APX: AI Data Services - Is China Enough?
APX: AI Data Services - Is China Enough?
In a Nutshell
Executive Summary
In a Nutshell
Appen collects and labels the human-generated data that AI companies use to train their models. At A$1.81 versus our fair value of A$1.50, the stock is 17% overvalued. The market is treating a strong fourth quarter as proof of a structural recovery — we think it was largely one project, and the next six months will tell us whether that's right or wrong.
Investor Profiles
| Profile | Rating | Rationale |
|---|---|---|
| Income | ★☆☆☆☆ | No dividend is expected across any forecast period. The business is reinvesting to survive a competitive transition, not to reward shareholders. Income investors should look elsewhere. |
| Value | ★★★☆☆ | At 13.7x forward EBITDA, the stock sits at par with higher-quality IT services peers but carries far greater execution risk. Our fair value of A$1.50 implies 17% downside from current prices, with a margin of safety only emerging below A$1.15. Not compelling at today's price. |
| Growth | ★★★★☆ | Revenue is forecast to grow from US$231M to US$278M in FY26 — a 20% step-up — driven by China's structural expansion. EBITDA is expected to more than double over two years. The AI training data market is growing faster than Appen itself, which is both the opportunity and the risk. |
| Quality | ★★☆☆☆ | ROIC is negative, the moat is narrow and under competitive pressure from better-funded rivals, and recurring restructuring costs have been framed as one-offs for three consecutive years. Business quality scores 5.5 out of 10. Not suited to quality-focused mandates. |
| Thematic | ★★★★☆ | Appen is one of the few listed ways to access AI training data demand directly. Its China segment provides rare on-the-ground exposure to China's domestic AI investment boom. The structural tailwind is genuine, though execution risk is high and a better-capitalised private competitor looms. |
The best fit for Appen is the growth or thematic investor with a two-to-three year horizon and a high tolerance for quarterly volatility. This is a recovery story in an early-stage market — the reward is real if execution holds, but it requires patience and the stomach for wide outcome ranges.
Executive Summary
Appen provides the human-labelled data that AI companies need to train and evaluate their models. It operates two distinct businesses: a dominant China segment serving domestic AI labs, and a Global segment competing for work from the world's largest technology companies.
FY25 was a tale of two halves. The first half was loss-making as the global business struggled to replace the revenue lost when Google terminated its contract. The second half reversed sharply, with the fourth quarter delivering an 18.2% EBITDA margin — the strongest result in years — driven by a surge in generative AI project work. Full-year normalised EBITDA came in at US$9.6M on revenue of US$231M. China grew 75% to US$103M and is now the engine of the business.
The investment case rests on three things happening simultaneously: China sustaining its growth above 25% annually, the Global segment securing recurring generative AI programs beyond the one project that drove the Q4 result, and operating leverage converting revenue growth into meaningful profit expansion. All three are plausible. None are certain.
At A$1.81 versus our fair value of A$1.50, the stock is 17% overvalued.
Results & Outlook
What happened?
FY25 revenue grew 4.5% to US$231M on an underlying basis, masking an enormous half-year split. China delivered US$103M — up 75% — while the Global segment posted US$128M, recovering from the Google loss but still inconsistent quarter to quarter. The cost restructuring program was delivered in full, reducing the operating base and enabling Q4's margin inflection. Annual free cash flow turned positive for the first time in several years at US$6.4M.
| Metric | FY25A | FY26E | FY27E |
|---|---|---|---|
| Revenue (US$M) | 230.8 | 278.0 | 322.0 |
| EBITDA (US$M) | 9.6 | 20.9 | 30.6 |
| EBITDA Margin | 4.2% | 7.5% | 9.5% |
| Free Cash Flow (US$M) | 6.4 | 0.6 | 8.0 |
| China Revenue (US$M) | 102.9 | 133.8 | 160.6 |
| GenAI % of Revenue | 33% | ~40% | n/a |
What's next?
China exited FY25 at an annualised run-rate above US$135M and is forecast to reach US$134M for the full FY26 year — a target that looks achievable if domestic Chinese AI lab investment continues at its current pace. The more important unknown is the Global segment's first half of FY26, where results are expected around August 2026.
If Global revenue exceeds US$65M in H1 with positive EBITDA, the recovery thesis gains real credibility. If it comes in below US$55M with losses, the bear case — at A$0.86 per share — becomes the more likely outcome. That single data point will tell investors more about Appen's trajectory than any other event in the next twelve months.
EBITDA is forecast to more than double to US$20.9M in FY26 as operating leverage takes hold, with the margin path from 4.2% today to 7.5% next year dependent on Global stabilisation and continued China growth.
Valuation & Risks
| Metric | Value |
|---|---|
| Fair Value | A$1.50 |
| Current Price | A$1.81 |
| Upside / Downside | −17% |
| 90% Confidence Range | A$1.13 – A$1.88 |
| Bull Case (20% probability) | A$2.25 |
| Bear Case (25% probability) | A$0.86 |
| Forward EV/EBITDA (FY26E) | 13.7x |
The single biggest risk is customer concentration. Appen's top five customers account for 74% of revenue — a figure that has risen this year despite an explicit strategy to diversify. The company lost Google, its largest customer, and spent two years absorbing the damage. Today's concentrated revenue base means that the departure or significant reduction of any one major customer would replicate that outcome. The Global segment's Q4 strength was driven largely by a single large project whose identity, renewal status, and whether it represents a new relationship or a one-time expansion has not been disclosed. Investors are pricing that quarter as a run-rate. If H1 FY26 reveals it was episodic, the bear case at A$0.86 — a 52% decline from today's price — is the relevant reference point. The US$59.8M cash balance (with no debt) means the company will survive the disappointment, but the share price may not be as forgiving.