POET is the Nasdaq ticker for POET Technologies, a Toronto company that designs a chip-scale packaging platform called the Optical Interposer. It puts lasers, detectors and light-guiding channels on a standard silicon wafer, beside the copper traces chips already use.
It is, for practical purposes, pre-revenue. The quarter ended 30 June 2026 brought in $569,925 against a net loss of $11.3 million, according to the Q2 2026 results the company furnished on Form 6-K.
What the Optical Interposer actually does
A conventional interposer is a slab of silicon carrying high-speed metal traces between chips. POET adds layers on top: waveguides that steer light, filters and gratings, and alignment features that let a laser or detector drop into place.
The company’s technology page says the stack is built with standard CMOS processes on a wafer of any size. Its claim, not yet proven at volume, is that this removes the part-by-part alignment and testing conventional photonics assembly needs, and therefore the cost.
Where the parts go: 800G, 1.6T and co-packaged optics
The first target is optical engines, the light-producing and light-detecting core of a pluggable transceiver. POET sells these to module makers, who sell finished 800G and 1.6T modules to hyperscale data centres running AI clusters.
The 20-F also names co-packaged optics (CPO) as a target format. CPO moves the optical engine onto the switch package itself, because copper runs out of reach inside AI racks, the same physics that keeps Astera Labs’ retimers busy.
The second line is light sources: external lasers for CPO systems and chip-to-chip optical links, built around an in-house hybrid laser branded Blazar.
Revenue, loss and the $796 million cash pile
Q2 2026 revenue was $569,925, up from $268,469 a year earlier and the sixth consecutive sequential increase. The company labels it non-recurring engineering and product revenue, so paid development work counts alongside shipped parts.
R&D was $5.8 million in the quarter and operating cash flow was negative $12.2 million. For 2025 as a whole, per the annual report on Form 20-F, net loss was $62,963,213 and the accumulated deficit reached roughly $297 million.
Cash and short-term investments stood at $796.3 million on 30 June 2026. Most of it arrived on 18 May 2026: one institutional investor bought 19,047,620 units at $21 for $400 million gross, each with a warrant exercisable at $26.25 until May 2029.
One reading trap: net loss swings on non-cash warrant revaluations. Q4 2025’s $42.7 million loss carried a $30.7 million warrant charge. Watch R&D and operating cash flow instead.
Orders, capacity and what to watch
In Q2 the company signed a supply agreement with Lumilens that included an initial $50 million purchase order for Optical Interposer-based engines. Management says it could reach $500 million or more over five years; the word “could” carries that sentence.
Capacity sits in Penang, Malaysia: a clean room of about 10,000 square feet at partner Globetronics rated for one million optical engines a year, per the Q2 MD&A.
The company expects to ship production units for qualification through the rest of 2026. Qualification is not revenue. Check each 6-K for whether the Lumilens order appears as recognised product sales, and how much the share count grew to fund the wait.
Is POET stock pre-revenue?
Effectively yes. Quarterly revenue is under $600,000, much of it engineering services, against a net loss above $11 million, per the Q2 2026 filing.
Who are POET’s customers?
Optical module makers, most of them unnamed. Lumilens is the only customer disclosed by name, with the $50 million initial order; the Q2 release also cites a $2.4 million post-quarter order from an unidentified existing customer.
Is POET a semiconductor stock or a photonics stock?
Both. It belongs with the silicon photonics names by product, and as a fabless designer building on CMOS wafers it sits in the same AI rack as Navitas’ power semiconductors, on the optical side of the board.
Daniel Reyes is a markets writer for S4Tips covering the AI infrastructure and semiconductor supply chain. He focuses on the companies that build and power the AI compute stack. His articles are for information only and are not financial advice.