The monthly briefing for engineers and operators who run serious compute workloads.
Ax3.Ai is rapidly expanding its global footprint with immediate and future-ready capacity across multiple regions. We provide white-glove support for infrastructure, hardware sourcing, and GPU instance sales to help you move fast.
| Site | Capacity | Location | Note |
|---|---|---|---|
| CENTAURI-60 | 60MW | Texas | Build-to-suit |
| SPICA-285 | 285MW → GW | North Dakota | IPP direct supply |
| ALTAIR-12 | ~12MW | N. America | Flowing today · ~5¢ |
| VEGA-60 | ~60MW (20MW now) | N. America | Nuclear-heavy, low-carbon |
| PROCYON-24 | 24MW (16MW flowing) | N. America | Transferable thru 2027 |
| ORION-100 | 100MW | N. America | Delivery 2028 · 100Gb fiber |
| POLARIS-200 | ~200MW | Texas | RFS Q3 2026 |
| ANDROMEDA-500 | 500MW (phased) | N. America | Phase 1 100MW · 2027 |
| RIGEL-24 | 24MW (16MW flowing) | N. America | 8MW headroom · 100Gb fiber |
| PERSEUS-20 | 20MW + gas exp. | N. America | GW-scale roadmap |
| CASSIOPEIA-60 | 60MW | USA | — |
Rentable NVIDIA Blackwell clusters across the US and Asia, available on 36-month terms. Reserve capacity now ahead of ready-for-service dates.
| Cluster | GPU | Qty | Rate | Location | RFS |
|---|---|---|---|---|---|
| TITAN | B300 | 2,948 | $4.2 | USA | 12 wks |
| NEBULA | B300 | 20,736 | Contact | Asia | Nov 26 |
| SUPERNOVA | GB300 | 20,736 | Contact | Asia | Q3 2026 |
| PULSAR | B300 | 10,368 | Contact | Asia | Jan 27 |
All terms 36 months · Full specs & pricing shared after we connect
Inquire NowCross-provider pricing snapshot, updated every Monday at 06:00 UTC from live provider dashboards. Spot prices reflect unreserved, on-demand availability in the US-East region.
| Configuration | Spot / hr |
|---|---|
| GB300 Blackwell | $4.40 STARTING |
| B300 Blackwell | $3.75 STARTING |
| H200 SXM5 | $2.50 STARTING |
| H100 SXM5 | $2.35 STARTING |
| MI3XX Series | $2.35 BEST |
Data sourced June 06 2026 · Global regions · Prices exclude egress and storage
See moreReal-time provisioning status across key Ax3 regions. Capacity is currently tight in US-East but remains stable across our emerging European and Asian hubs.
| GPU Configuration | Region | Status |
|---|---|---|
| GB300 / B300 / H200 | United States · USA | Available Now |
| H200 / H100 / MI3XX | Germany · EMEA Central 1 | Available Now |
| GB300 / B300 / H200 | Romania · EMEA Central 2 | Available Now |
| H200 / H100 | India · APAC South 2 | Available Now |
| B300 / H200 / H100 / MI3XX | Finland · EMEA North 2 | Available Now |
Datacenter Built to Suit — 60MW in Texas. Available to rent · Financing available.
See availabilityLambda demonstrated NVIDIA's upcoming co-packaged optics (CPO) networking switch, which integrates optical components directly into the switch package, eliminating traditional transceivers and significantly cutting power draw. The freed wattage can be redirected toward additional GPU compute — effectively delivering more tokens per watt.
Read the technical overviewCREZ transmission corridors, built in the early 2010s to move West Texas wind power across the ERCOT grid, are now a magnet for hyperscale AI infrastructure hungry for bulk electricity. AI developers prioritising power availability over geography are turning West Texas into a leading compute deployment region — a second life for infrastructure built for a completely different purpose.
