The 2026 backdrop: the end of easy AI multiple expansion

The most important change in small-cap AI is not that the artificial-intelligence theme has disappeared. It has not. The real change is that the market has become far less willing to pay for a story without financial confirmation. In 2023 and 2024, the AI label itself was often enough to pull valuation multiples higher. By 2026, that shortcut is much less reliable. Investors are now differentiating between companies that have a real commercial engine and companies that still depend mainly on presentation-deck language.

This is a healthy but painful transition. Small-cap AI companies typically sit in difficult territory: they are too small to enjoy the credibility and balance-sheet strength of hyperscalers, but they often ask the market to value them as if they already had platform economics. That gap can survive during a speculative cycle. It becomes harder to defend when investors focus again on revenue quality, gross margin, free cash flow, customer concentration and diluted share count.

The result is a more selective environment. Some AI-linked small caps still deserve attention, but the bar is higher. The best candidates are not simply those with the most exciting press releases. They are the companies that can show signed contracts, durable customers, improving margins, limited dilution risk and a path toward positive cash generation. In this framework, BigBear.ai ($BBAI) and Innodata ($INOD) are useful case studies because they represent two different versions of the small-cap AI opportunity.

Case study 1 — BigBear.ai ($BBAI): national security AI and the Ask Sage bet

Company profile

BigBear.ai ($BBAI) is an artificial-intelligence company focused mainly on defense, national security, intelligence, data fusion, decision support, pattern recognition and mission-oriented analytics. The company is not a consumer AI story and it is not a cloud infrastructure giant. Its core market is institutional, government-linked and highly regulated.

That positioning gives BigBear.ai ($BBAI) a narrative advantage because defense AI is one of the few areas where buyers can be large, budgets can be strategic and compliance barriers can create real moats. At the same time, it also creates concentration risk. Government procurement cycles are slow, political, bureaucratic and exposed to budget timing. A contract pipeline can look attractive but still take longer than expected to convert into recognized revenue.

BigBear.ai ($BBAI) came public through the SPAC route, which matters for the analysis. Many SPAC-era small caps inherited complicated capital structures, high expectations and shareholder bases that were not always patient enough for the realities of enterprise or government software. For $BBAI, the central question is whether the company can move from “AI defense story” to “measurable revenue and margin compounder.”

Ask Sage: strategic logic and shareholder math

The acquisition of Ask Sage is the major event that changes the $BBAI discussion. BigBear.ai ($BBAI) announced the definitive agreement alongside its third-quarter 2025 results and later finalized the acquisition on December 31, 2025. The transaction was valued at approximately $250 million. Ask Sage brings a generative-AI platform built for defense, national security and other highly regulated environments. BigBear.ai ($BBAI) disclosed that Ask Sage supported more than 100,000 users across more than 16,000 government teams and hundreds of commercial organizations.

The strategic rationale is easy to understand. BigBear.ai ($BBAI) already had government and defense exposure. Ask Sage adds a generative-AI layer that can potentially sit closer to the user and broaden the company’s product relevance. If $BBAI can cross-sell Ask Sage into its existing relationships and use Ask Sage as a front-end layer for broader data and decision-support workflows, the deal could make the company more relevant inside secure government AI adoption.

The financial question is tougher. Ask Sage was expected to deliver approximately $25 million of annual recurring revenue in 2025, based on BigBear.ai’s transaction materials. Paying roughly $250 million for about $25 million of expected ARR means the company paid a substantial growth multiple. That does not automatically make the deal wrong, but it raises the burden of proof. The acquisition must generate durable growth, meaningful cross-selling, improved revenue quality and enough operating leverage to compensate shareholders for the capital used and any dilution connected to the transaction.

KPI checklist for $BBAI

Competitive reality

BigBear.ai ($BBAI) competes in a market where scale matters. Defense and government AI are not empty fields. Palantir ($PLTR), Booz Allen Hamilton ($BAH), Leidos ($LDOS), SAIC ($SAIC) and other larger players have deeper relationships, longer operating histories and stronger balance sheets. This does not mean a smaller company cannot win. It means the smaller company must be very clear about its niche.

Ask Sage could help BigBear.ai ($BBAI) define that niche more sharply: secure generative AI for defense, national security and other regulated environments. That is more specific than generic AI software and potentially more defensible than broad enterprise AI messaging. But defensibility must still be proven commercially. The market will need evidence that customers are not only testing the product but paying for it, renewing it and expanding usage.

Case study 2 — Innodata ($INOD): the data layer behind generative AI

Company profile

Innodata ($INOD) sits in a different part of the AI value chain. The company is not trying to become a consumer AI application and it does not compete directly with the largest model labs as a foundation-model provider. Its role is closer to AI data infrastructure: data engineering, data preparation, annotation, evaluation, structuring, validation and domain-specific workflows that help enterprise customers build or improve AI systems.

This positioning matters because high-quality AI depends on high-quality data. Large language models and agentic systems need clean, structured, audited and domain-aware datasets. In regulated areas such as healthcare, pharma, legal and compliance, the data problem is not merely technical. It is also procedural, regulatory and quality-driven.

