Innodata vs. TaskUs: Which AI Services Stock Has Better Upside?

Innodata (INOD) delivered Q2 revenue up 58% YoY to $92.1M with adjusted EBITDA up 92% YoY to $25.4M and gross margin rising to 49% from 43%, but the stock is down 43.8% over the past three months and trades at 37.29x forward earnings. TaskUs (TASK) posted Q2 revenue up 5% to $308.9M and adjusted EBITDA of $57.7M (18.7% margin), with adjusted free cash flow of $36.4M and net leverage below 1.3x; shares are up 22.9% and it trades at just 5.24x forward earnings. Analyst revisions favor INOD (2026 EPS $1.18 vs $1.08; +28.3% EPS growth projected in 2026) while TASK faces slower EPS/growth expectations, resulting in INOD having higher upside potential but higher execution/valuation risk.

Analysis

INOD is the cleaner expression of AI infrastructure spend because it sits closer to the scarce budget line items frontier labs cannot easily commoditize: data curation, evaluation, and specialized training loops. The second-order implication is that reusable datasets and long-duration program wins can create operating leverage faster than headline revenue suggests, but they also make the stock highly sensitive to a single customer pause or reprioritization. TASK, by contrast, is a hybrid: its AI services can grow, but the legacy trust/safety and outsourcing base is vulnerable to automation, so “cheap” may simply reflect a business mix that is being structurally re-rated lower.

Over the next 1-3 months, the key catalyst is not sector enthusiasm but proof of sustainable mix shift. If INOD continues converting new model-eval and agentic-AI work into multi-quarter backlog, the market can tolerate the premium multiple; if not, the drawdown can persist because project deferrals hit sentiment before they hit EPS. For TASK, any upside depends on AI Services accelerating enough to offset erosion elsewhere; otherwise the low multiple may stay anchored as a value trap rather than a re-rating candidate.

The consensus appears to be over-weighting TASK’s valuation cushion and under-weighting its secular substitution risk, while also assuming INOD’s growth is fully durable. My base case is that the relative winner is still INOD on a 6-12 month horizon, but only on pullbacks or via a pair that isolates execution quality from market beta. Falsifiers: INOD guide-downs or customer concentration re-widens; TASK AI Services fails to reaccelerate above 30% and legacy declines deepen.

Key Decisions for Investors

• Long INOD / short TASK as a 2-6 month relative-value pair: expresses conviction that true AI infrastructure monetization deserves a premium over transitioning outsourcing. Size modestly; thesis breaks if INOD revenue growth decelerates materially or TASK AI Services reaccelerates well above expectations.
• Buy INOD only on weakness, not momentum-chase: wait for post-earnings or a 10%+ pullback to enter. Risk/reward is better if the market gives you a lower basis ahead of the next guide-up cycle; cut if management stops implying 40%+ growth.
• Avoid outright long TASK here unless you are explicitly playing cash-flow defense. The low multiple is support, but the stock is capped if legacy business erosion keeps offsetting AI services gains; use it only as a hedge against broader AI exuberance.
• Set a watch item on customer concentration and program duration for INOD. If the largest-customer share re-accelerates or a frontier-lab deal slips, treat it as a multiple-compression event rather than a one-quarter miss.