Innodata Inc.

What they do

Innodata partners with leading technology companies, frontier AI laboratories, and enterprises to provide training and post-training data development, alignment and preference optimization, capabilities/safety evaluation, and AI enablement including agentic systems. The company differentiates through a dual role supporting both AI builders and enterprise deployers, purpose-built platforms combining automation with human oversight, and a research-driven approach to measurement and safety.

Revenue drivers

• AI Training and Post-Training Data — Design and execution of large-scale data pipelines for supervised fine-tuning and post-training alignment of foundation models. This is a core driver as model developers prioritize data quality and provenance.
• Model Evaluation, Alignment, and Safety (Evals) — Capabilities and safety evaluation services for probabilistic AI systems, analogous to testing in traditional software. Growing as AI deployment scales.
• Off-the-Shelf Datasets and High-Value Pre-Training Programs — Innodata retains intellectual property on datasets monetized across multiple customers; mix shift to these programs contributed to Adjusted Gross Margin expansion to 49% in Q2 2026, above the 40% target.
• AI Enablement and Operationalization — Support for agentic and tool-using systems, including reinforcement learning for computer-use tasks and personalization of long-horizon agents. Recent programs with largest customer scaling.

Recent performance

Revenue for Q2 2026 was a record $92.1 million, up 58% year-over-year, beating consensus by 7%. Net income was $14.4 million, or $0.41 diluted EPS, versus $0.20 in the prior-year quarter. Adjusted EBITDA reached $25.4 million (27.5% of revenue), up 92% YoY and 50% above consensus. Cash and short-term investments totaled $250.4 million as of June 30, 2026; net of customer prepayments for pass-through costs, cash was approximately $134 million.

Strategy

Innodata aims to broaden its customer base and reduce concentration: in Q2 2026 the largest customer fell to 37% of revenue from 56% in Q1, while a new Big Tech customer scaled to 34%. The company is investing in research-driven growth, including agentic reinforcement learning, public benchmarks for AI failure modes, a Cyber Training Suite for secure coding, and egocentric data-collection pilots with robotics firms. Management emphasizes converting innovation into revenue within quarters, not years.

Risks

• Customer Concentration — Historically, one customer accounted for 58% of total revenue in 2025 and 48% in 2024; although Q2 2026 dropped to 37%, loss of or reduced spending from major customers would materially affect results.
• Project-Based, At-Will Contracts — Services are typically provided under master agreements with project-based statements of work that customers can reduce, delay, or cancel at will, creating revenue unpredictability.
• Dependence on AI Market Development — Growth relies on continued adoption and investment by AI developers and enterprises; if the market for specialized data engineering and evaluation develops slower than expected, revenue may not materialize.
• Competition and Technological Change — The AI services market is competitive with emerging technologies and players; failure to keep pace with advances in model architectures or evaluation methods could erode Innodata's differentiation.

Outlook

Management reaffirmed full-year 2026 revenue growth guidance of at least 40% year-over-year. Large potential programs from new and anticipated customers, considered likely wins, are not yet included in that guidance; once scope and timing are finalized, guidance will be updated. The company expects continued margin expansion driven by mix shifts toward high-value datasets and pre-training programs.