypto GPU clouds are still in the narrative stage, with no real users or revenue.
Unique technology to accelerate consumer level GPUs to provide the best cost-effective solution
In addition to rare data center-grade GPUs like the H100/A100, high-end consumer-level GPUs offer strong computational performance at a relatively lower price, making them a cost-effective choice for many inference scenarios. Exabits develops unique technology to accelerate consumer-level GPUs, enabling them to perform similarly to H100/A100 in some cases.
This allows Exabits to leverage a large number of existing high-performance consumer GPUs to provide the most cost-effective computing services to clients. This capability is crucial, considering many companies face years-long waiting lists to acquire enterprise-grade GPUs or experience bottlenecks securing instances at other cloud compute providers.
Last updated