Independent verification shows NOL8 delivers up to 118x more data per dollar than Google RE2
FortifAI (ASX: FTI) announced on 12 October 2026 that Professor Amit Klein, Head of the Federmann Cyber Security Center at The Hebrew University of Jerusalem, independently verified the performance of its NOL8 AI Data Plane against Google RE2. NOL8 delivered between 4.1x and 118x more data per dollar of infrastructure cost than RE2, with the advantage widening at every step up in workload complexity.
According to the company, the testing shows enterprises no longer have to trade tighter data protection for higher compute costs, as scrutiny of AI governance grows. Google RE2 is described in the announcement as a widely adopted engine for data pattern matching at speed and scale.
Key takeaways from the announcement:
- Peak advantage: NOL8 delivered 118x more data per dollar than RE2 at the largest ruleset tested, 5,000 rules.
- Flat cost: NOL8’s cost held at approximately US$2 per GB/s per hour across all six tests.
- Class C milestone: The verification satisfies the vesting condition for 52,516,667 Class C Performance Rights.
Kelly Herrell, Chief Executive Officer of FortifAI, said the results remove the trade-off between protection and compute cost.
Kelly Herrell, Chief Executive Officer, FortifAI
“Every enterprise deploying AI faces the same trade-off. The more tightly it protects its data, the more it pays in compute. Independent testing shows NOL8 removes that trade-off where our cost holds flat as protection tightens…”
When big ASX news breaks, our subscribers know first
How the testing worked and what the numbers show
Test design
Professor Klein’s assessment was based on his independent review of NOL8’s testing methodology and results. Six tests were run on both systems, and each test increased the number of rules, measured latency and throughput, and divided the result by hardware cost to derive a data-per-dollar figure.
The tests pushed an increasing number of rules (128 to 5,000) at chaotic intervals, which the company says reflects peaks and idle time in real-world load scenarios. Both systems ran on AWS, with NOL8 Technology V1.0 utilising FPGA compared to Google RE2 utilising CPUs.
Cost, latency and throughput
The table below summarises the results reported in the announcement. Cost is shown in US$ per GB/s of throughput per hour.
| Metric | NOL8 | Google RE2 |
|---|---|---|
| Cost per GB/s per hour | Approximately US$2 across all six tests | Rose from US$8 to US$240 |
| Latency | 3 to 5 milliseconds up to 98% of the workload | Rose quickly to 10 milliseconds |
| Throughput | Over 950 MB/s | Fell from approximately 100 MB/s to 2 MB/s |
At the smallest ruleset of 128 rules, NOL8 delivered 4.1x more data per dollar than RE2. At the largest ruleset of 5,000 rules, the figure was 118x, and RE2’s cost rose thirty-fold over the same tests.
What deterministic data processing means for AI governance
In the announcement, rules set how tight a filter’s mesh is. The tighter the mesh, the more it catches, and more rules give enterprises greater certainty that sensitive data is not passing into their AI systems. The company notes this is also where software-based filtering begins to break down.
“Deterministic” means the same input produces the same outcome every time. FortifAI describes NOL8 as designed to be the “world’s first fully deterministic data processing platform”, inspecting, filtering and applying security and governance rules to every byte of data in motion, with identical outcomes regardless of complexity or load.
For investors, the relevance lies in the cost profile. With scrutiny of enterprise data sensitivity growing, the company’s position is that full, deterministic inspection of every byte no longer comes at growing infrastructure cost.
Megaport and Equinix channels gain a third-party proof point
The announcement places the results in the context of NOL8’s distribution through the Megaport and Equinix networks, where customers pay for every byte of data processed. The sequence of channel milestones was:
- June 2026: NOL8 joined the Megaport AI Exchange (ASX: MP1) alongside Megaport’s compute, network and storage layers for proof-of-concept deployments.
- August 2026: Multi-continent testing across Megaport’s private network in Virginia, Tokyo and Sydney delivered approximately 3 milliseconds of in-region latency and 10x more governed data per dollar than legacy CPU-based software.
- September 2026: NOL8 became available through Equinix Fabric (NASDAQ: EQIX) across more than 60 locations globally.
The two sets of figures measure different things. The company states the purpose of the Megaport test was to prove 100% byte coverage across a global network, not to stress-test the system with increasing rule counts as this independently verified benchmark did.
The August multi-continent testing produced byte-identical governance outputs from Virginia, Tokyo and Sydney, which speaks to the consistency of NOL8’s deterministic design on a live private network.
The announcement says a data processing technology whose cost stays flat as complexity grows means more filtered, governed data reaching every GPU for each dollar spent. It adds that independent verification gives Megaport, Equinix and their customers a third-party basis on which to assess NOL8 through current proof-of-concept engagements ahead of general availability in the coming weeks.
Class C milestone achieved and general availability ahead
The independent verification satisfies the vesting condition for 52,516,667 Class C Performance Rights. These were issued pursuant to the company’s acquisition of FastAI Pty Ltd, first announced on 15 December 2025, and shareholder-approved CEO equity incentives.
The results extend NOL8’s published Google RE2 benchmarks announced in April 2026, and the February 2026 product roadmap set the path to commercialisation. General product availability is targeted for late-October, which the company says is ahead of NOL8’s forecast timeline.
Proof-of-concept engagements with Megaport and Equinix customers continue ahead of general availability. The announcement does not disclose pricing, revenue or share price information.
Investors exploring the path from benchmarks to revenue can read our detailed coverage of the NOL8 commercialisation roadmap, which sets out the three-phase plan, the cash position and the execution risks that remain before contracts are signed.
Get Tech News Before the Market Moves
Join 20,000+ investors receiving FREE breaking ASX news within minutes of release, complete with in-depth analysis. Click the “Free Alerts” button at Big News Blast to get the next tech announcement in your inbox the moment it breaks, and stay ahead of the market.