Highlights

  • FTI shares gained 15.5% to touch a 52-week high of $0.64 in today's session on volume of 1,075,654 shares, with the stock's market capitalisation standing at $176.37 million.
  • Nol8's FPGA-accelerated AI Data Plane maintained a constant 1,500 MB/s throughput across all complexity tiers and all load conditions, compared with Google RE2's collapse to just 0.007 MB/s at P99 extreme load under AI-grade workloads — a difference of more than 200,000 times.
  • The benchmark results were released ahead of Nol8's planned Enterprise Ready Benchmarking Engine release in June 2026, with the company noting that the published data represents only a portion of the technology's total performance capability.

 

FortifAI Limited (ASX:FTI) reached a new 52-week high of $0.64 in today's trading session (14 April 2026), advancing $0.084 or 15.5% on volume of 1,075,654 shares — a high-conviction move for a company operating in the ASX's emerging AI infrastructure space. The stock's bid and offer range of $0.620 to $0.635 at the close reflects the sustained buying interest that characterised the session, and the day's volume significantly exceeded what would be considered a typical trading day for a company at this market capitalisation of $176.37 million. The catalyst for the session's move was the release of benchmark testing results published by FortifAI in an ASX announcement dated 1 April 2026, in which the company's Nol8 subsidiary documented a greater than 200,000-times throughput advantage over Google's RE2 engine — the globally accepted software standard for data pattern-matching performance — under the most demanding AI-grade workload conditions. For investors tracking the intersection of hardware acceleration and artificial intelligence infrastructure, the result represented the most quantitatively definitive validation of the Nol8 technology that the company has published to date.

About FortifAI and the Nol8 Technology Platform

FortifAI Limited is an ASX-listed AI infrastructure company focused on developing and commercialising technology with a specific emphasis on artificial intelligence applications. Alongside its broader technology portfolio — which includes a video games business — the company's primary growth narrative is centred on Nol8, a subsidiary developing what it describes as the foundational AI Data Plane for the era of autonomous agents. Nol8's technology combines neural-network-based algorithms with Field-Programmable Gate Array hardware acceleration to deliver data processing at speeds that the company argues are categorically beyond what software-based architectures can achieve. The core value proposition of Nol8 addresses what the company characterises as a critical and previously unresolved bottleneck in AI infrastructure: the layer between raw, unstructured data and AI inference models. As the global AI industry moves toward agentic and autonomous systems — which operate continuously, processing data in real time without human intervention at each step — the volume and complexity of data that must be filtered, classified, enriched, and routed before reaching a model is growing at an exponential rate. Nol8 describes this processing layer as the AI Data Plane, and positions its technology as the purpose-built hardware infrastructure layer that the industry now requires but has not previously had access to.

The Benchmark: Nol8 vs Google RE2

The benchmark announced today involved testing Nol8's FPGA-accelerated engine against Google RE2 — Google's open-source regular expression engine and the most widely adopted software standard for high-performance data pattern matching. RE2 is used across Google's own infrastructure and is embedded in a broad range of modern enterprise data platforms and cloud services. Its primary architectural advantage over alternative software regex engines is a guaranteed linear processing time: performance degrades predictably rather than collapsing catastrophically under complex or adversarial inputs. However, as Nol8's testing demonstrates, predictable degradation is still degradation — and at AI-grade workload complexity, the practical performance outcome of RE2 becomes operationally untenable. The benchmark tested both systems across three complexity tiers and three load percentiles. At the lowest complexity tier, representing basic web and API filtering with approximately 10 rules, RE2 produced a P99 extreme load result of 10.8 megabytes per second — adequate for the use case it was designed for. As complexity scaled to the enterprise security tier at approximately 1,000 rules, RE2's P99 result fell to 0.28 megabytes per second. At the AI-grade tier — 6,000 or more simultaneous classification rules representative of real-time agentic AI pipelines and large language model data governance — RE2's P99 throughput collapsed to 0.007 megabytes per second. Nol8's FPGA engine, by contrast, maintained a constant 1,500 megabytes per second across every complexity tier and every load condition tested. The P99 advantage at the AI-grade tier is therefore 1,500 divided by 0.007 — a factor of more than 200,000 times.

