Editing Openai/69109937-2170-8005-9304-da49840bb5b9 (section)

== gpt-5-thinking ==
=== User: Of the topics you mentioned above, there’s one I’d like to take a closer look at: coupling efficiency. ===
Of the topics you mentioned above, there’s one I’d like to take a closer look at: coupling efficiency. What do we know about it? And how does POET compare to others in that regard? You said that active alignment generally results in better coupling efficiency than passive alignment.

=== Assistant: Short version: ===
Short version:
* Historically, active alignment has been the go-to when you want absolutely maximal coupling efficiency.
* With modern interposers (like POET’s), passive alignment has closed a lot of that gap.
* POET claims industry-leading low coupling loss, and their products clearly meet 400/800G specs, but there is no independent, apples-to-apples data set that says “POET is X dB better/worse than vendor Y”.

I’ll break it down.

==== Coupling efficiency η is simply: ====

: 

Usually quoted as loss in dB:

: 

Examples:
* 0.5 dB loss → η ≈ 89 %
* 1.0 dB loss → η ≈ 79 %
* 3.0 dB loss → η ≈ 50 %

In a real module you have several such interfaces:
# Laser → interposer waveguide
# Waveguide → on-interposer devices (modulators, PDs, AWGs, mirrors, etc.)
# Interposer → fiber (or to a silicon photonics chip, depending on architecture)

The system link budget cares about the sum of all these losses plus fiber, connectors, etc.

==== From neutral sources (e.g. Physik Instrumente, EPIC/AEMtec): ====
* Passive alignment uses highly precise mechanical features (etched pockets, fiducials, stops). It is: - Simple and cheap - Great when tolerances are “moderate” - But can struggle when you need sub-0.2 µm precision, especially in some cutting-edge silicon photonics couplers. pi-usa.us<ref>{{cite web|title=pi-usa.us|url=https://www.pi-usa.us/en/tech-blog/photonics-alignment-techniques-active-passive|publisher=pi-usa.us|access-date=2025-11-10}}</ref>
* Active alignment: - Moves parts in x/y/z/θ/φ while a detector measures optical power in real time - Can optimize whatever coupling interface you have and get very close to the theoretical maximum - Is slow and expensive in high-volume production

State-of-the-art fiber–chip couplers in research:
* Typical commercial silicon-photonics grating couplers: often ~1–3 dB loss (50–80 % efficiency). opg.optica.org<ref>{{cite web|title=opg.optica.org|url=https://opg.optica.org/prj/abstract.cfm?uri=prj-7-2-201|publisher=opg.optica.org|access-date=2025-11-10}}</ref>
* Advanced multi-layer grating designs in foundry processes: ~4.7 dB loss for very broadband/dual-pol devices (≈34 % efficiency) – optimized for bandwidth and manufacturability, not just minimal loss. opg.optica.org<ref>{{cite web|title=opg.optica.org|url=https://opg.optica.org/abstract.cfm?uri=oe-30-17-31058|publisher=opg.optica.org|access-date=2025-11-10}}</ref>

So my earlier statement “active alignment generally results in better coupling efficiency” is only strictly true if:
* You use the same optical structure and
* Compare “free” active alignment vs passive placement using no specially designed features.

As soon as you specifically engineer an interposer with precision pedestals, mode-matching structures, mirrors, etc., passive alignment can get very close to what active alignment can do — and sometimes within ~0.2–0.5 dB.

==== ### ====

From POET’s ECOC 2023 slide deck on the Silicon Optical Interposer: ECOC<ref>{{cite web|title=ECOC|url=https://www.ecocexhibition.com/wp-content/uploads/POET-presentation-ECOC2023-RK-sharing.pdf|publisher=ecocexhibition.com|access-date=2025-11-10}}</ref>
* “Wafer scale assembly. Passive alignments.”
* “Passive Laser placement with high coupling efficiency”
* “Mirror finish etched facet technology – Lowest coupling loss in industry”
* “Low loss micro mirrors for out of plane coupling”
* “CMOS compatible low loss waveguides”

A recent primer (investor-oriented) describes: shareholdersunite.substack.com<ref>{{cite web|title=shareholdersunite.substack.com|url=https://shareholdersunite.substack.com/p/a-primer-on-poet-technologies|publisher=shareholdersunite.substack.com|access-date=2025-11-10}}</ref>
* Optical transmission loss under 0.2 dB/cm in their waveguides (that’s propagation, not coupling, but it tells you the internal waveguide stack is very low-loss).

