You get the full GPU, the full memory, and a cost structure you can model
In the demanding world of professional media production, where a missed deadline is never merely a technical inconvenience, 1Legion has introduced dedicated bare-metal servers built around NVIDIA's RTX Pro 6000 Blackwell GPUs — offering studios and broadcast operators something rarer than raw compute power: certainty. The announcement reflects a broader tension in cloud infrastructure, where shared environments and variable pricing have long sat uneasily alongside the unforgiving schedules of film, television, and visual effects work. By removing shared tenancy and hidden costs from the equation, 1Legion is wagering that predictability itself is a form of performance.
- Production teams have long faced a quiet crisis: render farms running out of GPU memory mid-project, transcoding pipelines stalling on oversubscribed hardware, and technical failures cascading into missed delivery dates.
- 1Legion is responding with eight NVIDIA RTX Pro 6000 Blackwell GPUs per server node — 768GB of total GPU memory, no virtualization, no shared tenancy — giving studios exclusive access to the full machine they rent.
- Pricing is fixed at $1.34 per GPU per hour on 12-to-24-month terms, with no egress fees and no variable performance pricing, directly targeting the budget unpredictability that haunts production teams in hyperscaler environments.
- The infrastructure is available now, positioning 1Legion as a purpose-built alternative for studios and broadcast operators who need hardware that won't be oversubscribed at the moment it matters most.
1Legion has launched dedicated bare-metal servers built around NVIDIA's RTX Pro 6000 Blackwell GPUs, targeting the specific pressures of professional media and visual effects production. The offering is a direct response to a familiar problem: when shared cloud infrastructure runs out of video memory mid-render, or a transcoding pipeline chokes because hardware is split among multiple customers, the technical failure becomes a scheduling crisis and a missed delivery date.
Each server node houses eight RTX Pro 6000 Blackwell Max-Q GPUs, delivering 768 gigabytes of total GPU memory — 96GB per card — alongside 24,064 CUDA cores per GPU and the latest RT and Tensor cores. That memory headroom matters for the software these machines are designed to run. Blender Cycles, V-Ray, Arnold, and Redshift regularly encounter scenes too large for conventional server GPUs, and the Pro 6000 accommodates dense volumetric data, high-polygon models, and 4K texture sets without forcing artists to segment scenes or fall back on slower processing methods. For broadcast operations, multiple NVENC and NVDEC engines handle real-time transcoding and low-latency streaming across the full node.
The infrastructure is bare metal throughout — no virtualization layer, no partitioned instances, no egress fees. Pricing starts at $1.34 per GPU per hour on 12-to-24-month contracts, and 1Legion emphasizes that this is the actual cost, fixed regardless of platform demand. CEO David Vargas-Racero put it plainly: media and VFX teams don't have unpredictable workloads, they have deadlines. The servers are available now for studios and broadcast operators ready to provision for their next project.
1Legion has rolled out a new infrastructure option built specifically for the people who make movies, television, and visual effects: dedicated servers packed with NVIDIA's latest RTX Pro 6000 Blackwell GPUs, eight cards per machine, no sharing, no surprises.
The company positions this as a direct answer to a recurring problem in production environments. When a render farm runs out of video memory mid-project, or a transcoding pipeline chokes because the underlying hardware is being split among multiple customers, the delays cascade. A technical constraint becomes a scheduling crisis becomes a missed delivery date. 1Legion's approach is to remove the constraint entirely by giving production teams exclusive access to the hardware they rent, with transparent pricing that doesn't shift based on demand or hidden fees.
Each server node houses eight RTX Pro 6000 Blackwell Max-Q GPUs. That translates to 768 gigabytes of total GPU memory—96 gigabytes per card—along with 24,064 CUDA cores per GPU and the latest generation of RT and Tensor cores. The memory capacity matters enormously for the work these machines are meant to handle. Professional rendering software like Blender Cycles, V-Ray, Arnold, and Redshift regularly encounters scenes that exceed what conventional server GPUs can hold in memory. Large volumetric data, dense particle systems, high-polygon models, and 4K texture sets all demand space. The Pro 6000 accommodates them without forcing artists to segment scenes or resort to slower out-of-core processing.
For broadcast and media technology operations, the GPU includes multiple encoding and decoding engines—NVENC and NVDEC—that handle real-time transcoding, playout, and low-latency streaming across the full eight-GPU node. The infrastructure itself is bare metal, meaning no virtualization layer, no shared tenancy, no partitioned instances. Customers get the full machine. There are no egress fees either, a detail that matters when moving large files in and out of production pipelines.
1Legion prices the service starting at $1.34 per GPU per hour on 12-month and 24-month contracts. The company emphasizes that this is the actual cost—no variable performance pricing, no storage surprises, nothing that shifts based on how busy the infrastructure is on any given day. For production teams accustomed to hyperscaler environments where costs can balloon unexpectedly, the predictability is the feature.
David Vargas-Racero, 1Legion's CEO, framed the offering as a response to the specific nature of media work. "Media and VFX teams don't have unpredictable workloads, they have deadlines," he said. The infrastructure is designed around that reality. When a render hits memory limits or a transcoding pipeline stalls due to contention on shared hardware, the problem isn't just technical—it's financial and logistical. Dedicated bare metal eliminates both the bottleneck and the uncertainty that comes with it. You know what you're paying, you know what you're getting, and you know it won't be oversubscribed when you need it most.
The servers are available now through 1Legion's platform, ready for studios and broadcast operators to provision for their next project.
Notable Quotes
When a render farm hits VRAM limits mid-production or a transcoding pipeline stalls because of shared infrastructure contention, that's a billing problem that compounds into a delivery problem.— David Vargas-Racero, CEO of 1Legion
The Hearth Conversation Another angle on the story
Why does a VFX studio care whether the GPU is "dedicated" versus shared? Isn't it just compute?
It's about what happens when you run out of space. A shared GPU might have 96GB total, but if you're one of five customers, you're fighting for maybe 20GB. A scene that needs 80GB of memory either doesn't run, or it runs slowly because the software has to swap data in and out. On a dedicated machine, all 768GB is yours. No contention, no slowdown mid-render.
But couldn't a studio just buy their own hardware and own it outright?
They could, but then they're managing the machines themselves, dealing with maintenance, upgrades, power, cooling, and they're paying for capacity they might not use every day. Renting dedicated infrastructure means you scale up and down with your projects without the capital cost or the operational burden.
The pricing is $1.34 per GPU per hour. That sounds cheap, but what's the actual monthly bill?
Eight GPUs running 24/7 for a month is roughly $7,900. For a studio running a major project, that's a known cost they can budget before production starts. The point is there are no surprises—no egress fees, no performance degradation because someone else's job is running on the same hardware.
What makes the RTX Pro 6000 Blackwell different from a consumer GPU?
The memory is the biggest difference. Consumer cards max out at 24GB. The Pro 6000 has 96GB per card. It's also built for professional software—Blender, V-Ray, Arnold—and it includes encoding and decoding engines for broadcast work. It's purpose-built for the workloads that media companies actually run.
Is this a response to cloud providers like AWS or Google Cloud?
It's a response to the limitations of those environments. Hyperscalers optimize for scale and flexibility, which means shared infrastructure and variable pricing. 1Legion is saying: if you know your workload and you need predictability, there's another option. Dedicated hardware, transparent costs, no hidden fees.