Innodata (NASDAQ:INOD) executives said the company closed fiscal 2025 with strong fourth-quarter results, driven by continued demand for data engineering across the generative AI lifecycle and supported by ongoing investments in capacity and innovation. Management also provided an initial outlook for 2026, calling for year-over-year revenue growth of approximately 35% or more based on current visibility, while noting potential upside as customer programs spin up quickly.
Fourth-quarter and full-year financial results
Chairman and CEO Jack Abuhoff said fourth-quarter revenue was $72.4 million, representing 22% year-over-year growth. For the full year, revenue totaled $251.7 million, up 48% year-over-year.
Interim CFO Marissa Espineli added that fourth-quarter revenue increased 15.7% sequentially from $62.6 million in the third quarter. She said adjusted gross profit was $30.1 million, up 6% year-over-year and 9% sequentially, and net income for the quarter was $8.8 million. Espineli also noted that cash increased from $73.9 million at the end of the prior quarter and from $46.9 million at year-end 2024, and that Innodata did not draw on its $30 million Wells Fargo credit facility.
Investing ahead of demand and margin expectations
Management emphasized that results were achieved while continuing to invest for growth. Abuhoff said the company carried capacity ahead of revenue ramp within cost of goods sold and increased spending in SG&A, including engineers, data scientists, and customer-facing account leadership. He characterized those investments as prudent, contributing to innovation and expanding opportunity.
Looking ahead, Abuhoff said Innodata expects early 2026 adjusted gross margins to be in the 35% to 40% range as new programs ramp, with normalization toward the company’s target of 40% or better adjusted gross margins as those programs scale and as “innovation-driven workflows” expand. He also said automation, synthetic systems, and evaluation platforms are expected to increase operating leverage over time.
During the Q&A, Abuhoff said the company is focused on seizing opportunity while remaining profitable, and that it may reinvest more as opportunity grows. Responding to a question on margin expansion, he pointed to gross margin expansion over time as Innodata brings more hybrid software-and-human solutions to market, suggesting those offerings could carry gross margins “well in excess” of current targets as they mature.
2026 outlook and customer mix
Abuhoff said the company anticipates another year of “potentially extraordinary growth” in 2026 and currently estimates year-over-year growth of approximately 35% or more. He said the estimate reflects active programs, recently awarded wins, late-stage evaluations, and other opportunities where the company has a “clear line of sight,” while acknowledging potential variability in customer ramp schedules, budget approvals, and research priorities.
Embedded in the outlook, according to Abuhoff, is an expectation that spending from Innodata’s largest customer will increase somewhat, while the rest of the customer base grows faster in aggregate. He said he expects that growth to come from a mix of the “Mag Seven,” domestic AI innovation labs, sovereign AI initiatives, and leading enterprises, contributing to customer diversification.
In response to questions about forecasting, Abuhoff said Innodata is maintaining a conservative methodology by excluding opportunities without high confidence. He added that the company’s aspiration is to beat expectations and that it may raise guidance as visibility increases, similar to prior years.
Innovation focus: from training data efficacy to agent reliability
Abuhoff devoted a significant portion of the call to describing innovation initiatives, framing them around a common theme: “Every innovation…is fundamentally a data innovation,” including data quality, composition, validation, and engineering at scale.
He said customers are increasingly asking Innodata not only to provide training data, but also to diagnose model performance gaps, design targeted datasets, and demonstrate performance impact through evaluation frameworks and fine-tuning (either on a customer model or a proxy model) before scaling. He also said the company is advancing methods to create datasets aimed at improving long-context reasoning, which he described as one of the industry’s most important technical challenges.
On agentic AI, Abuhoff said Innodata has developed three “highly complementary hybrid solutions” to address the degradation of autonomous agents in real-world production environments:
- An agent evaluation and observability platform, which he said supports annotation of agent trace data, LLM-as-a-judge evaluators, business-aligned rubrics, golden datasets for regression testing, and scaled test-data generation, with continuous monitoring and root-cause analysis post-deployment. He said Innodata anticipates a managed services engagement with a hyperscaler involving scaled test data, automated evaluations, and vulnerability identification for a customer-facing intelligent virtual assistant.
- A managed agent optimization pipeline, which he said generates realistic scenarios, automates evaluation, measures constraint satisfaction, and produces reinforcement-learning datasets. Abuhoff said Innodata has demonstrated improvements “of up to 25 points” in constraint satisfaction and claimed a widening performance gap versus standard approaches in more demanding scenarios.
- An adversarial simulation system, which he said generates diverse adversarial attacks—including jailbreaks, prompt injection via retrieval-augmented generation pipelines, multi-turn social engineering, steganographic payloads, and compound attacks—and produces mitigation datasets to strengthen guardrails. Abuhoff said this work is drawing interest from security leaders and has led to early-stage engagements, positioning Innodata in AI trust and safety and “prompt layer security.”
Physical AI and robotics-related work
Abuhoff said Innodata expects 2026 to also mark an acceleration in physical AI, with dataset quality and scale as a major bottleneck. He described a large-scale data engineering system incorporating structural validation, distribution monitoring, temporal consistency checks, and model-in-the-loop instrumentation to identify and correct defects before they affect model performance.
He said components of that system are already being used in engagements recently announced with Palantir, and that Innodata secured a “significant engagement” to create foundational datasets for next-generation robotic datasets, including egocentric data. He also said Innodata is working with a leading robotics lab to create “affordance data” at scale, which teaches systems what actions are possible in a given setting. Abuhoff added that this work positions the company to support the development of “world models,” which require richly structured datasets capturing interactions over time.
He also said Innodata developed an AI model for drone and other small object detection that exceeds prior state-of-the-art benchmarks by 6.45%, and that the company is exploring “dual-use implications” with potential customers.
Closing the call, Abuhoff said 2025 was a “great year” and that 2026 “holds the promise of being even better,” reiterating expectations for growth, potential customer diversification, and the importance of engineered data ecosystems in advancing AI systems.
About Innodata (NASDAQ:INOD)
Innodata Inc (NASDAQ: INOD) is a digital services and technology company that specializes in data engineering and artificial intelligence solutions. Founded in 1988 and headquartered in East Brunswick, New Jersey, the company provides structured content and digital transformation services to publishers, media companies, legal and compliance organizations, and other information-intensive industries. Innodata’s platform enables clients to convert unstructured text, images and multimedia into high‐quality, machine‐readable formats that support search, analytics and AI model training.
The firm’s offerings include content enrichment, metadata management, taxonomy development, digital asset management and data annotation services.