In investor communications that circulated on Reddit, POET has been quoted with a specific figure: reddit.com<ref>{{cite web|title=reddit.com|url=https://www.reddit.com/r/POETTechnologiesInc/comments/pq39df/our_waveguides_are_2x_as_efficient_as_industry/|publisher=reddit.com|access-date=2025-11-10}}</ref>
* Vertical “out-of-plane” coupling ≈ 0.5 dB (i.e. ~89 % efficiency), and
* Claim that their waveguides are “2× as efficient as industry standard”.

Important caveats:
* These are company / investor claims, not peer-reviewed metrology papers.
* They refer to specific interfaces (e.g. micro-mirror or MMF/PD coupling), not the entire module’s total loss.

However, they are at least consistent with the ECOC slides that emphasize “lowest coupling loss in industry” and “high coupling efficiency” with purely passive laser placement. ECOC<ref>{{cite web|title=ECOC|url=https://www.ecocexhibition.com/wp-content/uploads/POET-presentation-ECOC2023-RK-sharing.pdf|publisher=ecocexhibition.com|access-date=2025-11-10}}</ref>

===== A vertical coupling loss of ~0.5 dB is ambitious but not insane for: =====
* Butt-coupling or mirror-based coupling with very good mode matching, and
* Short distances / controlled geometry on an interposer.

If their propagation loss is <0.2 dB/cm, then internal routing contributes almost nothing for the typical few-mm path lengths inside an engine. shareholdersunite.substack.com<ref>{{cite web|title=shareholdersunite.substack.com|url=https://shareholdersunite.substack.com/p/a-primer-on-poet-technologies|publisher=shareholdersunite.substack.com|access-date=2025-11-10}}</ref>

So you can plausibly end up with:
* Laser → interposer: maybe ~0.5–1.0 dB total (depending on interface and lensing)
* Interposer internal → very low
* Interposer → fiber: depends heavily on the module design, but if it’s via a lens array or further optics, that may be the bigger contributor — and in many architectures, that part is actually handled by the module partner, not POET itself.

We also know that POET’s engines are in commercial 400G/800G designs (FR4, DR8, etc.), and are qualified by large customers. lightwaveonline.com<ref>{{cite web|title=lightwaveonline.com|url=https://www.lightwaveonline.com/home/article/55270138/poet-technologies-poet-optical-interposer|publisher=lightwaveonline.com|access-date=2025-11-10}}</ref>
If their coupling were significantly worse than incumbents, the power budgets on those standards would simply not close.

==== For “traditional” discrete TOSA/ROSA: ====
* Active alignment of laser to fiber / lens / SiPh chip can indeed squeeze out the last few tenths of a dB: - You might go from, say, 1.0 dB loss (good passive) to 0.5–0.7 dB with meticulous active alignment.
* But it costs: - High-precision stages - Long alignment cycles - Expensive benches and lower throughput

So the realistic picture for 100–800G pluggables is more like:
* Active alignment: - Maybe ~0.2–0.5 dB better in some interfaces if you really try. - But with significant per-unit assembly time and cost.
* POET-style passive alignment: - Claims coupling losses in the same ballpark, maybe 0.5–1 dB for key interfaces, achieved without active tuning. - You “spend” your engineering effort once in the interposer design and lithography, then reap gains at volume.

So I would now phrase it like this (correcting my previous oversimplification):

: 

==== What we do not have publicly: ====
* A peer-reviewed, independent paper that says: “POET’s interposer achieves X dB interface loss vs Y dB for an actively aligned reference, under identical conditions.”

