Skorppio Challenges Cloud with On-Premise HPC Rentals

LOS ANGELES, CA – February 23, 2026 – A new Los Angeles-based startup, Skorppio, launched a self-serve rental platform today, aiming to reshape how industries access high-performance computing (HPC). The company offers on-premise, enterprise-grade servers and workstations, including systems with the latest NVIDIA and AMD processors, on flexible, short-term contracts. The service directly targets AI startups, visual effects (VFX) studios, and research organizations struggling with the high costs, data control issues, and vendor lock-in associated with public cloud services.

Skorppio enters a market where the demand for computational power is insatiable, driven by the explosive growth of artificial intelligence and the ever-increasing complexity of digital content creation. By providing a physical, on-site alternative to cloud infrastructure, the company is betting on a growing trend of on-premise resurgence, where control and cost predictability are paramount.

The On-Premise Comeback

For years, the public cloud has been the default solution for scalable computing. However, as workloads mature and run continuously, the usage-based pricing models of hyperscalers like AWS, Azure, and Google Cloud can become prohibitively expensive. Industry reports indicate that global cloud operating costs are projected to rise by approximately 19% in 2025 alone, driven largely by AI services. This financial pressure, combined with unpredictable data transfer fees—often called egress fees—has led many organizations to re-evaluate their cloud-first strategies.

Skorppio's model is designed to counter these pain points. It offers what it calls “bare-metal” performance, meaning users get direct, unshared access to the hardware, eliminating the performance variability that can occur in multi-tenant cloud environments. This is particularly crucial for latency-sensitive or I/O-heavy HPC workloads where network bottlenecks can starve powerful GPUs of data, leading to underutilized and costly resources.

"Access to performance compute has gotten expensive enough that teams end up buying what they can afford or what's available quickly, instead of what the work demands," said Founder and CEO Jonathan Goldstein in the company's launch announcement. "Skorppio delivers dedicated, workflow-built systems with bare-metal performance, on ultra-flexible terms, without the cost of ownership."

This approach also addresses critical concerns around data sovereignty and security. For industries handling sensitive intellectual property, such as AI model development or pre-release film assets, keeping data within their own physical perimeter is a significant advantage. Market analysis supports this shift, showing that on-premise deployments are expected to retain a majority market share in the HPC space, with many organizations adopting a hybrid or "Cloud+" model that strategically repatriates key workloads from the public cloud.

Democratizing Access to Powerhouse Hardware

At the core of Skorppio’s offering is access to cutting-edge hardware that is often difficult and expensive to procure. Through a partnership with PNY Pro, the company's fleet includes systems powered by high-end AMD processors like the EPYC and Threadripper PRO, as well as NVIDIA GPUs. The press release highlights the inclusion of the NVIDIA Blackwell GPU architecture, the company's next-generation platform for AI.

While the Blackwell data center GPUs like the B100 and GB200 are just beginning to roll out in late 2024 and early 2025, their inclusion in a rental catalog would place Skorppio at the forefront of hardware accessibility. The announcement also specifies the "RTX 6000 PRO," which likely refers to the powerful NVIDIA RTX 6000 Ada Generation workstation card. Based on the Ada Lovelace architecture, this GPU is a workhorse for professional visualization and local AI development, offering 48GB of VRAM per card. Skorppio's claim of a flagship workstation with up to 786 GB of VRAM suggests configurations with multiple interconnected GPUs, a setup that would traditionally require massive capital investment.

By renting these systems, Skorppio effectively removes the primary barrier to entry for smaller companies and research teams: a high upfront cost. A single professional-grade GPU can cost tens of thousands of dollars, making multi-GPU systems a non-starter for many. This rental model could significantly accelerate innovation by allowing more developers and creators to experiment with and build on the latest hardware, a trend reflected in the broader GPU rental market, which is projected to grow tenfold to over $30 billion by 2032.

From Box to Bare Metal: Simplifying Deployment

Beyond just providing hardware, Skorppio aims to simplify the entire deployment process through a digital-first platform and curated hardware packages. The company offers pre-configured 'KIT Collections' tailored for specific tasks, such as 'AI Startup Dev Kits' and 'VFX Render Farm' configurations. These kits are validated by domain experts, including PhD-level AI researchers and VFX technical directors, ensuring they are optimized for real-world workflows out of the box. The packages also include necessary infrastructure components like high-speed interconnects and network switches to accelerate on-site setup.

A key technical selling point is the claim that its flagship workstation can run on standard electrical circuits, eliminating the need for specialized, high-amperage power infrastructure typical of data centers. While a multi-GPU system delivering nearly a terabyte of VRAM would still present significant power and cooling challenges, the ability to deploy powerful workstations in a conventional office environment without an electrician is a compelling proposition for teams that lack dedicated server rooms.

This focus on practical deployment is rooted in the experience of founder Jonathan Goldstein. His background includes founding a creative technology agency that built large-scale technical installations for Fortune 500 companies and global music artists. This hands-on experience with complex, on-premise technology projects provides a credible foundation for Skorppio's ambitious logistics and support model, which promises nationwide delivery and setup.

A New Investment Model: Trading Compute for Equity

Looking beyond its launch, Skorppio has announced plans to introduce an innovative program that further lowers the barrier to entry for the most resource-constrained innovators. Later this year, the company intends to roll out an equity-for-compute model, providing high-performance computing resources to select early-stage AI startups in exchange for an equity stake.

This model, reminiscent of startup accelerators like Y Combinator, aligns Skorppio's success directly with that of its startup clients. For a pre-seed or seed-stage AI company, the ability to trade a small percentage of ownership for the six- or seven-figure compute resources needed to train a foundational model could be a game-changer. It preserves precious cash reserves for hiring talent and other operational expenses, while providing access to the very tools needed to build their core product.

While this model carries inherent risks for both sides—startups dilute ownership, and Skorppio takes on the high risk of early-stage investing—it positions the company as more than just a hardware rental service. It becomes a strategic partner and enabler in the burgeoning AI ecosystem, betting on the future success of the companies it helps empower. This move signals a deep understanding of the unique challenges facing the next generation of technology companies and a bold strategy to grow alongside them.