Compute, concentrated.
US adjacent reserved GPU clusters.
Reference architecture sized, rentable by the node or the half site. Built for US teams and US neocloud partners who need Central US latency and do not need US soil.
The offer
Reserved GPU clusters and GPU hours on dedicated HGX B300 topology. Kubernetes and Slurm on bare metal, Spectrum X or InfiniBand inside the POP, private interconnect toward Dallas.
512 GPUs is one scalable unit of the NVIDIA B300 reference architecture (64 nodes). Splits cleanly as 256 + 256 or 128 × 4.
Six ways to slice one fabric
From a single reserved node to the full dedicated site. Pick a size here and it carries into the request form.
| SKU | GPUs | Nodes | Typical use | |
|---|---|---|---|---|
| 8 | 1 | Single node reserved | select | |
| 16 | 2 | Small reserved | select | |
| 64 | 8 | Mid training / serving | select | |
| 128 | 16 | Serious reserved pod | selected | |
| 256 | 32 | Large enterprise or wholesale block | select | |
| 512 | 64 | Full site dedicated | select |
Serious reserved pod
Enterprise reserved from 12 months. Wholesale and capacity partner blocks run 24 to 36 months take or pay.
Why Monterrey, for a US buyer
Nearshore power and space with a private path into the Texas interconnection hubs. What that buys you, need by need.
Latency to US ops
Texas adjacent nearshore: typically sub 20 ms class RTT toward the Dallas and Houston corridors.
Capacity without quota fights
Dedicated HGX B300 topology with reserved delivery. No hyperscaler quota queue, no waitlist.
Clean cluster math
512 GPUs, one reference architecture scalable unit (64 nodes). Splits as 256 + 256 or 128 × 4, so your topology design survives contact with procurement.
Power without the queue
Power and space without a US interconnection queue. Phase 1 goes live October 2027.
Interconnect
Cross border fiber pathways and a private link story into Texas interconnection hubs, operated by our interconnect partner on a carrier backbone. Six terrestrial border crossings into the US.
Commercial posture
Reserved clusters, not marketplace best effort hosts. USD contracts under US law, named support.
Honest framing. This is Nearshore Americas, not a US data residency region. Residency is Mexico and we disclose it clearly. Ideal when your policy allows Mexico hosted compute and you want Texas class latency with reserved B300 capacity.
One standard cluster, replicated
Monterrey is first. The same B300 cluster standard is planned for the sites below. Pipeline, not promise: only Monterrey is open for contracts today.
Texas corridor. Sub 20 ms to Dallas and Houston.
West coast corridor toward San Diego and Los Angeles.
Central Mexico data center cluster.
Southeast edge toward Florida and the Caribbean.
For workloads that require US data residency.
Nearshore capacity for the isthmus and the Caribbean basin.
Long term capacity on low cost power markets.
Who it is for, and who it is not
We would rather lose a deal on the first call than on the compliance review.
Strong fit
- ✓US startups, AI labs and mid-market tech with no hard US soil mandate
- ✓US neoclouds and aggregators needing wholesale or white label overflow capacity
- ✓Nearshore manufacturing and US to Mexico industrial AI workloads already spanning the border
Weaker fit, we will say so upfront
- ✕Deals that require US data residency, FedRAMP, or strict healthcare / government US only hosting
- →Those belong on a future US POP, or a split design: 256 in Monterrey + 256 in the US
Two ways to buy the same capacity
One ledger, two contracts. Enterprise wins on SLA and support attach; wholesale wins on term and volume.
Enterprise Reserved Cluster
Your brand facing reserved pods, run the way a US team expects.
- →USD contracts under US law, named support
- →Kubernetes and Slurm, high performance storage, egress friendly posture
- →Private cross connect toward Texas
- →Reserved terms from 12 to 36 months
Capacity Partner Agreement
Blocks from 128 GPUs up to the full 512, behind your brand.
- →White label and “powered by” friendly, non exclusive by default
- →Backline L2/L3, API and ops handoff, VRF/VLAN and BGP friendly interconnect
- →Longer take or pay (24 to 36 months), volume pricing
- →Co-exist behind your brand; we stay in the footer
What you get on day one
The bar every cluster clears before we hand over credentials.
Compute
NVIDIA B300 HGX. NVLink domain of 8 GPUs per node. Rail aligned scale out fabric across the POP.
Network
Spectrum X or InfiniBand, ~800 Gb/s class inside the POP. Private interconnect toward Dallas via private link.
Orchestration
Kubernetes and Slurm on bare metal. Your images, your schedulers, your topology awareness.
Storage
Local NVMe plus parallel and AI object tiers sized to the cluster.
Compliance
Colocation operator facility certifications; dense² SOC 2 program in progress. Residency: Mexico, disclosed in every document you sign.
Commercial
Reserved clusters and GPU hours only, 12 to 36 month terms, billed in USD under US law. Not a marketplace, not best effort hosts, and we do not sell tokens.
Where we sit against the alternatives
Egress is not a tax. A fixed topology you can design against, delivered on a date.
Nearshore cost and power, non exclusive blocks, and we co-exist behind your brand.
Fixed topology, signed SLA, named support. Not spotty hosts with a ratings page.
Tell us your target. We return a delivery window and a term sheet.
Size, buyer type and how you want to land in Texas. That is all we need for a first answer.
GPU hours and reserved clusters only. We do not sell tokens.