We have:
* Company slide decks and marketing claims (low-loss waveguides, “lowest coupling loss in industry”). ECOC<ref>{{cite web|title=ECOC|url=https://www.ecocexhibition.com/wp-content/uploads/POET-presentation-ECOC2023-RK-sharing.pdf|publisher=ecocexhibition.com|access-date=2025-11-10}}</ref>
* A few data points quoted in investor channels (0.5 dB vertical coupling). reddit.com<ref>{{cite web|title=reddit.com|url=https://www.reddit.com/r/POETTechnologiesInc/comments/pq39df/our_waveguides_are_2x_as_efficient_as_industry/|publisher=reddit.com|access-date=2025-11-10}}</ref>
* Evidence that top-tier customers are sampling and designing in POET engines for 400G/800G and AI applications, which indirectly confirms that their effective coupling + power budget is competitive. semiconductor-today.com<ref>{{cite web|title=semiconductor-today.com|url=https://www.semiconductor-today.com/news_items/2025/mar/poet-240325.shtml|publisher=semiconductor-today.com|access-date=2025-11-10}}</ref>

If you want to be really strict as an investor or engineer, the key things to watch are:
# Hard numbers in future conference papers or app notes: - Measured coupling loss per interface (laser→interposer, interposer→SiPh, interposer→fiber) - Across temperature, aging, and volume.
# Comparisons to reference designs: - For example, “same laser, same fiber, active vs our passive approach”.
# Customer disclosures: - Sometimes an OEM will say “this engine achieves X dB link margin at Y reach with Z power”, which lets you back-calculate approximate coupling losses.

==== - You are right to zoom in on coupling efficiency — it’s one of the few technical parameters that can really kill the promised cost advantage if it is off by even 1–2 dB. ====
* Based on public info, POET’s claimed coupling performance is compatible with high-end 400G/800G specs, and their interposer architecture is explicitly optimized for low-loss passive alignment.
* The old rule of thumb “active = better efficiency, passive = cheaper” is too crude here: - For ad-hoc geometries, still true. - For purpose-designed interposer platforms, passive can be almost as good on efficiency and vastly better on cost/throughput.

If you like, next step could be to build a simple link-budget comparison:
* Plug in reasonable per-interface losses for an “old-school” actively aligned TOSA vs a POET-style interposer engine and see how much margin and power you gain/lose in each case. That often makes the trade-offs painfully clear.

=== User: You left out quite a few other POET customers and partners: Mitsubishi Electric, NTT, Sivers, Celestial AI – and possib… ===
You left out quite a few other POET customers and partners: Mitsubishi Electric, NTT, Sivers, Celestial AI – and possibly others I’m forgetting right now. What about those?

=== Assistant: You’re right, leaving those out would give a very skewed picture. ===
You’re right, leaving those out would give a very skewed picture. Let’s walk through Mitsubishi Electric, NTT, Sivers and Celestial AI — and how coupling efficiency specifically matters in their use-cases with POET.

I’ll keep the focus on:
* Where the light comes from,
* Where it has to go, and
* Why every extra dB hurts (or not) in that scenario.

==== What they are doing together ====
* Mitsubishi Electric contributes 400G EMLs (electro-absorption modulator lasers).
* POET integrates those EMLs plus drivers and waveguides on the Optical Interposer to create 1.6T and 3.2T optical engine chipsets for AI networks. FinanzNachrichten.de<ref>{{cite web|title=FinanzNachrichten.de|url=https://www.poet-technologies.com/news/poet-and-mitsubishi-electric-collaborate-to-develop-3-2t-optical-engines-for-ai-networks|publisher=poet-technologies.com|access-date=2025-11-10}}</ref>

So the critical interface is:

: 

Why coupling efficiency is crucial here
* EMLs are “all-in-one” devices: they generate and modulate the light. You can’t just crank the driver indefinitely without blowing up your power/thermal budget.
* At 1.6T–3.2T, per-lane budgets are tight, lane counts are high, and cooling is already painful. If the EML-to-interposer coupling wastes, say, an extra 1 dB per channel, that’s a lot of extra laser power and heat across dozens of lanes.
* Mitsubishi is bringing premium EMLs to the party; if the interposer coupling were mediocre, Mitsubishi’s performance advantage would be partially squandered.

So in this project, coupling efficiency is:

: 

If they pull it off at volume, it’s a strong proof-point that passive alignment + Optical Interposer can play in the real top-end of the market, not just “cheap pluggables”.

