🔆

AAOI

💡
Thesis. Revenue is exploding — up 86% — while gross margin fell to a six-quarter low. That combination is the clearest evidence in this group that AAOI is a price-taker passing through someone else's scarcity rather than a bottleneck owner. A fine high-beta trade; a poor investment.
Reviewed 11 September 2026, after Q2 2026 results. Next review after Q3 earnings on 5 November 2026.

What it does

AAOI makes optical transceivers — the plug-in devices that convert electrical signals into light and back again — for AI datacentres, plus cable TV and HFC equipment for cable operators.
Unusually for a module maker, it runs its own indium phosphide laser fab, and manufactures in Sugar Land, Texas. CEO is founder Dr. Thompson Lin.
Owning a laser fab sounds like it should confer the same pricing power Coherent and Lumentum have. The margin line says otherwise. Revenue nearly doubled and gross margin went down — which is what happens when you are buying scarce inputs and selling into a competitive market.

Bull case

  • Q2 2026 revenue $191.9M, up 86% — a fifth consecutive record quarter
  • Datacentre is where the growth is. $107.7M, up 140%, with 800G more than doubling sequentially and guided to roughly 5x in Q3
  • Guidance is well above consensus. Q3 $255–290M, up about 130% at the midpoint. FY26 around $1.1B, which management describes as entirely capacity-limited rather than demand-limited
  • The order book is full. More than $324M of 800G and 1.6T orders booked, including a $200M-plus 1.6T volume order, with orders solid through mid-2027
  • Real non-AI ballast. CATV hit a record $80.6M on DOCSIS 4.0 and 1.8GHz upgrade cycles — genuinely uncorrelated with anything else in this file
  • US manufacturing is a differentiator if the draft US ban on Chinese transceivers proceeds
  • Management targets 32–33% gross margin near-term, and 55–65% on co-packaged optics laser modules by late 2027

Bear case

  • Gross margin 27.7% — a six-quarter low, against 29.1% last quarter and 30.3% a year ago. Margin fell while revenue nearly doubled. That single fact is the thesis
  • It isn't profitable. GAAP net loss $22.8M. The "returned to profitability" headline is a non-GAAP $5.5M helped by a $14.26M tax adjustment, and adjusted EBITDA is still negative
  • Dilution is relentless. Diluted share count up 42% year over year. Revenue grew 86% but revenue per diluted share only around 31%. A $707M outflow was funded with $980M of equity
  • Guidance is split. Q3 revenue guidance beat, Q3 profit guidance came in below consensus
  • It sits downstream of the real bottleneck. The indium phosphide laser scarcity belongs to Coherent and Lumentum, and AAOI pays for it
  • Chinese module pricing runs 20–25% below Western, which caps what it can charge
  • Highest beta in the category, around 3.7

Major customers

Concentration is the single biggest structural issue on this page.
  • Microsoft — reported at roughly 44% of revenue. This is a third-party estimate rather than a filed figure, and should be verified against the next annual report
  • Amazon — holds 7,945,399 warrants issued in March 2025, vesting against purchase volume. The warrant structure is effectively a disclosed commitment
  • Oracle — named among datacentre customers
  • Cable MSOs — buying DOCSIS 4.0 and 1.8GHz headend and node equipment. A record $80.6M in Q2 2026
One customer at around 44% means a single order deferral moves the entire quarter.
🔗
Read-through. Hyperscaler capex guidance from Microsoft, Amazon and Oracle is the leading signal for AAOI's datacentre line — more directly than for any other name in this category, because of the concentration.
The more useful cross-read runs the other way. AAOI's falling gross margin on doubling revenue is positive evidence for
🔆
COHR and
🔆
LITE. It shows the scarcity rent is being captured upstream at the laser layer, not downstream at the module layer. Read AAOI's margin line as a live measurement of where the pricing power actually sits.
The CATV line is driven by cable operator upgrade cycles and is genuinely uncorrelated with AI — the reason this name has any ballast at all.

What would change the view

  • Q3 gross margin — guided 29–30.5%, and it needs to beat that and head toward the promised 32–33%
  • Share count — further ATM or equity issuance against the 42% increase already taken
  • The Chinese transceiver ban — whether it is adopted or withdrawn
  • 1.6T shipments starting late Q3 against the $200M-plus order
  • Customer diversification away from the dominant account
  • Adjusted EBITDA turning positive

Update log

  • 11 Sep 2026 — Reviewed. No change to the thesis or the verdict.
  • 7 Aug 2026 — Full review after Q2 2026 results. The six-quarter-low gross margin on 86% revenue growth became the centre of the thesis and the basis for the Avoid rating.
Research and education only — not investment advice.