Appen’s (ASX:APX) post-Google turnaround has finally showed up!

Appen’s FY26 guidance of US$270m to US$300m resets the investor thesis

Few other ASX tech players were written off as much as Appen (ASX:APX). After losing Google in early 2024, a failed Innodata bid, and a 90% drawdown from the 2020 peak, the bull case effectively required investors to believe the AI data labelling business could be rebuilt from scratch.

Today’s AGM address from CEO Ryan Kolln put numbers on that rebuild. FY25 group revenue landed at US$230.8 million, up 4.5% once you strip out the Google headwind, and underlying EBITDA before FX hit US$12.2 million, a 251% jump on FY24. The Q4 EBITDA margin of 18.2% is the highest quarterly print since 2021.

The real surprise sits in the China segment. The company’s revenue from China grew 75% to US$102.9 million, with EBITDA up 640% to US$10.6 million. Management has also reaffirmed FY26 guidance of US$270 million to US$300 million in revenue.

We think this is the first AGM in three years where the slides and the cash flow are pointing in roughly the same direction.

China is now the engine, not the side bet

For most of Appen’s history, the bull case was big tech advertising work funnelled through the US business. That story died with Google. What has replaced it is something investors barely modelled, a Chinese AI data business doing US$102.9 million in revenue at a 10.3% full-year EBITDA margin.

December annualised revenue in China already exceeds US$135 million, driven by LLM work for Chinese tech majors expanding internationally. Management notes Appen China revenue is now meaningfully larger than its main listed local competitor, Data Ocean.

The skeptical read is that Chinese tech demand is lumpy and politically exposed. The constructive read is that Appen has built a defensible vendor-of-choice position western competitors cannot easily enter.

Generative AI is doing the heavy lifting on margin

Group gross margin lifted 100 basis points to 40.3%, with management attributing the move to a higher mix of generative AI work. GenAI now accounts for 44.1% of Q4 group revenue, up from 34.8% a year ago.

The project examples are worth paying attention to. Appen delivered over 50,000 units of robotics annotation data, built reinforcement learning environments across 20 domains, and shipped multi-lingual speech data across roughly 60 languages for a leading AI lab.

Several programs have been extended into follow-on phases. That conversion from pilot to expansion is the single hardest thing for a data services business to demonstrate, and it is what investors needed to see.

The Appen Global turnaround is not finished yet

The legacy Appen Global division remains the weak link. FY25 revenue of US$127.9 million was down 21% year on year excluding Google, with three soft quarters before Q4 finally broke higher to US$41.4 million at a 24.6% EBITDA margin.

Our concern is that a single strong quarter does not establish a trend, particularly in a business management itself describes as project-based and H2-skewed. The US$10 million in annualised cost efficiencies helps, but Global has to demonstrate Q4 was the floor on demand.

We would want to see Q1 and Q2 hold the Q4 margin trajectory before declaring the turnaround complete.

The Investors Takeaway for Appen

The setup heading into FY26 is genuinely different from anything Appen has offered investors since 2021. Cash of US$59.8 million removes the existential risk, China is scaling profitably, and the GenAI revenue mix is doing real work on gross margin.

What still needs to happen is straightforward. Appen Global has to prove Q4 was a turn rather than a bounce, China has to keep converting LLM pilots into expansions, and the H2-skewed guidance has to land closer to the top of the US$270 million to US$300 million range than the bottom.

We have tracked this name through its worst stretch and our prior coverage sits at stocksdownunder. The Q1 trading update will start to tell investors whether the bloodbath actually ended in 2025, or whether FY26 is just another false dawn.