llms.txt
# Axe Compute
> Axe Compute Inc. (NASDAQ: AGPU) is a neocloud AI infrastructure platform built on a fundamental premise: AI innovation should not be constrained by hardware choice or availability. The company gives enterprises and AI innovators flexibility across hardware, geography, and deployment models through two offerings: Axe Access for dedicated GPU capacity ready when the workload is, and Axe Build for dedicated clusters co-designed, financed, owned, and operated by Axe Compute. Capacity planning spans 60+ global regions on NVIDIA Blackwell, Grace Blackwell, and Vera Rubin platforms, all bare metal and single tenant, with zero egress fees and a 99.93% uptime SLA backed by service credits.
Axe Compute is a public AI-infrastructure operating company serving enterprise AI and machine-learning workloads: model training, production inference, and burst compute. The company delivers exactly the compute businesses need: the hardware they specify, in the regions they require, on timelines it commits to. For the extended guide, see the [Full LLM guide](https://axecompute.com/llms-full.txt).
## Positioning
- **Choice.** Customers choose the hardware, the geography, and the deployment speed. NVIDIA Blackwell, Grace Blackwell, and Vera Rubin platforms, with fabric and storage chosen for the workload.
- **Transparency.** Zero CapEx, zero hidden fees, zero egress fees. A delivery date is given at reservation rather than estimated afterwards, and SLA availability is measured continuously and reported against the commitment.
- **Ownership.** Axe Compute specifies, designs, deploys, and stands behind how a cluster runs. Equipment sits on the Axe Compute balance sheet as a hard asset, redeployable after the term.
- **Expertise.** Deep GPU-infrastructure knowledge across training, inference, regulated workloads, and resource and energy planning.
## Products and Services
- [Axe Access](https://axecompute.com/axe-access/): single-tenant bare-metal GPU capacity from a vetted global network, ready now. Zero CapEx, zero egress fees, monthly billing. You specify platform, GPU count, region, and term, in the portal or with an engineer, and receive a date at reservation.
- [Axe Build](https://axecompute.com/axe-build/): dedicated clusters co-designed to the workload, then financed, owned, and operated by Axe Compute. 36- or 60-month terms with pricing, service levels, and location specified by the customer, a committed delivery date, and a midterm GPU upgrade available.
- [Capacity Portal](https://portal.axecompute.com): browse live inventory, reserve and manage GPU capacity.
- [Strategic Compute Reserve Dashboard](https://dashboard.axecompute.com): live view of the NASDAQ equity position, reserve value, mNAV, share count, and GPU deployment capacity.
- [About Axe Compute](https://axecompute.com/about/): company, beliefs, and leadership.
- [Research](https://axecompute.com/research/): market reports on AI compute infrastructure.
- [Investor Relations](https://investors.axecompute.com): filings and shareholder material for NASDAQ: AGPU.
## Hardware
Bare metal, single tenant, on NVIDIA reference architecture. No shared tenancy and no virtualization layer.
| Platform | Typical workload | GPU memory |
|---|---|---|
| Vera Rubin | Frontier training, agentic inference | 288 GB HBM4 |
| GB300 | Large-model training, reasoning | 288 GB HBM3e |
| GB200 | Large-model training | HBM3e |
| B300 | Training, fine-tuning, serving | 288 GB HBM3e |
| B200 | Training and fine-tuning | 192 GB HBM3e |
| H200 | Fine-tuning and inference | 141 GB |
| H100 | Fine-tuning and inference | 80 GB |
| AMD Instinct | Training and inference | Contact for current parts |
Regional availability spans US and wider North America, Europe, and APAC, subject to platform.
