NOL8 proves technical and economic superiority across three continents
FortifAI (ASX: FTI) has completed its most comprehensive testing program yet on the NOL8 AI Data Plane, proving version V1.0 across three continents on Megaport’s global network. The standout outcome was 10X more governed data per dollar of infrastructure compared to legacy software running on CPUs.
The live testing was conducted by NOL8, FortifAI’s subsidiary, over Megaport’s private network. Workloads traversed Virginia (US), Tokyo and Sydney, delivering agentic AI data workflows at near-zero latency for enterprises running Private AI.
Across every test, performance met or exceeded expectations. NOL8’s engine recorded in-region latency of 3 milliseconds, with distance the only variable across sites due to the speed of light.
This represents applied, real-world proof extending earlier benchmark results released to the ASX on 1 April 2026. For investors, it marks a validation milestone from lab benchmark to live multi-continent deployment.
When big ASX news breaks, our subscribers know first
The headline numbers: performance meets or exceeds expectations
The testing compared NOL8 V1.0 directly against FortifAI’s own benchmark legacy software implementation. Importantly, per the source, these comparisons refer to the Company’s own benchmark software, not any third-party product.
NOL8 governed data at a rate of 34 terabytes/day, while the legacy benchmark software reached only 2.6 terabytes/day, delivering the 10X+ advantage. Throughput declined by less than 1% even as governance policies grew by 400%.
Every figure reported is a floor, not a limit. NOL8’s ceiling was never reached during testing, and the Company notes the advantage widens as workloads grow larger. Every output was checked against an independent reference.
| Metric | NOL8 V1.0 | Legacy Benchmark Software | Advantage |
|---|---|---|---|
| Governed data throughput | 34 TB/day | 2.6 TB/day | 10X+ |
| Data governance accuracy | 100% (byte-for-byte verified) | n/a | n/a |
| In-region engine latency | 3 milliseconds | n/a | n/a |
| Throughput decline as policies grew 400% | Less than 1% | n/a | n/a |
According to the Company, matching NOL8 in software would require exponentially more CPUs, with the associated power draw, and still not match NOL8’s latency.
The CPU replacement economics behind this advantage are significant: benchmarked data shows the Nol8 appliance costs under A$50,000 per year to operate at workloads that would otherwise require over 60,000 CPUs costing A$6-7.5 million annually, reframing the technology as an infrastructure efficiency play rather than a point performance upgrade.
Proven globally for enterprises running Private AI
Across Megaport’s Virginia, Tokyo and Sydney sites, NOL8 returned byte-identical governance outputs, with distance (the speed of light) the only variable. The multi-continent result covered pre-embedding, inference and agent-to-agent simulated use cases.
Per the source, Megaport provides “private connectivity to all major global hyperscalers and a global ecosystem of NeoCloud GPU providers and private data centres.” For large enterprises, that means one policy governing AI data wherever workloads run, over private connections.
The stated benefit is freedom to move data to any cloud in the world with complete governance control. It should be noted that cross-site testing was conducted on public-cloud environments; references to deployment across hyperscalers, NeoCloud providers and private data centres describe connectivity reach and architectural capability.
Michael Reid, Chief Executive Officer, Megaport
“Enterprises increasingly want to run AI wherever it makes the most sense, across clouds, GPU providers and their own data centres, with their data governed consistently everywhere. As AI infrastructure becomes more distributed, enterprises will need new ways to move data privately between these environments while maintaining consistent governance. NOL8’s work reflects the growing industry focus on addressing this challenge, while also highlighting the increasing importance of the private, flexible connectivity provided by the Megaport network.”
Per the source, references to Megaport do not imply endorsement.
Understanding the AI Data Plane and why governance economics matter
An AI Data Plane is technology purpose-built to govern and deliver the right data to every AI agent at wire speed by enforcing policy inline. In practical terms, it ensures the correct data reaches each AI process, with governance rules applied as the data moves rather than after the fact.
The concept of “governed data per dollar” relates to total cost of ownership. As AI deployments scale, efficiency per unit of infrastructure is becoming a key buying decision, because delivering more governed data from the same hardware directly lowers cost.
There is also a GPU efficiency angle. The testing measured NOL8 removing up to 64% of data from the AI pipeline before it reached the GPUs. According to the Company, this means every megawatt of GPU capacity delivers up to 2.8X more useful AI work, while NOL8 itself consumes a fraction of the power of a CPU-powered software pipeline.
Key efficiency metrics from the testing include:
-
Up to 64% of data removed before reaching GPUs
-
Up to 2.8X more useful AI work per megawatt of GPU capacity
-
10X+ more governed data per dollar versus legacy software on CPUs
As the industry focuses on infrastructure efficiency, or “more AI per megawatt”, this economic advantage could become an important differentiator in customer decision-making.
What this means for FortifAI investors
For investors, this testing represents applied proof extending the benchmark results released to the ASX on 1 April 2026. It builds a validation trail from lab benchmark to live multi-continent deployment.
Scalability is a central theme. Throughput held steady as policy complexity grew by 400%, and the Company states the advantage widens with larger workloads. Every figure reported is described as a floor, not a limit.
Kelly Herrell, Chief Executive Officer, FortifAI Limited
“This is the applied proof of the NOL8 V1.0 engine. Performance met or exceeded our expectations on every test we ran, and when we took the same workloads to three continents across the Megaport global network, NOL8 returned the same answer, byte for byte, from every one, with a 10X cost advantage. That is what large enterprises running Private AI need: the same cost-effective governance with the freedom to choose between cloud providers, over private connections. That is the foundation for scaling NOL8 globally with our design partners.”
Next steps: global scaling with design partners
On the strength of these results, the Company is moving to global scaling of the NOL8 V1.0 platform with design partners across the Megaport global network. Per the source, timing and conversion remain subject to partner arrangements and are not assured.
The forward path outlined includes:
-
Continued global scaling of NOL8 V1.0 across the Megaport network
-
Engagement with design partners, with timing and conversion not assured
-
Extending workloads beyond the current test scope, where NOL8’s advantage is expected to widen
The Company has framed the results as the foundation for scaling NOL8 globally, building on a validation trail that now spans from benchmark testing to live deployment across three continents.
The NOL8 commercialisation roadmap published earlier in 2026 set commercial contracts as the end-CY2026 target, with the three-phase structure providing the milestone framework against which this multi-continent validation now sits as a completed execution step.
Don’t Miss the Next ASX Tech Breakout
Big News Blast delivers FREE breaking ASX tech news directly to your inbox within minutes of release, complete with in-depth analysis already done. Join 20,000+ investors who stay ahead of the market the moment announcements drop. Click the “Free Alerts” button at StockWire X to start receiving alerts today.