Innodata Inc. — Blitz Index 48 (C)
An AI data engine growing 55% on frontier-lab demand — held back by two customers supplying 73% of revenue and a margin warning in Q3.
Last re-scored: November 13, 2026.
Confidence: High · Category: ai-infra
Score breakdown
| Section | Score | Detail |
|---|---|---|
| A · Speed | 20 / 25 | Thirteen straight quarters of YoY growth, all organic — FY2024 +96.4%, FY2025 +47.6%, 9M 2026 +55.0% — with FY2026 guidance raised to 42%+ |
| B · Moat | 29 / 60 | Five of the Mag 7 as customers and expansion inside existing accounts, but no platform lock-in — a project-services moat that converts to only 48.3 |
| C · Creative destruction | 9 / 15 | Expert human data and RLHF displacing scraping and crowdsourced labelling as frontier labs move to post-training — but the displacement is a service, not a standard |
| D · BlitzFailing | −10 | Two customers at 73% of nine-month revenue on at-will contracts (D2 −7), plus stock comp at 7.2% of revenue (D1 −2) — no flag, but 10 points of drag |
Think we scored it wrong? Score it yourself →
Radar axes (0–100): Moat 48.3 · Speed 80 · Creative 60.
Watch item
Gates: G1 ✓ (confirmed — 13 straight quarters of YoY growth, all organic, with repeat purchases and expansion inside existing customers) · G2 ✓ (reported — the narrow AI training-data market reaches $17–24B by 2030–2035 against a $3.50B bar, 5–7x, though filling it needs about 15% share by 2035). 17 points below the 65-point threshold — the speed is real and the cash is strong, but two customers supply 73% of revenue and Q3 operating profit fell while sales rose. Watch the Q4 wind-down of a large program, whether S&A growth is investment or cost drift, and customer diversification.
Strengths
- Revenue up 4x in three years, 13 straight quarters of YoY growth, FY2026 guidance raised to 42%+ — organic blitz, proven
- Customer diversification advancing: a customer under 5% a year ago is now the largest, with five hyperscaler programs pointing to $20M+ in 2027
- Debt-free with cash of $164.0M and 9M operating cash flow of $80.5M — self-funded growth
- Second curve: physical AI and robotics data (motion-capture lab), enterprise, and federal work
- Five of the Mag 7 are customers — a position inside the frontier-lab ecosystem
Concerns
- Two customers at 73% of revenue on at-will project contracts — the structural factor behind D2 −7
- Q3 operating profit down 26.4% YoY, S&A up 96%, gross margin down 3.6 points — a warning light on the quality of growth
- Stock comp surging (7.2% of revenue in 9M 2026) plus $38.3M of unrecognised RSUs — dilution pressure
- Labour-intensive structure: revenue per employee of $24.9k, with headcount growing slightly faster than revenue
- No platform lock-in — when a customer leaves, almost no moat remains
Research notes
Thesis: human data for frontier AI
Innodata enters the BlitzLab record with a first scoring of 48 out of 100, grade C, built from Speed 20 out of 25, Moat 29 out of 60, Creative Destruction 9 out of 15, and deductions of 10. The total sits 17 points below the 65-point line, and the shape of the score is the story: on the radar conversion used by the model, Speed converts to 80.0, Creative Destruction to 60.0, and Moat to only 48.3. This is a global data-engineering and AI systems company — trailing-twelve-month revenue of $350.2 million — whose frontier-lab customers buy training and post-training data, RLHF, and evaluation, whose revenue has grown for 13 straight quarters, and whose largest customer a year ago was under 5% of revenue. Both gates pass, so the scoring proceeds, but the second passes on reported market estimates, and the deductions are the heaviest part of the record: 10 points, seven of them from customer concentration. The fourth quarter of 2026 has not been reported at T, and its reporting date is unconfirmed — it is not a company-confirmed date and must not be presented as one — and no outcome is anticipated here. That is the balance of the record at T: speed that is real, a moat that is thin by the nature of project services, concentration that is extreme, and a margin warning in the latest quarter.
The gates: both pass — growth proven, market ample on the stated definition
Gate 1, two pieces of product-market-fit evidence for a growth company, passes on a confirmed basis. Growth is organic and unbroken: fiscal 2024 up 96.4%, fiscal 2025 up 47.6%, the first nine months of 2026 up 55.0% year on year, and the third quarter up 52.9%. Repeat purchase and expansion are proven inside existing customers: a customer under 5% a year ago is now the largest, a fourth-quarter program adds $18 million, five hyperscaler programs point to $20 million or more in 2027, and five new logos arrived in the third quarter. The second curve is also measured: physical AI and robotics data, enterprise work, and federal business. Gate 2, a ten-year market ceiling at least ten times current revenue, also passes. Revenue is GAAP revenue. The bar is about $3.50 billion, ten times trailing-twelve-month revenue of $350.2 million. The addressable market is AI training datasets plus data collection and labelling, narrowly defined: $17–24 billion by 2030–2035, or 5 to 7 times the bar. Filling it needs about 15% share by 2035 on the narrow definition, which is why the verdict is reported rather than confirmed. The deductions are the other half of the gate story: 10 points — debt at minus 2, revenue quality at minus 7, and regulatory cost at minus 1 — so the company is not flagged. If the large program winds down without replacement or the S&A surge proves to be cost drift, the deductions deepen; if diversification continues and the margin recovers, they ease.
More AI Infra companies
Other AI Infra names on the Blitz Index, highest score first: NVIDIA (87/A), Arista Networks (73/B), Advanced Micro Devices (62/B).
Next earnings
Not announced.
Coverage
Innodata Inc. · Q3 2026 reported Nov 5, 2026 · Score 48 · Grade C
Embed this score
Copy the HTML below to show this score on your site. The badge links back to this page and updates automatically when the score changes.
Preview:
BlitzLab scores businesses against a scholar’s standard — not investment advice. No price targets, no buy/sell recommendations. Figures are from SEC filings (XBRL) as labeled in the source scorecard.