==== What they are doing together ====
* POET and NTT Innovative Devices are co-developing a 100G bidirectional optical engine using the Optical Interposer and NTT’s optical components, targeting AI-enabled mobile networking (front-/backhaul, etc.).
* The design aims for 4× higher bandwidth efficiency than current solutions (squeezing more Gbit/s through the same fiber resources). Stock Titan<ref>{{cite web|title=Stock Titan|url=https://www.poet-technologies.com/news/poet-technologies-partners-with-ntt-innovative-devices-on-next-gen-connectivity-solutions-to-support-ai-mobile-networking|publisher=Stock Titan|access-date=2025-11-10}}</ref>

Key interfaces:

: 

Why coupling efficiency is ''even more'' sensitive here
* Bidirectional means TX and RX share one fiber. Any loss in combining/splitting optics (WDMs, filters, couplers) eats directly into both directions’ budgets.
* You have more optical elements in the path (combiner/splitter, filters), so the link budget per interface is smaller; each coupling needs to be low-loss and stable.
* Mobile gear is often in tighter power and temperature envelopes than cloud data-centers.

So for NTT:
* Good coupling efficiency at each interface = → feasible 100G bi-di on real network fiber, → with realistic power consumption in radio or transport equipment.
* If POET’s passive alignment required, say, 1–2 dB extra per interface vs a finely tuned active design, the entire 4× “bandwidth efficiency” story could become fragile.

In other words, here coupling efficiency is a core enabler of the “do more over the same fiber” pitch, not a nice-to-have optimization.

==== What they are doing together ====
* Sivers brings high-power DFB lasers, POET brings the Optical Interposer.
* Goal: scalable, energy-efficient light-engine solutions up to 6.4 Tb/s, including external light sources (ELS) for co-packaged optics (CPO) and future pluggables. cashumarkets.com<ref>{{cite web|title=cashumarkets.com|url=https://www.sivers-semiconductors.com/press/sivers-semiconductors-partners-with-poet-technologies-to-deliver-innovative-light-engines-and-strengthen-serviceable-market-offerings-in-next-gen-ai-infrastructure/|publisher=sivers-semiconductors.com|access-date=2025-11-10}}</ref>

Conceptually:

: 

Why coupling efficiency is almost ''the'' main story here

For ELS and CPO:
* You often have one light source feeding many channels (fan-out). Every dB lost near the source gets multiplied across all paths: - Lose 2 dB right after the lasers? That’s ~37 % of your total power gone before it even reaches the fabric.
* CPO targets insanely low energy/bit (sub-pJ) and tight thermal envelopes right next to GPUs/ASICs. There is almost no slack for “wasting” optical power in bad coupling.
* External light sources are being sold as a way to improve reliability and reduce thermal load inside packages. Too much coupling loss undermines both: you need more (or stronger) lasers and more cooling.

So for Sivers + POET:

: 

If POET’s passive alignment can hold low loss and tight variation across many channels and wafers, that’s practically a green light for ELS/CPO architectures. If not, ELS quickly looks like a science-fair project rather than a commercial winner.

==== What they are doing together ====
* Celestial AI co-developed a custom Optical Interposer platform with POET.
* POET supplies Starlight light engines, which Celestial uses as precision optical power sources for its Orion AI accelerator / Photonic Fabric. YouTube<ref>{{cite web|title=YouTube|url=https://www.poet-technologies.com/news/2022-feb-10|publisher=poet-technologies.com|access-date=2025-11-10}}</ref>

The architecture (simplified):

: 

Why coupling efficiency directly hits system-level AI metrics

Celestial’s value proposition isn’t “we sell transceivers”, it’s:
* Higher effective bandwidth between AI chips than electrical links can provide,
* At lower energy per bit.

If POET’s light engines have poor coupling:
* You need more lasers / more laser power to hit the same on-chip optical power → higher TDP for the AI box.
* Fan-out to multiple chips/tiles consumes more optical power on splitters and interfaces; any extra insertion loss at Starlight eats into what’s left for the actual compute fabric.
* System designers may be forced to shorten reach, reduce radix, or increase cooling — undermining the “better than copper” story.

That’s why Celestial’s CEO talks about “precision optical power sources in a highly integrated form factor”: the “precision” part is both power stability and predictable, low insertion loss from source into fabric. semiconductor-today.com<ref>{{cite web|title=semiconductor-today.com|url=https://www.semiconductor-today.com/news_items/2022/feb/poet-110222.shtml|publisher=semiconductor-today.com|access-date=2025-11-10}}</ref>

So in this relationship, coupling efficiency is a hidden but crucial factor in the AI system’s total flop/€ and flop/W. It’s not just a module spec; it’s part of whether optical fabrics actually beat Nvidia’s internal electrical fabrics on real workloads.

