BrainChip Holdings Ltd

All figures in USD millions unless stated. Fiscal year ends 31 December.

Interim detail (most recent period)

Source: Appendix 4D Half-Year Financial Report to 30 June 2026. The loss per share is reported as "0.51" in secondary summaries; this is US cents, consistent with about 2.28 billion shares.

1H26 R&D includes US$1,348,509 of AKD2500 development costs (1H25: nil). Gross margin rose to about 78% from about 27% as licences replaced lower-margin services.

Balance sheet

At 30 June 2026, cash was US$20,304,971 and consolidated net assets were US$19,781,894. Net tangible assets were 0.85 US cents per share.

Debt: BrainChip carries no material financial borrowings. Lease liabilities are small; one aggregator shows total debt of about A$1.2 million, consistent with leases. The LDA POA was an equity facility, not debt. Goodwill and intangibles are immaterial, with capitalised development written off or impaired over time. Working capital is dominated by cash.

Cash flow

Quarterly cash trajectory (USD M, from Appendix 4C filings):

Equity financing is the funding source. Since August 2020, about A$68 million was raised through LDA (as at January 2025). A further A$25 million came in July 2024 and about A$38 million (A$35 million placement plus A$2.98 million SPP) in November–December 2025. Free cash flow is consistently negative. At the 1H26 burn rate of about US$5 million per quarter, the US$20.3 million cash balance implies roughly four quarters of runway without new revenue or capital. This is a derived estimate.

Ratios

ROE is derived as net loss divided by year-end net assets. ROA, ROIC, current ratio, interest coverage, asset turnover and cash conversion cycle cannot be computed meaningfully from the extracted data, or are not meaningful for a pre-revenue company with no debt.

Revenue CAGR FY2021–FY2025 (derived): (1.89 / 1.59)^(1/4) − 1 ≈ 4.4% a year. This understates the volatility: revenue swung from US$5.07 million to US$0.23 million within the period.

Commentary

Three phases stand out. The first, FY2021–FY2022, was the first licensing wave: upfront Renesas and MegaChips-type licence fees inflated FY2022 revenue to US$5.07 million, but almost all of it landed in 1H22. The second, FY2023, was a "product-gap" collapse. Customers waited for Akida 2.0, revenue fell about 95%, and the loss widened to US$28.9 million as the company built out sales and marketing. The third, FY2024–1H26, has been a slow rebuild. Revenue rose to US$1.89 million in FY2025 and is now led by licences. FY2025 cost discipline narrowed the loss to US$20.5 million, but 1H26 opex rose 33% to fund AKD1500 commercialisation, AKD2500 and GenAI.

The key inflection to watch is whether licence income turns into royalty income. No material royalty revenue has yet been disclosed. Until it appears, revenue covers less than 10% of operating expenses.

Management guidance

Management gives no numerical guidance. Its qualitative expectations are:

• Significant licence and product revenue growth from AKD1500 commercialisation.

• Recurring revenue from licences and royalties.

• AKD2500 tape-out in early December 2026.

• Internal GenAI demonstrations in late Q3 2026, with the GenAI product to market in 2026.

• AkidaTag mass production from Q3 2026.

Consensus

No reliable sell-side consensus exists.

Bull case

1. Mix inflection. Licences reached 50% of 1H26 revenue and gross margin jumped to about 78%. If EDGEAI, ASICLAND, ForwardEdge and MicroIP programmes reach production, royalties could compound on a near-fixed cost base.

2. Silicon now monetisable. AKD1500 is in production, and about 25% of the first inventory was sold within weeks. Boards and modules add a second, faster revenue stream that does not depend on long SoC cycles.

3. Defence tailwinds. Size, weight and power constraints, together with AFRL, Parsons, Raytheon and Lockheed Martin relationships, could yield funded development and sole-source design-ins at premium pricing.

Bear case

1. Revenue is still immaterial. TTM revenue of about US$2.1 million against an annualised opex run-rate of about US$27 million (1H26 × 2) means the company is years away from breakeven, even with strong growth.

2. Dilution treadmill. With US$20.3 million of cash, outflows of about US$5 million a quarter and the LDA facility gone, a further raise by 2027 is highly likely. Each raise has come at a discount, and the share count has grown about 33% since 2022.

3. Competitive displacement. SoC vendors can reach "good-enough" low-power inference from Arm, CEVA, Synopsys or in-house NPUs. Neuromorphic differentiation has not yet produced disclosed royalty revenue after roughly six years of licensing.

Catalysts and monitorables (next 12 months)

Analyst verdict

BrainChip is a technology-validated, commercially unproven company. Over the past 18 months it has made real operational progress. It moved from licensing reference designs to holding production silicon (AKD1500). It signed royalty-bearing Akida 2.0 licences with EDGEAI and ASICLAND. It built a credible defence and aerospace channel through Parsons, Raytheon, AFRL, Frontgrade Gaisler and Lockheed Martin's ForwardEdge ASIC. And for the first time, licences made up half of its revenue. These are the leading indicators that an IP business needs before royalties can appear.

The financial reality has not yet caught up with that narrative. FY2025 revenue was US$1.89 million, and 1H26 revenue of US$1.22 million came against a net loss of US$12.0 million. Operating costs are rising faster than revenue as the company funds AKD2500 and GenAI at the same time. Cash fell by more than a third in six months to US$20.3 million. With the LDA facility gone, another equity raise looks very likely within about a year, probably at a discount.

At roughly US$175 million of enterprise value and about 84x trailing sales, the market is valuing a royalty stream that has not been disclosed. The investment case therefore depends on a few binary checkpoints: the AKD2500 tape-out on schedule in December 2026, a credible GenAI demonstration, and, above all, the first disclosed royalty revenue or a production licence with a top-tier SoC or automotive customer.

Our assessment is that BrainChip is a speculative, high-risk technology option. Execution in 2026 has been better than in 2023. However, the evidence does not yet support a view that the business can fund itself. The company should be judged on royalty disclosures and cash burn, not on partnership announcements.