The GPU Hypervisor

The First Hypervisor for
GPU/CPU AI Infrastructure.

The first hypervisor for GPU/CPU AI infrastructure. Break vendor lock-in. Run any workload on any hardware — faster, cheaper, without a single line of code changed.

0
Faster than native CUDA.
Not an emulation layer. A performance multiplier — proven on real hardware, real workloads.

Savings that change
the economics of AI

ABD doesn't just free your infrastructure. It transforms your P&L.

4.4x

Cost Efficiency

$0.23 per million tokens vs. $1.02 on NVIDIA. When you're processing billions of tokens, this isn't savings — it's survival.

90%+

Workloads Locked In

Over 90% of AI workloads depend on CUDA. That's not an ecosystem — it's a dependency. ABD breaks it without breaking your code.

$250B+

Addressable Market

AI compatibility, enablement, and hardware extraction infrastructure. The market isn't emerging — it's overdue.

Born from a simple
but urgent realization

The future of AI was becoming too constrained by the infrastructure beneath it. That realization emerged through the work of Alex Veytsman — former CTO and Chief AI Officer of Dell Technologies — who has designed and built more than 100 data centers across the globe.

Over years of advising enterprises, data center developers, and hardware designers, a consistent concern surfaced again and again: AI infrastructure was becoming increasingly concentrated, rigid, and dependent on narrow technology paths.

"How do we preserve choice in AI infrastructure? How do we diversify compute without disrupting existing systems? How do we scale AI sustainably?"

These weren't theoretical questions. They were operational, economic, and strategic — shared by nearly every stakeholder shaping the next generation of AI data centers.

The founding team assembled leaders with decades of experience — a common conviction that AI's long-term success depends on infrastructure that is open, resilient, and capable of evolving beyond single-vendor dependence.

DellAppleVerizonPixar Georgia TechNTT DataeBayPayPal IntelHPECiscoCoca-Cola Accenture
Data Center
Innovation

The hypervisor layer for
hybrid AI data centers

Three steps. Zero code changes. Full hardware freedom.

01

Intercept

ABD sits transparently below the CUDA API layer, intercepting workloads at runtime. No rewrites. No recompilation.

Zero Code Changes
02

Translate & Execute

Workloads run on AMD, Intel, ARM, and custom ASICs while preserving full compatibility and operational continuity.

Hardware-Accelerated
03

Operate at Scale

A unified Control Plane with Runtime Nodes and APIs supports orchestration, scheduling, telemetry, and automation.

Enterprise-Ready

Interoperability without compromise

ABD Fabric enables Hybrid AI Infrastructure so data centers can deploy alternative GPU systems while maintaining compatibility with existing AI software.

  • Control Plane for Orchestration, Runtime emulation, Workflow scheduling
  • Support for AMD, ARM and Customer ASICs
  • 6+ provisional patents covering 19 inventions
Data Center Infrastructure

Faster than native.
Not just compatible.

NVIDIA H100 (Native CUDA)1.0x
AMD MI300X + ABD Fabric2.19x
NVIDIA H100 — Cost / M tokens$1.02
AMD MI300X + ABD — Cost / M tokens$0.23

Not an emulation layer.
A performance multiplier.

ABD Fabric doesn't just translate — it optimizes. Hardware-accelerated semantic translation delivers throughput that exceeds native CUDA performance on equivalent hardware.

+3.6%
SGEMM vs native ROCm
+11%
LayerNorm improvement
+3.6%
FFN acceleration
48
GPUs per ART-1 device

See what freedom saves you

Estimate your savings when you break free from single-vendor lock-in.

10 MW
1,000 GPUs
3 years

// Estimated Savings

Hardware & Compute$135M
Power & Cooling$15M
Time-to-Revenue$28M
Total Estimated Savings$178M

The operators scaling AI

🏢

Data Center Operators

Validate and operate NVIDIA-alternative infrastructure before and after capital is committed. De-risk billion-dollar hardware decisions.

☁️

NeoCloud Providers

Offer hybrid AI environments that expand hardware choice while preserving complete CUDA compatibility for your customers.

⚙️

OEM System Builders

Deliver future-proof AI infrastructure aligned with multi-vendor procurement. Factory-install ABD Fabric as a differentiator.

Built by operators,
not observers

Combined experience includes exits to Arctic Wolf, leadership at Apple, Pixar, Dell, Cisco — and 17 U.S. patents.

Greg Fitzgerald

Greg Fitzgerald

Founder & CEO
Serial exit (Sevco → Arctic Wolf). Cylance, BMC Software.
Jason Green

Jason Green

Founder & President
Enterprise API. John Lewis, Marvel, UFC, Webflow.
Alex Veytsman

Alex Veytsman

Founder & CTO
Former Dell CTO. HPE, Cisco, eBay, PayPal. 100+ data centers built.
Jack Finegold

Jack Finegold

Founder & Chief of Staff
Fintech + cybersecurity serial entrepreneur.
Chief Revenue Officer

Coming Soon

Chief Revenue Officer
Megan Dahlgren

Megan Dahlgren

CMO
Global VP at NTT Corporation.
Aritomo Shinozaki

Aritomo Shinozaki

Chief Product Officer
Physics PhD, Cornell MBA.
Mark Saliman

Mark Saliman

CFO
30+ years. Chiron/Novartis $5.1B acquisition.
Dr. Kyriakos Vamvoudakis

Dr. Kyriakos Vamvoudakis

Chief AI & Data Security
Endowed Professor, Georgia Tech.
VP of Sales

Coming Soon

VP of Sales
David Shefter

David Shefter

VP Alliances
Citi, Verizon, IBM. $1.8B revenue growth, $250M savings via AI & digital innovation.
Dr. Min-Yi Shih

Dr. Min-Yi Shih

Hardware Advisor
PhD EE, UCLA MBA. 17 U.S. patents. Defense + energy.
Strategic Partners & Early Customers
GigaIO Sanity Solutions Hivelocity Datum Dev Corp

Request the deck

Get access to ABD's investor materials, technical white paper, and market analysis. Controlled distribution — all requests reviewed by the founding team.

I'm interested as...
Materials sent directly by the ABD team within 24 hours.
✓ Request received. The ABD team will follow up shortly.
Coming Soon

Insights & Research

Deep dives into AI infrastructure economics, GPU market dynamics, and the technical architecture behind ABD Fabric. Thought leadership from the team that's lived it.

We're Hiring

Join the Team

We're building the foundation for hybrid AI. If you've got deep domain experience in GPU architecture, data center operations, or AI infrastructure — we want to talk.

See Open Positions →

Ready to break free?

Join the operators building the next era of AI infrastructure — without lock-in, without limits.