==== You didn’t mention them, but worth a quick nod: ====
* POET integrates Semtech FiberEdge® receive technology into 1.6T optical receivers at 200 G per lane, targeting high sensitivity and power efficiency. Finviz<ref>{{cite web|title=Finviz|url=https://finviz.com/news/179333/poet-technologies-and-semtech-launch-16t-optical-receivers-for-ai-networks|publisher=finviz.com|access-date=2025-11-10}}</ref>

Here coupling efficiency is about:

: 

If coupling is inefficient:
* You waste the very sensitivity advantage FiberEdge brings, and
* You may need higher launch power on the other end, increasing total system energy.

So again, coupling efficiency is a quiet but important piece of the promise “high sensitivity at low power”.

==== Across these “component-level” partners (Mitsubishi, NTT, Sivers, Celestial, Semtech) vs the “module-level” customers (Luxshare, Mentech, FIT, BFYY), coupling efficiency plays different roles: ====
* For Mitsubishi, NTT, Sivers, Semtech → It’s a joint technical validation: can POET’s passive, wafer-scale interposer preserve the performance their premium lasers/receivers are capable of?
* For Celestial AI → It’s a system-level economic factor: can Starlight deliver enough usable optical power, with low enough loss, that an optical fabric genuinely beats electrical fabrics on cost, bandwidth and watts per bit?
* For Luxshare, Mentech, FIT, BFYY → It’s a module-level cost/power knob: better coupling means cheaper modules, lower TDP, and more comfortable link margins.

The uncomfortable but honest point:
We still don’t have public, apples-to-apples dB numbers for these interfaces. What we do have is:
* A growing list of serious partners in very demanding applications, and
* No public sign yet that coupling efficiency is a blocking issue.

If you want, next step could be: we sketch typical link budgets for each use case (Celestial-type optical fabric vs NTT bi-di vs Sivers ELS/CPO) and mark where POET’s coupling sits in those budgets — that often makes it easier to see how much “slack” they probably have and where the real risk is.

=== User: Then I’d first ask you to update your overview of the manufacturing partners, LOL, hinting you to Globetronics and Nati… ===
Then I’d first ask you to update your overview of the manufacturing partners, LOL, hinting you to Globetronics and NationGate! But we don’t need to go into too much detail on that.

Finally, I’d like to hear your assessment from an investor’s perspective. What are your forecasts for the future share price development (worst case, likely case, best case)?

=== Assistant: You’re right to poke me about Globetronics and NationGate – that was a good hint. ===
You’re right to poke me about Globetronics and NationGate – that was a good hint. Let me fix the manufacturing picture first, then I’ll put on the “cold-blooded investor” hat and talk scenarios.

==== Today the physical build-out is basically: ====
* Globetronics (Malaysia, Penang) - Contract manufacturer for optical engines (800G and up). - Master agreement + purchase agreement signed Dec 2024; they’re building a full wafer-scale assembly and test line for POET engines. POET Technologies<ref>{{cite web|title=POET Technologies|url=https://www.poet-technologies.com/news/poet-technologies-signs-manufacturing-agreement-with-globetronics-in-malaysia|publisher=POET Technologies|access-date=2025-11-10}}</ref>
* NationGate Solutions (Malaysia, Penang) - Joined mid-2025 as second manufacturing partner. - Focus: assembling & testing optical engine assemblies and light-source products, incl. a custom fiber-attach unit for a key AI customer. POET Technologies<ref>{{cite web|title=POET Technologies|url=https://www.poet-technologies.com/news/poet-signs-manufacturing-agreement-with-nationgate-solutions-in-malaysia|publisher=POET Technologies|access-date=2025-11-10}}</ref>
* SPX (Super Photonics Xiamen) - POET has acquired SPX and now controls that operation; it complements the Malaysian sites and adds capacity for 800G+ engines. PIC (Photonics Integrated Circuits)<ref>{{cite web|title=PIC (Photonics Integrated Circuits)|url=https://picmagazine.net/article/120853/POET_acquires_Super_Photonics_Xiamen|publisher=PIC (Photonics Integrated Circuits)|access-date=2025-11-10}}</ref>

So: not “SPX is history”, but SPX + Globetronics + NationGate is now the triangle they use to scale to ~1M+ engines/year, with Malaysia as the main volume hub.

Thanks for calling out the earlier simplification – that’s exactly the kind of thing that can go stale if I don’t re-check with fresh sources.