Read the full storyDell and CoreWeave have completed the world's first deployment of NVIDIA's Vera Rubin GPU architecture, marking a new frontier in AI infrastructure capability. The deployment signals the beginning of a new compute generation with significant improvements in training throughput and memory bandwidth.
Read the full storyNebius projects its annualised revenue run rate will surge from $1.25 billion to $7–9 billion by end of 2026, anchored by major contracts with Meta and Microsoft. Its expanding software layer for AI development and inference — beyond pure hardware — is the key reason analysts see the stock as potentially undervalued even after recent gains.
Read the analysisSoftBank plans to invest up to €75 billion in French AI infrastructure, targeting up to 5 GW of data center capacity across three initial sites by 2031. The deal underscores a fundamental shift in site selection: electricity availability has become the primary deciding factor, overriding geography and tax incentives.
Read the full storyTrain Globally, Inference Locally. Ax3 provides cutting-edge, rack-scale and server-level AI solutions optimized for trillion-parameter LLMs and agentic AI workloads.
Liquid-cooled rack-scale system. 72× Rubin GPUs, 36× Vera CPUs, 75TB unified memory, 4× E1.S NVMe. Built for trillion-parameter training, large-scale inference, and agentic AI workloads.
Liquid-cooled rack-scale system. 2,592 Arm Neoverse V2 cores, up to 17.28TB LPDDR5X, 18.4 TB/s bandwidth. Optimised for AI, HPC, and data analytics at scale.
8× SXM GPUs via CX8 NVSwitch. 2× Intel Xeon 6 (up to 350W), 32× DDR5 DIMM (up to 4TB at 6400MHz), 8× 2.5" hot-swap NVMe + 2× M.2 boot drives.
8× NVIDIA Blackwell GPUs, 2× 6th-Gen Intel Xeon Scalable, 32× DDR5 6400MT/s DIMMs, 16× mixed SATA/NVMe storage options.
8× PCIe GPUs. Dual AMD EPYC 9005 or 2× 5th-Gen Intel Xeon with Neptune liquid cooling, 32× DDR5, 12× 2.5" U.2/U.3 NVMe SSDs.
8× AMD Instinct MI350 OAM accelerators. 2× AMD EPYC 9005-class CPUs, 24× DDR5 RDIMM (up to 6400MT/s), 8× U.2 2.5" SSDs.
8× AMD Instinct MI300X. 2× AMD EPYC (4th/5th Gen, up to 400W), 24× DDR5 DIMM (up to 3TB), 16× 2.5" hot-swap NVMe SSDs. LLM inference and cloud services.
Supports 8 FHFL + 2 double-width PCIe GPUs. 2× AMD EPYC 9005 Series, 32× DDR5 RDIMM (5200–4400MT/s), 16× PCIe Gen5×4 E1.S drives.
Ax3 helps organizations harness the full power of Computer Vision (CV) to automate operations across retail, manufacturing, energy, and healthcare. Our platform reduces deployment time and cost by up to 70%.
Ax3.Ai offers comprehensive consulting to OEMs, ODMs, and datacenter operators, helping them expand market reach and execute effective go-to-market strategies.
Accelerating the AI era with modular, high-density infrastructure built for speed and extreme efficiency. Precision-engineered factory builds eliminate construction uncertainty.
Scale AI Infrastructure Without Equity Dilution. Our coordinated global leasing model helps you avoid heavy CapEx and retain ownership of your company while deploying high-performance compute.
We provide the capital assets—covering hardware, shipping, and installation—allowing you to focus on growth and software development without giving up control.
✅ Industry Expertise: Years of experience in datacenter hardware, cloud infrastructure, and enterprise IT sales.
✅ Extensive Network: Access to top-tier customers, technology providers, and investors.
✅ Proven Results: Helping businesses accelerate revenue, establish partnerships, and drive market adoption.
Ax3.Ai is your trusted partner for breaking into new markets,connecting with key customers, and driving business growth.
