We aim to remove the barriers between your team and the compute you need—so you can focus on building.
Axe Compute (NASDAQ: AGPU) is introduced today as a new corporate identity and enterprise compute platform initiative, as Predictive Oncology expands into enterprise GPU infrastructure. Here’s why we built Axe Compute—and what it means for teams building AI at scale.
GPU-as-a-Service Is Broken
If you’re building AI at scale, you already know the pain. You need GPUs—yesterday—but procurement timelines have stretched to 40–52 weeks for high-end hardware. Hyperscaler waitlists average 12 weeks. When you finally get capacity, you’re often stuck with whatever hardware and region are available—not what your workload actually needs.
Then come the other problems: complex pricing with surprise egress fees, virtualization overhead eating into performance, and rigid contracts that lock you into a single vendor’s ecosystem. Most providers operate in just 3–4 countries, forcing you to move data to their infrastructure rather than bringing compute to where your data and users are.
The current state of enterprise GPU infrastructure forces teams to spend more time managing compute than building products. That’s backwards. Your engineers should be training models and shipping features—not fighting for capacity, deciphering invoices, or compromising on hardware because it’s all that’s available.
We Built Axe Compute to Fix This
Axe Compute, an enterprise-focused GPU infrastructure platform leveraging Aethir’s distributed network, is designed to remove the obstacles standing between your team and the compute you need. Our platform is designed to support provisioning of dedicated GPU clusters in as little as 24–48 hours, rather than weeks or months. We aim to offer transparent flat-rate pricing with no egress fees, bare-metal GPU configurations designed to eliminate typical virtualization overhead, and real choice—in hardware, configuration, and where workloads run.
Unmatched Choice: Hardware, Configuration, Location
The hardware you need
We plan to offer access to a broad range of GPU types through our network relationships, spanning multiple tiers from B300-class GPUs to H200s and beyond, with diverse cluster configurations and performance profiles. Whether you’re running large-scale training, real-time inference, or research workloads, you can select compute aligned to your project’s needs, rather than being forced into a one-size-fits-all offering.
The configuration you want
Beyond GPU selection, we provide flexible fabric and bandwidth options. Choose interconnects, topology, and network architecture designed for distributed training or inference workloads. Configurations are designed to optimize for price, performance, or both, without forcing teams into rigid templates.
The location your data demands
This is where we’re fundamentally different. While many competitors operate in just a handful of countries, Axe Compute enables access to distributed capacity across 200+ global locations through the Aethir network. That means workloads can be placed closer to users and data, helping address latency, data residency, and compliance requirements—without requiring data to be centralized in a single cloud region. We aim to bring compute to your data, not the other way around.
Economics That Actually Work
Choice means little if the pricing doesn’t make sense. Our model is designed to reduce total cost of ownership compared to traditional hyperscaler pricing, in some cases by up to 60%, based on illustrative comparisons to publicly available on-demand cloud pricing.
Illustrative example: H200 GPUs priced at approximately $1.35 per hour compared to $6.85 per hour on AWS, based on publicly listed rates at the time of writing. For an 8-GPU cluster running continuously, that equates to approximately $94,608 annually versus $480,048, representing potential savings of $385,440. Actual pricing and savings will vary based on configuration, region, duration, and availability.
Pricing is intended to be flat and transparent, with no bandwidth or egress fees. Workloads are designed to remain portable, with flexibility to scale up or down based on business needs, without long-term lock-ins or forced architectures.
A New Name, An Expanded Mission
Today we’re announcing that Predictive Oncology (NASDAQ: POAI) has rebranded as Axe Compute, with our common stock now trading under the ticker AGPU. This represents an expansion of the company’s mission.
The logic is simple: without infrastructure, innovation cannot occur. Transformers required compute to train them. Scaling laws require hardware to test them. Production AI requires capacity to run it. Global enterprise spending on AI cloud services is projected to exceed $400 billion in 2025, with demand continuing to outpace supply. Axe Compute is positioning itself to help address this infrastructure gap.
Predictive Oncology continues as a subsidiary focused on AI-driven drug discovery. Axe Compute operates as the parent company, bringing the governance, transparency, and accountability of a publicly traded, SEC-regulated entity to the GPU infrastructure market.
How We Deliver
Our offering leverages Aethir’s distributed GPU network, which includes over 435,000 GPU containers across 93 countries. As an enterprise-facing provider leveraging this network, Axe Compute coordinates distributed capacity into an enterprise-grade service layer designed to support availability targets, service-level agreements, dedicated support, and traditional commercial contracting structures.
We intend to operate as an active infrastructure company rather than a passive treasury. Our model is designed to acquire access to infrastructure capacity and deploy it to serve enterprise clients under service agreements, with revenue intended to be generated from the margin between infrastructure acquisition costs and enterprise billing rates. There can be no assurance that these economics will be realized as anticipated.
What This Means for You
If you’re building AI, you shouldn’t have to become an expert in GPU procurement, cloud pricing models, or infrastructure management. You shouldn’t have to compromise on hardware because it’s all that’s available, or move your data halfway around the world because your provider operates in only a few regions.
You should be able to access the compute you need—the specific GPU, the right configuration, in the region that makes sense—and then get back to the work that actually matters: building and delivering value to your customers.
That’s what Axe Compute is designed to offer: broad choice in hardware and configuration, global reach through a distributed network, transparent pricing, and enterprise-grade reliability.
Get Started
Ready to explore what GPU infrastructure can look like? Reserve compute, calculate potential savings, or talk to our team about your specific requirements.
— — —
NASDAQ: AGPU