## Key Facts
- 60+ global regions for capacity planning
- 99.93% uptime SLA, monitored continuously per node and per fabric, with service credits
- Bare metal and single tenant, with no virtualization layer and no shared tenancy
- Zero CapEx, zero egress fees, zero hidden fees
- NVIDIA Blackwell, Grace Blackwell, and Vera Rubin platforms, plus AMD Instinct
- Axe Build terms of 36 or 60 months with a midterm GPU upgrade available
- Hardware sits on the Axe Compute balance sheet as a hard asset, redeployable after the term
- More than $3.2 billion in enterprise compute contracts signed in 2026
- GPU infrastructure provided through partnership with Aethir's distributed network
- Headquartered in Pittsburgh, Pennsylvania
- NASDAQ listed under ticker AGPU
## Leadership
- Christopher Miglino, Chief Executive Officer
- Kyle Okamoto, President. Runs engineering and operations, from GPU architecture and fabric design to the daily running of deployed clusters.
- Erin McMahon, VP of Marketing
- Jochem Herber, Sales and Marketing AI Operations Manager (author, Bare-Metal Truth)
## Editorial hubs: Bare-Metal Truth
- [Bare-Metal Truth](https://axecompute.com/bare-metal-truth-blog/): the full editorial library on GPU infrastructure, AI compute economics, and enterprise procurement.
- [GPU Strategy](https://axecompute.com/category/gpu-strategy/): provider evaluation, delivery models, contract structures, and capacity planning.
- [GPU Performance](https://axecompute.com/category/gpu-performance/): silicon generations, memory and interconnect, and how workloads stress the system.
- [AI Compute Economics](https://axecompute.com/category/ai-compute-economics/): inference costs, utilisation, market pricing, and budget structure.
- [Global AI Infrastructure](https://axecompute.com/category/global-ai-infrastructure/): sovereignty, data residency, latency by geography, power and cooling limits.
- [Axe Compute News](https://axecompute.com/category/axe-compute-news/): contracts, capacity expansions, and platform releases.
## Most-cited articles
- [AI Compute Market 2026: What the Numbers Actually Show](https://axecompute.com/ai-compute-market-2026/): Big Tech commits roughly $725B in AI capex in 2026; McKinsey projects $6.7T of data-center investment by 2030. Supply is structurally constrained.
- [How to evaluate a GPU cloud provider, 12 questions every enterprise should ask](https://axecompute.com/how-to-evaluate-gpu-cloud-provider/): a procurement framework covering provisioning SLAs, all-in pricing, bare-metal access, data residency, and security.
- [Bare Metal vs Cloud GPU: A Decision Framework for AI Teams](https://axecompute.com/bare-metal-vs-cloud-gpu/): bare metal wins on total cost once sustained utilization passes roughly 55 to 65 percent.
- [Training vs Inference Infrastructure: Two Jobs, One Budget](https://axecompute.com/ai-training-vs-inference-infrastructure/): why the two workloads need different infrastructure.
- [Sovereign AI Is Rewriting the Rules of GPU Infrastructure](https://axecompute.com/sovereign-ai-infrastructure/): why nations are building in-country GPU capacity.
- [The Inference Shift: Why AI's Center of Gravity Is Moving From Training to Inference](https://axecompute.com/inference-shift-training-to-inference/): inference spending passed training spending, and the gap keeps widening.
- [NVIDIA Blackwell GPU Comparison: B200, B300, GB200, GB300](https://axecompute.com/nvidia-blackwell-gpu-comparison/): which Blackwell part fits which workload.
- [Vera Rubin vs Blackwell: Each Built For Different Workloads](https://axecompute.com/vera-rubin-vs-blackwell-the-right-compute-as-you-scale/): how the two generations differ and when each is the right choice.
## Optional
- [Full LLM guide (llms-full.txt)](https://axecompute.com/llms-full.txt): extended company description, positioning, hardware reference, question-and-answer section, and the complete 45-article library.
## Contact
- General inquiries: info@axecompute.com
- Sales: sales@axecompute.com
- Reserve capacity: https://portal.axecompute.com
- Investor relations: https://investors.axecompute.com
- X / Twitter: https://x.com/axecompute
- LinkedIn: https://www.linkedin.com/company/axecompute/