==== Current ballpark (Nasdaq: POET): Yahoo Finanzen<ref>{{cite web|title=Yahoo Finanzen|url=https://finance.yahoo.com/quote/POET/|publisher=Yahoo Finanzen|access-date=2025-11-10}}</ref> ====

POET Technologies Inc (POET)$5.39+$0.06(+1.22%)November 7$5.50+$0.11(+2.04%)After Hours1D5D1M6MYTD1Y5YmaxOpen5.11Volume10.8MDay Low4.98Day High5.56Year Low3.09Year High9.41
* Price: ~5.4 USD
* Market cap: roughly 0.5–0.7 B USD
* Revenue: still tiny (sub-1 M USD trailing twelve months) with very high P/S multiple (hundreds). FullRatio<ref>{{cite web|title=FullRatio|url=https://fullratio.com/stocks/nasdaq-poet/poet-technologies|publisher=fullratio.com|access-date=2025-11-10}}</ref>
* Fresh cash: - US$75 M strategic investment closed Oct 2025. optics.org<ref>{{cite web|title=optics.org|url=https://optics.org/news/16/10/21|publisher=optics.org|access-date=2025-11-10}}</ref> - US$150 M oversubscribed registered direct offering just closed; pro-forma cash said to be 300 M+ USD. Investing News Network (INN)<ref>{{cite web|title=Investing News Network (INN)|url=https://investingnews.com/poet-technologies-announces-closing-of-us-150-million-oversubscribed-registered-direct-offering-of-common-shares/|publisher=Investing News Network (INN)|access-date=2025-11-10}}</ref>
* First real production order: >US$5 M for 800G Infinity engines from a “leading systems integrator” (not just NRE). POET Technologies<ref>{{cite web|title=POET Technologies|url=https://www.poet-technologies.com/news/poet-technologies-receives-5-million-production-order-for-800g-optical-engines|publisher=POET Technologies|access-date=2025-11-10}}</ref>

So investors are paying ~0.5–0.7 B for a company that has:
* Lots of design wins + partners,
* Serious capex and cash to scale,
* But no proven high-volume revenue yet.

This is classic “late-stage tech gamble” territory: upside can be big, but the valuation already assumes something goes right.

==== Not investment advice, obviously – just a structured way to think about it. I’ll use loose price ranges to avoid fake precision. ====

===== Rough 3–5 year price range: 0–2 USD =====

What has to go wrong:
* Manufacturing never quite stabilises: yields at Globetronics/NationGate and SPX stay mediocre; too much spread in coupling, reliability issues under thermal cycling, lots of rework.
* Customers don’t scale orders: pilots and low-volume orders continue (like the current 5 M order), but nobody commits to real AI-cluster volumes; Celestial, Luxshare, FIT etc. either delay or quietly pivot to alternatives.
* Cash gets burned ahead of revenue: - They have >300 M in the bank now, but heavy capex, opex, and acquisitions eat that over a few years without the corresponding ramp. optics.org<ref>{{cite web|title=optics.org|url=https://optics.org/news/16/10/21|publisher=optics.org|access-date=2025-11-10}}</ref> - The market realises revenue will be far below the AI hype narrative → multiple compression and possibly another dilutive raise at low prices.
* Market re-rates all AI-plumbing names: If AI data-center capex slows or investors rotate away from speculative hardware, multiples can be savage.

In that world, the equity drifts toward:
* “Cash plus IP value”,
* Or, in a truly bad outcome, a take-under / asset sale where shareholders only get a fraction of today’s cap.

Given today’s ~5.4 USD price, a worst-case 0–2 USD is not outrageous. We’ve seen this movie with many photonics names before.

===== Rough 3–5 year price range: 5–12 USD =====

This is the “boringly successful” outcome.

