Phase Space AI

Trade Construction

Innodata Inc [INOD]

Innodata Inc [INOD] — Trade Construction & Risk

As of 2026-07-29 · spot $57.56 · This memo issues NO position verdict. The book decides whether any of this is taken. What follows is vehicle analysis and risk parameters, conditional on the book choosing to act.

1. Liquidity Criteria — the actual chain was pulled first

criteria.md: "Any proposed options structure requires the actual chain pulled first — open interest and quoted size for the specific strikes and expiry. A vehicle that cannot be filled is not a vehicle." Pulled live from Alpaca, 2026-07-29.

1.1 Equity — PASS, comfortably

Metric Value
21-day ADV $65.3m
60-day ADV $172.8m
Shares outstanding 32.655m
Public float (10-K, 2025-06-30) $1.56bn

A $5m position is ~7.7% of one 21-day ADV. Entry and exit are not constrained.

1.2 Options — MARGINAL. Real, but expensive.

Listed call expiries, Nov 2026 – Jul 2027:

Expiry Strikes Total call OI Deepest strike
2026-11-20 34 3,079 $120 (629)
2027-01-15 41 7,263 $50 (658), $70 (651), $135 (447)
2027-02-19 28 240 $75 (60) — uninvestable

Live quotes and Greeks, Jan-2027 (2026-07-29):

Strike Bid × size Ask × size Spread as % of mid IV Delta
50 16.22 × 211 19.68 × 371 19.3% 1.024 0.703
60 13.19 × 16 15.28 × 141 14.7% 1.027 0.608
70 10.31 × 56 12.80 × 436 21.5% 1.041 0.525
80 8.03 × 26 8.97 × 10 11.1% 0.987 0.433
90 6.13 × 212 8.09 × 431 27.6% 1.011 0.374
100 4.47 × 229 6.39 × 213 35.3% 0.987 0.308

Assessment. Only the 2027-01-15 expiry is usable. OI is in the hundreds per strike, not the tens, so this is not the HCA failure (18 contracts across an entire chain). But bid-ask runs 11–35% of mid, which is a 20%+ round-trip friction on a defined-risk spread. Conclusion: options are executable at small size only, and the spread cost is a material drag on any structure held less than several months.

Vol context. IV ≈ 100%. My realized: 252d 105.8%, 126d 125.8%, 63d 156.5%, 21d 57.6%. IV sits below mid-term realized and well above short-term realized. There is no clean vol edge in either direction — do not construct this as a vol trade.

2. If the book takes exposure — vehicle ranking

Vehicle Assessment
1. Common stock The only vehicle with clean economics. Options friction is 20%+ round-trip; equity is frictionless at any size the book would run. Sizing (below) does the risk work
2. Jan-2027 call spread (e.g. 70/100) Defined risk, ~5.5 months, ~21% and ~35% spread cost on the two legs. Only if the book specifically wants convexity into the FY2026 print and can accept the friction at small size
3. Long calls outright 100% IV with 106–156% realized is not obviously mispriced, and theta at that IV is punitive. Not recommended
4. Nov-2026 or Feb-2027 expiries Nov OI is thinner; Feb-2027 total call OI is 240 contracts across the whole expiry — uninvestable, matching the HCA failure pattern. Rejected

3. Sizing — the active protection

criteria.md: "Inverse-volatility sizing is the active protection, and it works because a fat-left-tail name is almost always a high-volatility name and is sized down automatically."

INOD is the highest-volatility name in the batch and the mechanism is doing exactly its job. At 105.8% realized (252d), inverse-vol sizing puts it at roughly one quarter to one fifth the weight of a 25%-vol name for the same risk contribution.

Three characteristics that compound and argue for the low end of whatever the mechanism produces:

  1. 56% single-counterparty revenue concentration on 30–90 day at-will terms. This is idiosyncratic, binary-ish risk that volatility measures only after the fact.
  2. A demonstrated −52.6% drawdown in eight weeks on no disclosed negative news (2026-06-04 $121.50 → 2026-07-28 $57.56; guidance was reaffirmed on 2026-06-17 inside that move).
  3. A demonstrated +86.0% single-day gap (2026-05-08). Gap risk runs in both directions and cannot be stopped out of.

Any stop-loss placed inside 40% is a coin flip on this name and will be gapped through. Position size, not stop placement, is the risk control here.

4. Entry timing — Momentum Criteria (MEASURED; governs when, never whether)

The momentum reading is genuinely conflicting and both halves are reported:

Measure Value Reading
12-1 momentum +55.0% Strong cross-sectionally
1-year total return +16.4% Unremarkable
From 2026-06-04 peak −52.6% Severe drawdown
63-day realized vol 156.5% Extreme
21-day realized vol 57.6% Sharply decaying — the June/July disorder is subsiding

The +55% 12-1 figure is an artifact of the window straddling the +86% single-day gap on 2026-05-08. It is not evidence of trend. The honest description: the name is in a violent post-spike drawdown with short-dated volatility now compressing hard (156.5% → 57.6%). Momentum Criteria is MEASURED and blocks nothing; it argues for staging entry rather than a single fill, and it is the type-discipline test case the framework flags as most likely to be silently re-promoted to a blocking gate. It has not been.

5. Invalidation — what would kill the thesis

Ranked by how fast it would be observable:

  1. Top-customer concentration rises rather than falls. Management explicitly guided that the largest customer will be a smaller percentage of FY2026 revenue. Q1 already delivered 61% → 56%. If Q2 or Q3 shows concentration going back up, the diversification mechanism — the single load-bearing claim in the growth case — is refuted by the company's own disclosure.
  2. The ≥40% FY2026 guide is not raised again, or is cut. Q1 annualises to $360m against a $352m guide. A guide that is merely maintained through Q2 and Q3 implies sequential stagnation and would break the 12-month target's revenue base.
  3. The second Big Tech programme under-delivers against ~$51m. ~$36m remains to be recognised in Q2–Q4. It is checkable each quarter against the "another customer generated approximately X%" disclosure.
  4. The unresolved $17.2m accrued-expense build proves to be a cost timing item that reverses, which would halve the apparent Q1 cash generation. First question for the Q2 call.
  5. Further large insider selling at lower prices. $158.3m was sold at ~$97.5 average. Selling continuing at $55–60 would be a different signal entirely.
  6. Anything indicating in-sourcing or synthetic-data substitution by a frontier lab. This is the structural terminal risk and it would arrive from outside the filings.

6. Recommendation ledger

No trade is recommended and nothing is appended to trade_recommendations.jsonl — the memo issues no position verdict, and the ledger records specific recommendations, not analyses. If the book acts on this analysis, the ledger entry should carry: vehicle (common), the inverse-vol size, the bear case at −27% (top customer halves), severe at −46%, impairment at −67%, with the named cause at-will termination or volume reduction by the 56% customer on 30–90 days' notice, and the probabilities in INOD_Research.md §8 for Brier scoring.