That gives Innodata ($INOD) a more fundamental profile than many small-cap AI names. The company has reported strong revenue growth and has provided an official 2026 growth outlook. The key issue is whether the business can keep scaling without becoming a purely labor-intensive services model. For $INOD, margin expansion and customer diversification are as important as headline revenue growth.

FY2025 results and the 2026 setup

Innodata ($INOD) reported full-year 2025 revenue of $251.7 million, representing 48% annual growth. Fourth-quarter revenue was $72.4 million. Management also indicated an expectation for approximately 35% or more revenue growth in 2026, while continuing to emphasize customer diversification and investment in its generative-AI capabilities.

The next relevant near-term checkpoint is the company’s first-quarter 2026 report, scheduled after the market close on May 7, 2026, with the conference call at 5:00 p.m. ET. For investors following $INOD, that release matters because it is the first test of whether the 2026 guidance tone is being confirmed by actual quarterly execution.

Agentic AI and regulated workflows

One of the most interesting parts of the Innodata ($INOD) story is the potential application of agentic AI to complex enterprise workflows. Agentic AI refers to systems that can complete multi-step tasks, use tools, evaluate outputs and operate with more autonomy than a simple chatbot. In pharma, legal and compliance workflows, that could mean organizing clinical-trial documents, reviewing safety data, supporting regulatory documentation or structuring large bodies of specialized information.

The investment question is not whether this sounds useful. It does. The investment question is who is paying for it, at what scale, with what contract duration, at what margin and with what renewal behavior. A small pilot has limited analytical value. A multi-year deployment with a Tier 1 customer has much greater value. The filings and earnings calls must eventually show enough detail to distinguish between the two.

KPI checklist for $INOD

The main structural risk for Innodata ($INOD)

The core risk is commoditization. Large AI labs and hyperscalers are investing heavily in internal data pipelines, synthetic data, automated labeling, evaluation systems and model-improvement tools. If those systems reduce the need for third-party data-engineering providers, Innodata’s addressable market could become more competitive.

The counterargument is that regulated and domain-heavy data workflows are not easy to automate fully. Healthcare, pharma, legal and compliance data often require auditability, traceability and specialized human review. That complexity can protect parts of Innodata’s business, especially if the company keeps moving toward higher-value data and AI workflow services rather than low-margin annotation alone.

How to read a small-cap AI balance sheet

The broader lesson from BigBear.ai ($BBAI) and Innodata ($INOD) is simple: the quality of growth matters more than the label attached to the company. A business growing 25% with recurring revenue, expanding margins and low dilution may be more attractive analytically than a business growing faster through low-quality, low-margin or heavily dilutive revenue.

1. Revenue quality before revenue quantity

Readers should first ask whether revenue is recurring, contractual, usage-based, project-based or one-time. Subscription revenue, multi-year contracts and repeat enterprise deployments usually deserve more credibility than short-term project revenue. In filings, the relevant sections are revenue recognition, customer concentration, remaining performance obligations and management discussion of operating results.

2. Backlog and RPO

Remaining Performance Obligation, or RPO, is one of the most useful indicators in software and services businesses. It represents contracted revenue that has not yet been recognized. A growing RPO can signal better future visibility, especially if the company discloses how much of it is expected to convert within the next 12 months.

3. Cash runway and dilution

For unprofitable small caps, cash runway is not a side issue. It is central. The practical calculation is cash and equivalents divided by average monthly cash burn. A company with less than 12 months of runway may be forced to raise capital under unfavorable conditions. That does not make every capital raise negative, but it changes the shareholder math.

4. Gross margin as the scalability test

AI software companies should eventually show margin leverage. If revenue grows but gross margin compresses, the market may conclude that the business is closer to consulting or manual delivery than scalable software. For AI data-service companies, this test is especially important because human-in-the-loop work can be valuable but also labor-intensive.

5. Diluted share count

Diluted share count is often where the retail story breaks. A company can grow revenue and still destroy value per share if the share count grows too quickly. Warrants, convertibles, RSUs and acquisition financing all matter. The question is not only whether the company is getting bigger; it is whether each share is participating in that growth.

M&A in small-cap AI

The Ask Sage acquisition shows why M&A is likely to remain an important theme in small-cap AI. Smaller companies may need acquisitions to gain product depth, customer access or scale. Larger companies may eventually acquire specialized AI assets when it is faster to buy than to build.

Not all M&A is equal. A small company buying another small company can create value, but integration risk is high and the price paid matters enormously. A large incumbent buying or investing in a smaller AI specialist can serve as a stronger external validation, because major technology or defense players usually have the resources to perform deeper diligence. Readers should therefore distinguish between acquisition headlines and acquisition economics.

The Merlintrader Small-Cap AI Framework

This six-question checklist can be applied to any small-cap AI company before it enters a serious research watchlist.

Comparative snapshot — BigBear.ai ($BBAI) vs Innodata ($INOD)