Key Takeaways
- • FortifAI (ASX: FTI) reported an independently reviewed benchmark showing its NOL8 technology delivered up to 118 times more data per infrastructure dollar than Google RE2 under the most complex workload tested, with 5,000 filtering rules.
- • The advantage rose from 4.1 times at 128 filtering rules to 118 times at 5,000 rules, according to the six-test comparison.
- • NOL8 cost stayed near US$2 per gigabyte per second of throughput per hour, while RE2's cost rose from US$8 to US$240 on the same measure.
- • Professor Amit Klein of the Hebrew University of Jerusalem's Federmann Cyber Security Center independently reviewed the methodology and results.
- • General product availability is expected by late October 2026, ahead of the February commercialisation roadmap timetable.
Summary
FortifAI (ASX: FTI) has reported an independently reviewed benchmark showing its NOL8 data-processing technology delivered up to 118 times more data per infrastructure dollar than Google RE2 under the most complex workload tested, according to the company's announcement as reported on TradingView. The six-test comparison found NOL8's advantage widened from 4.1 times with 128 filtering rules to 118 times with 5,000 rules as the demands of inspecting data increased.
The tests compared NOL8 Technology V1.0 running on field-programmable gate arrays with Google RE2 running on central processing units, using shared cloud infrastructure assumptions. Each trial increased the number of filtering rules, introduced irregular bursts of activity, and measured throughput and response time relative to hardware cost. NOL8's cost remained around US$2 per gigabyte per second of throughput per hour across the trials, while RE2's rose from US$8 to US$240 on the same measure. NOL8 maintained latency of three to five milliseconds for up to 98% of the workload and throughput above 950 megabytes per second. RE2 latency reached 10 milliseconds, and its throughput fell from approximately 100 MB/s to 2 MB/s as workloads became more demanding.
Professor Amit Klein independently reviewed the testing methodology and results. The benchmark follows performance testing published in April. Commercially, NOL8 joined the Megaport (ASX: MP1) AI Exchange in June for proof-of-concept deployments. A separate multi-continent Megaport trial in August recorded roughly 3 milliseconds in-region latency and 10 times more governed data per dollar than legacy CPU-based software, though that trial tested network coverage rather than complex rule sets. In September, NOL8 became available over Equinix Fabric at more than 60 global locations.
Chief executive Kelly Herrell said NOL8 "inspects every byte and applies every rule with the same result every time," and that third-party verification matters to partners at Megaport and Equinix. FortifAI expects general product availability by late October, ahead of the timetable previously set out in its February commercialisation roadmap.
The tests compared NOL8 Technology V1.0 running on field-programmable gate arrays with Google RE2 running on central processing units, using shared cloud infrastructure assumptions. Each trial increased the number of filtering rules, introduced irregular bursts of activity, and measured throughput and response time relative to hardware cost. NOL8's cost remained around US$2 per gigabyte per second of throughput per hour across the trials, while RE2's rose from US$8 to US$240 on the same measure. NOL8 maintained latency of three to five milliseconds for up to 98% of the workload and throughput above 950 megabytes per second. RE2 latency reached 10 milliseconds, and its throughput fell from approximately 100 MB/s to 2 MB/s as workloads became more demanding.
Professor Amit Klein independently reviewed the testing methodology and results. The benchmark follows performance testing published in April. Commercially, NOL8 joined the Megaport (ASX: MP1) AI Exchange in June for proof-of-concept deployments. A separate multi-continent Megaport trial in August recorded roughly 3 milliseconds in-region latency and 10 times more governed data per dollar than legacy CPU-based software, though that trial tested network coverage rather than complex rule sets. In September, NOL8 became available over Equinix Fabric at more than 60 global locations.
Chief executive Kelly Herrell said NOL8 "inspects every byte and applies every rule with the same result every time," and that third-party verification matters to partners at Megaport and Equinix. FortifAI expects general product availability by late October, ahead of the timetable previously set out in its February commercialisation roadmap.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.