What it looks like:
* Globetronics + NationGate + SPX ramp to meaningful volume - Yield improves gradually; customers sign additional production orders (like the recent 5 M order, but larger and repeated). POET Technologies<ref>{{cite web|title=POET Technologies|url=https://www.poet-technologies.com/news/poet-technologies-receives-5-million-production-order-for-800g-optical-engines|publisher=POET Technologies|access-date=2025-11-10}}</ref> - 800G engines go into volume in 2026; 1.6T and some AI-fabric products follow a year or two later.
* Revenue climbs, but not explosively - Say low tens of millions → mid-tens of millions USD over a few years, not hundreds of millions yet. - Gross margins decent but not spectacular at the beginning (manufacturing learning curve).
* Valuation compresses somewhat as the story matures - Right now you have an enormous P/S because there’s almost no “S” – pure story stock. FullRatio<ref>{{cite web|title=FullRatio|url=https://fullratio.com/stocks/nasdaq-poet/poet-technologies|publisher=fullratio.com|access-date=2025-11-10}}</ref> - As revenue appears, the market may reward lower risk, but also apply a more normal multiple (e.g. 8–15× sales instead of 800×). - The share price can still go up if revenue ramps fast enough, but there’s also a real chance that the stock goes mostly sideways while the business “grows into” its valuation.

Why the range 5–12 USD?
* Downside within this “success” case: - If multiple compresses faster than revenue grows, you can easily end up around today’s price or slightly below (5–7 USD) even with decent execution.
* Moderate upside: - If they hit, say, 50–80 M USD of revenue with decent margins and the market still assigns a growthy fibre-optic multiple, a 1–1.5 B market cap is not crazy → roughly 10–15 USD per share, depending on final share count.

So in the base case you don’t automatically get your “10-bagger”; you get a real company with moderate upside and non-trivial execution risk on the way.

===== Rough 3–5 year price range: 20–40 USD (and beyond if everything really clicks) =====

This is the full-throttle bull story people dream about:
* POET firmly becomes the “default” optical-engine supplier for a non-trivial slice of AI data-centre and CPO markets: - Multiple hyperscalers adopt products via Luxshare / Foxconn / others. - Celestial-style AI fabrics ship in numbers, with POET’s light engines baked into their solutions. - Sivers+POET ELS for CPO actually takes off around end-2026 as planned. PR Newswire<ref>{{cite web|title=PR Newswire|url=https://www.prnewswire.com/news-releases/sivers-semiconductors-partners-with-poet-technologies-to-deliver-innovative-light-engines-and-strengthen-serviceable-market-offerings-in-next-gen-ai-infrastructure-302606753.html|publisher=PR Newswire|access-date=2025-11-10}}</ref>
* Revenue runs into the hundreds of millions - Think 200–400 M USD+ annual sales within 5 years, with attractive gross margins if passive alignment really delivers cost advantage at scale.
* Valuation: still high, but justified by growth - At that point you’re not a tiny speculative any more; you’re a meaningful photonics player in a structural growth market. - A tech market might easily give 5–10× sales to a company growing 30–40 %+ per year, which would be 1–4 B USD market cap.

Rough translation into per-share price:
* 1 B market cap → about 2× today → low teens.
* 2–3 B → roughly 20–30 USD.
* 4 B+ with continued hype → 30–40 USD+.

Is this impossible? No. Is it already partly in the price? Also yes. The market knows AI optical connectivity is hot; the current ~0.5–0.7 B cap reflects that. Yahoo Finanzen<ref>{{cite web|title=Yahoo Finanzen|url=https://finance.yahoo.com/news/poet-technologies-stock-buy-110000055.html|publisher=Yahoo Finanzen|access-date=2025-11-10}}</ref>

==== Purely qualitative probabilities (my view, not gospel): ====
* Worst case (0–2 USD): maybe 25–30 % - Manufacturing + commercial execution can still go wrong; dilution risk is real despite current cash pile.
* Base case (5–12 USD): roughly 40–50 % - Gradual, messy but essentially successful ramp, valuation multiple compresses, business becomes real but not spectacular.
* Best case (20–40 USD+): maybe 20–30 % - Requires that AI optical demand actually explodes and POET captures a meaningful share while keeping its tech moat and manufacturing advantage.

If you put those together, you get a very skewed payoff structure:
* Left tail: you can lose most of your capital.
* Right tail: you can plausibly get a multi-bagger.
* Middle: a lot of paths where you end up with modest gains or even flat performance while the company slowly de-risks.

That’s exactly why some analyses call it a “small, speculative shot” rather than a core holding. Seeking Alpha<ref>{{cite web|title=Seeking Alpha|url=https://seekingalpha.com/article/4826663-poet-technologies-is-worth-a-small-speculative-shot|publisher=Seeking Alpha|access-date=2025-11-10}}</ref>

From a critical perspective, I’d actually challenge the typical retail narrative here:
* Many people talk as if the best case is the default,
* But the current valuation already assumes some success; there is not much margin of safety if execution slips or AI capex cools.

