Pathkey’s Chipforge clears validation hurdle as commercial release approaches

Chipforge moves from demonstration to commercial test

Pathkey.AI Ltd (ASX:PKY) is taking its Chipforge acquisition into its next, more consequential phase: proving that an AI system capable of designing and verifying chips can become a product customers pay to use. At its 2026 AGM, the company said Chipforge had reached production-system readiness, completed an independent end-to-end validation and begun work under its first proof-of-value design contract.

The independent test is the strongest evidence presented in the update. A user took multiple designs from written intent through to deployment on a target FPGA, with the presentation stating that Chipforge required no assistance from its team and passed verification tests. That builds on the independent FPGA validation reported earlier this week, which involved working FPGA designs produced without manual code edits.

Benchmark gains come with important limits

Pathkey also compared Chipforge with published results across NVIDIA’s CVDP chip-design benchmarks, covering code completion, specification-to-RTL generation, code modification and debugging. The company’s chart showed first-pass-rate improvements of 14 to 28 percentage points against the comparison systems across the four task types.

Those figures are evidence of performance under defined benchmark conditions, not customer outcomes. Pathkey’s own disclaimer says the tests are specific, internal results unless stated otherwise, while comparison figures came from published sources and may reflect different testing methods. The presentation’s claim that the same team could complete four designs in the time of one conventional workflow is explicitly illustrative rather than a customer result.

Revenue roadmap now depends on delivery

Chipforge Studio version 1 is targeted for Q3 FY2027, with the company planning to extend support from completed AMD and Altera workflows to Lattice and Microchip in FY2028. A custom-silicon tool chain, secure infrastructure platform and fine-tuned model are also listed as FY2028 development priorities. The commercial model starts with fixed-price proof-of-value work, then moves towards additional project fees, usage charges or multi-year licences as customers bring more designs through the platform.

The first contract is therefore more significant than its immediate revenue contribution. Pathkey says the proof-of-value design is underway, while the stated ambition is to convert paid programs into recurring revenue. The earlier first Chipforge customer contract was disclosed alongside a September-quarter cash balance of A$5.83 million and only A$10,000 in customer receipts, underscoring how early the commercial cycle remains.

Cash provides runway but not proof of scale

Pathkey reported A$5.8 million in cash at 30 September, against 1,351.6 million shares on issue and a stated market capitalisation of A$21.6 million at a share price of A$0.016 on 31 October. The company said its A$2.8 million August capital raising helped fund the next stage of development, with the A$2.8 million placement also bringing Sam Weiss on as a strategic adviser.

That funding position gives Pathkey room to pursue product release and customer work, but the presentation provides no revenue forecast or confirmed timing for the transition to recurring income. Its forward-looking statements remain dependent on technical progress, customer adoption, staffing and further funding, while the financial information is historical and unaudited unless stated.

TrialKey remains an earlier-stage second engine

Pathkey’s original clinical-trial platform, TrialKey, was presented as a separate opportunity to identify weaknesses in study plans before money is committed. Analysis of 27,178 historical trials found that 88% changed plans within three years, while modelled redesign scenarios based on 100 historical records produced a median predicted 20.5 percentage-point uplift in success scores and a 31% median estimated cost reduction.

Those numbers are not clinical results. The proposed redesigns have not run in real trials, have not been assessed by subject-matter experts, and the presentation says there is no demonstrated improvement in actual trial success, cost or time to market. The next meaningful evidence for TrialKey will need to come from expert review and validation beyond historical-model scenarios.