Zaid Kahn
Founder & ChairmanMicrosoft · LinkedIn · Open Compute ProjectEx-Microsoft VP/GM of AI infrastructure. Chairs the Open Compute Project board.
§ 01 · Full-Stack AI Infrastructure
The foundation of AI.
Mantle designs, builds, and operates AI data centers across every layer of the stack. Land to inference, one company. GB200-class compute available now.
Tier III+ Pacific NW · Operating|Land to inference · Full stack|GB200 Available Now
§ 02 · Why Now
Mantle's compute is available now.
Mantle has compute running today, placed in existing facilities and ready to use. No interconnect queue, no megacampus timeline, no waitlist. We are also building our own liquid-cooled sites near enterprise demand, each online in 12 to 18 months, for the long term. Inference, not training. Capacity that arrives when your workload does, not years after.
§ 03 · Platform
Every layer of the stack, from land to inference.
Most AI infrastructure is assembled from a dozen vendors. Mantle owns the whole stack: the land, the power, the building, the hardware, and the software that runs on it.
Powered land, secured
01
Land
Mantle sources and secures powered land and shovel-ready sites near enterprise demand, the scarce input every AI build starts from.
Grid-aligned deployment
02
Power
We size each site to fit distribution-level grid opportunities, so we can often move faster than the large campuses that depend on major transmission interconnections.
Liquid-cooled by design
03
Structure
Mantle designs and builds its own data centers, engineered from the slab for direct-to-chip liquid cooling at the densities frontier GPUs demand.
NVIDIA NVL72 systems
04
Hardware
We deploy and operate NVIDIA NVL72 systems. GB200 is operating now, with GB300 NVL72 on the roadmap.
Orchestration and fleet
05
Software
Orchestration, fleet automation, and the inference platform that turns the hardware into capacity teams can request and run.
Five layers, one company. From the dirt to the inference endpoint, Mantle builds and runs all of it.
Two ways we deliver
Capacity now
Mantle racks operating today in a Tier III+ facility we designed and run, selling compute now while our owned sites come online.
Owned sites
Data centers Mantle builds and runs end to end, land to inference, for economics and control a leased footprint cannot match.
§ 04 · Compute
GB200-class inference compute, available now.
It is operating today, in a facility Mantle designed and runs. This is capacity you can request and run now, not a waitlist.
Platform
GB200 NVL72Operating now. GB300 NVL72 on the roadmap.
Cooling
Direct-to-chipLiquid-cooled by design
Network
400G fabricDual-bonded, redundant
Facility
Tier III+Pacific NW
The platform we run is the platform we ship into every site we build.
Mantle scales with AI.
Rack power density is climbing fast. Today's GPUs need 120 to 160 kW a rack; the next generation pushes toward 600. Because Mantle owns the land, the power, and the buildings, our sites are built to move up that curve, not age out at a fixed ceiling.
§ 05 · Team
Operators, not tourists.
Mantle is run by senior operators from Microsoft, Google, LinkedIn, Intel, Comcast, and AECOM. Roughly fifteen leaders deep, across land, power, infrastructure, hardware, cooling, networking, construction, and capital. The people who designed and ran data centers at this scale are the people running Mantle, team by team.
Leadership
Jordan Friedman
Chief Executive OfficerCommercial Development · Powered LandExperienced founder, commercial developer, and institutional fund manager. Expert in powered land and capital formation.
Infra Ops
Jacob Rose
Chief Development OfficerLinkedInEx-LinkedIn principal data center engineer. Designs and builds the facilities.
John Jason Brzozowski
President, GPUaaSMicrosoft · ComcastEx-Microsoft chief architect on the first NVIDIA GB200.
Hussain Moosajee
Chief Investment OfficerInstitutional Finance · Real EstateInstitutional finance and real estate. Structures the capital.
Sean Bondar
Chief of Business DevelopmentAECOM · AccentureEx-AECOM SVP for global tech. Hyperscale delivery and business development.
Software
Dima Pugachevich
SVP Software EngineeringLinkedInData center automation and Kubernetes at scale.
Alexander Kabanov
Principal EngineerLinkedIn · ApplePlatform engineering and GPU resource management.
Alexander Luberg
Principal EngineerLinkedInInfrastructure automation and DCIM at hyperscale.
Networking
Chris Tuska
Principal Network ArchitectImperium Data · Rocky Mountain IPv6Network automation and orchestration.
Hardware & Systems
Aaron Anderson
Principal Systems EngineerMicrosoft · IntelLiquid-cooled GPU systems. Ex-Microsoft GB200 systems.
Jason Chesser
SVP Systems EngineeringMicrosoft · IntelThermal and systems integration for HPC and GPU platforms.
Chris Gonzales
Principal Thermal Systems EngineerLenovo · IntelThermal engineering for next-generation HPC AI gear.
Structure
Karl Dzioba
Construction EngineeringSuffolk ConstructionMission-critical construction leader. Delivers the build.
Trevor Matthews
SVP EngineeringGoogle · LinkedInData center architecture and operations. Ex-Google and LinkedIn.
Land & Power
Michael Lavillotti
Head of Land & PowerEdged Energy · NYISOEnergy markets and grid leader. Sites the power and secures the interconnect.
§ 06 · For Customers
Looking for compute?
GB200 available now
What you operate day one
A complete platform, not raw GPUs. Mantle runs the infrastructure end to end, so your team ships models, not tickets.
A Mantle engineer on your workloads
Direct engineering support, not a ticket queue.
Storage included, not billed on top
High-performance all-flash storage carved out for your team, in the platform price.
Managed GPU Kubernetes · Dedicated all-flash storage · Private container registry · Hard tenant isolation · Live telemetry and dashboards.
SOC 2 Type II in progress. Backed by a 99.99% uptime SLA.
What you can run
Production inference
Multi-tenant or dedicated GB200-class capacity, liquid-cooled and available now, deployed in weeks, not quarters.
Batch and high-throughput
Reserved capacity for large-scale batch and high-throughput inference, on hardware Mantle deploys and operates itself.
Multi-year capacity reservations are available for teams running AI as core infrastructure, with terms set against real, operating availability.
§ 07 · For Investors
Mantle is raising.
Request access to our data room for the full materials. Send your information and we will open it.