Why P99 Is the Metric That Matters for AI Infrastructure

The significance of the P99 load condition — the 99th percentile, representing the most extreme 1% of real-world operating scenarios — is particularly acute in the context of agentic AI systems. Unlike consumer web applications or traditional enterprise software, agentic AI systems do not have predictable usage profiles. They operate continuously, 24 hours a day, seven days a week, and the data processing demands they place on infrastructure are not confined to defined peak periods that can be planned around. For enterprise AI systems operating at scale — across cybersecurity threat detection, financial services fraud monitoring, healthcare data classification, and AI agent orchestration — a data processing layer that functions adequately at median load but collapses under peak conditions is not a viable enterprise solution. It is the P99 scenario — the equivalent of Black Friday for every system simultaneously, as the company's announcement characterises it — that defines whether an infrastructure layer is genuinely enterprise-grade. The benchmark results confirm that RE2, as a software solution, cannot deliver enterprise-grade performance at AI workload complexity. Nol8, as a hardware-accelerated solution operating at the silicon level, processes data in parallel regardless of complexity load, and its performance does not degrade as rule complexity or traffic intensity increases. This architectural difference — software versus hardware at the processing level — is what Nol8's founders characterise as crossing a threshold that software alone cannot surpass.

Market Context: The AI Data Infrastructure Opportunity

The commercial context for Nol8's technology is framed by a data growth trajectory of considerable magnitude. The global datasphere is forecast to expand from 334 zettabytes in 2025 to 19,267 zettabytes by 2035 — an increase of nearly 60 times — driven overwhelmingly by unstructured data generated by AI agents, large language models, and autonomous systems operating at continuous scale. Approximately 90% of future data flows are projected to be unstructured data, which carries the highest computational processing burden of any data category. The current enterprise response to the data processing bottleneck — deploying large arrays of CPUs working in parallel to distribute the load — is characterised by Nol8's CTO as expensive, power-intensive, and structurally incapable of keeping pace with exponentially growing AI data volumes. FPGA-based hardware acceleration, as Nol8 deploys it, processes data in parallel at the hardware level and does not carry the latency or throughput limitations of CPU-based approaches. The benchmark results provide the first independently reproducible, publicly documented evidence of the scale of this performance differential in a head-to-head comparison against the dominant software standard. FortifAI has also noted that the published benchmark data represents only throughput performance results to date, with ongoing work to translate these gains into quantified infrastructure reduction outcomes — specifically, measuring the reduction in hardware footprint, computational load, and energy consumption achievable by substituting Nol8's engine for CPU arrays.

Enterprise Readiness and Upcoming Milestones

The timing of today's benchmark release is strategically positioned ahead of a significant milestone on Nol8's commercial roadmap: the planned release of its Enterprise Ready Benchmarking Engine by June 2026, previously announced to the ASX on 16 February 2026. The Enterprise Ready Benchmarking Engine is designed to enable prospective enterprise customers to conduct their own performance testing in their specific operating environments — a critical step in the commercial validation and sales cycle for an infrastructure technology of this nature. The ability to offer independently reproducible benchmark results in customer environments removes a significant barrier to enterprise adoption by allowing potential clients to validate performance claims against their own workloads rather than relying solely on vendor-published data. FortifAI has indicated that further benchmark testing will be conducted and published as the company continues to explore the full capability envelope of the Nol8 technology, with future results expected to address additional performance dimensions beyond throughput — including latency characteristics and the quantification of infrastructure cost reduction outcomes.

Conclusion

FortifAI's 52-week high of $0.64 and a 15.5% session gain reflect the market's response to what the company has characterised as a world-first validation of its Nol8 AI Data Plane technology. The benchmark confirmation of a greater than 200,000-times throughput advantage over Google RE2 at AI-grade workloads provides the most concrete, quantified evidence yet that Nol8 has developed a hardware-accelerated data processing capability that operates in a different performance category from any existing software solution. With the Enterprise Ready Benchmarking Engine milestone approaching in June 2026 and further performance characterisation work ongoing, the operational news flow from FortifAI has the potential to remain substantive through the balance of the calendar year as the company progresses from benchmark validation toward commercial engagement with enterprise customers.