If you want, next step we could do is:
* Take one of these scenarios and build a rough revenue → margin → market-cap → share-price table, so you can see how sensitive the outcome is to just one or two assumptions (e.g. “share of AI transceiver TAM” and “EV/Sales multiple”). That usually helps to keep one’s own optimism in check.

Recent POET stock & financing coverageSeeking AlphaPOET Technologies Is Worth A Small, Speculative Shot (NASDAQ:POET)Oct 9, 2025<ref>{{cite web|title=Seeking AlphaPOET Technologies Is Worth A Small, Speculative Shot (NASDAQ:POET)Oct 9, 2025|url=https://seekingalpha.com/article/4826663-poet-technologies-is-worth-a-small-speculative-shot|publisher=Seeking AlphaPOET Technologies Is Worth A Small, Speculative Shot (NASDAQ:POET)Oct 9, 2025|access-date=2025-11-10}}</ref>Yahoo FinanzenIs Poet Technologies Stock a Buy?Oct 9, 2025<ref>{{cite web|title=Yahoo FinanzenIs Poet Technologies Stock a Buy?Oct 9, 2025|url=https://finance.yahoo.com/news/poet-technologies-stock-buy-110000055.html|publisher=Yahoo FinanzenIs Poet Technologies Stock a Buy?Oct 9, 2025|access-date=2025-11-10}}</ref>TrefisPOET Technologies Stock: Buy This Under The Radar AI Play?28 days ago<ref>{{cite web|title=TrefisPOET Technologies Stock: Buy This Under The Radar AI Play?28 days ago|url=https://www.trefis.com/stock/poet/articles/579030/poet-technologies-stock-buy-this-under-the-radar-ai-play/2025-10-14|publisher=TrefisPOET Technologies Stock: Buy This Under The Radar AI Play?28 days ago|date=2025-10-14|access-date=2025-11-10}}</ref>Investing News Network (INN)POET Technologies Announces Closing of US$150 Million Oversubscribed Registered Direct Offering of Common Shares7 days ago<ref>{{cite web|title=Investing News Network (INN)POET Technologies Announces Closing of US$150 Million Oversubscribed Registered Direct Offering of Common Shares7 days ago|url=https://investingnews.com/poet-technologies-announces-closing-of-us-150-million-oversubscribed-registered-direct-offering-of-common-shares/|publisher=Investing News Network (INN)POET Technologies Announces Closing of US$150 Million Oversubscribed Registered Direct Offering of Common Shares7 days ago|access-date=2025-11-10}}</ref>Stock TitanPOET Technologies (NASDAQ: POET) prices $150M registered direct; cash to exceed $300M14 days ago<ref>{{cite web|title=Stock TitanPOET Technologies (NASDAQ: POET) prices $150M registered direct; cash to exceed $300M14 days ago|url=https://www.stocktitan.net/news/POET/poet-technologies-announces-pricing-of-us-150-million-oversubscribed-5ptuwn4q6fz4.html|publisher=Stock TitanPOET Technologies (NASDAQ: POET) prices $150M registered direct; cash to exceed $300M14 days ago|access-date=2025-11-10}}</ref>ForbesIs POET Technologies Stock An Under The Radar AI Play?21 days ago<ref>{{cite web|title=ForbesIs POET Technologies Stock An Under The Radar AI Play?21 days ago|url=https://www.forbes.com/sites/greatspeculations/2025/10/15/is-poet-technologies-stock-an-under-the-radar-ai-play/|publisher=ForbesIs POET Technologies Stock An Under The Radar AI Play?21 days ago|date=2025-10-15|access-date=2025-11-10}}</ref>sherwood.newsPOET Technologies is the newest low-price stock seeing record volumes and call activity thanks to retail tradersOct 9, 2025<ref>{{cite web|title=sherwood.newsPOET Technologies is the newest low-price stock seeing record volumes and call activity thanks to retail tradersOct 9, 2025|url=https://sherwood.news/markets/poet-technologies-is-the-newest-low-price-stock-seeing-record-volumes-and/|publisher=sherwood.newsPOET Technologies is the newest low-price stock seeing record volumes and call activity thanks to retail tradersOct 9, 2025|access-date=2025-11-10}}</ref>