Silicon Investor - The end of Moore's law - Poet Technologies

[toccodolce] Latest from RDDT: AGM II. - This Did Not Sound Like the Old POET Anymore I hav...

toccodolce — 6/29/2026 4:40:30 PM

Latest from RDDT: AGM II. - This Did Not Sound Like the Old POET Anymore

I have followed POET Technologies for a very long time, including AGMs going back many years. After reading the informal 2026 AGM transcript, my reaction is simple:

This did not sound like the old POET anymore.

For anyone newer to the story: POET Technologies is trying to commercialize its Optical Interposer platform — a semiconductor-style photonics integration platform for optical engines, light sources, 800G/1.6T transceivers, external light sources, and eventually near-packaged/co-packaged optics for AI infrastructure.

For years, the POET story was mostly about potential: brilliant technology, long development cycles, difficult qualification paths, delayed commercialization, and shareholders trying to infer whether the platform would ever become commercially meaningful.

This AGM felt different.

Not because of one single sentence.
Not because of one customer.
Not because of one hype number.

It felt different because the entire language of the company seems to have shifted from:

“We have promising technology.”

to:

“We have a platform, customers, orders, capacity planning, and now we need to execute.”

That is a major change.

1. The tone changedSuresh opened by calling this “the most exciting annual general meeting in POET’s history.”

That may sound like normal CEO language, but for long-term POET followers it matters. POET management has historically been careful, sometimes frustratingly so. This AGM sounded much more confident, much more direct, and much more commercially grounded.

The key theme was the “three Cs”:

Credibility, Capacity, Capability.

And the important part is that Suresh framed them not as talking points, but as “fact on the ground.”

That is the central shift.

2. Credibility now means customers putting roadmaps at riskOne sentence stood out strongly:

Customers are willing to write checks and put their own product roadmaps on the line with POET.

That is very different from ordinary sampling language.

If customers are aligning product roadmaps with POET technology, then this is no longer just about whether the Optical Interposer works technically. It becomes about design wins, qualification, customer lock-in, and multi-generation platform value.

That is the platform case.

Not one product.
Not one customer.
Not one 800G module.

A repeatable architecture.

3. “We’re not a one-product, one-customer story. We’re a platform.”For me, this may be the most important line in the entire AGM.

POET is explicitly positioning itself as a platform company.

That means:

800G today
1.6T ramping through 2027
3.2T and beyond
pluggables
near-packaged optics
co-packaged optics
external light sources
custom optical modules
wafer-scale photonics integration
hybrid integration of electrical and optical components
potentially sticky customer roadmaps

This is not a narrow transceiver component story anymore.

At least that is not how management is presenting it.

4. The design-win funnel is becoming more concreteSuresh talked about a design-win funnel that could translate into annualized revenue exceeding $100M over the next two years, dependent of course on markets, qualifications, and execution.

The important point is that he did not present this as abstract TAM.

He said the path is mapped to specific customers, specific products, and specific revenue ranges, with agreements in place.

That does not remove risk.

But it does change the nature of the risk.

The old question was:

Will anyone care?

The new question is:

Can POET deliver?

That is a much better problem to have, but it is still a very real problem.

5. H2 2026 is now the major proof windowThe transcript says POET expects to begin its production ramp in the second half of 2026.

Not prototypes.
Not engineering samples.
Volume customer shipments.

Suresh specifically referred to tens of thousands of optical engines for shipments in H2 2026 in support of production orders already received.

That is huge.

For me, this makes H2 2026 the first major validation window.

The key things to watch are:

actual shipments
revenue recognition
production yield
customer follow-on activity
Malaysia capacity ramp
ability to deliver against purchase orders

This is where POET has to cross from deep-tech promise into industrial execution.

6. Capacity is now the storyThe manufacturing comments were some of the most important parts of the AGM.

POET says it has two manufacturing partners in Penang, Malaysia. The China-to-Malaysia assembly/test transition is described as complete. Initial capacity is qualified for production. Additional capacity is being evaluated or built out through Globetronics and NationGate.

Suresh also said existing capacity is capable of producing around 1 million optical engines per year, while projected demand exiting 2027 would require around 1 million engines per month.

That implies roughly a 10x capacity expansion.

This is exciting, but also serious.

A 10x ramp is not a marketing exercise. It requires equipment, process control, supply chain, test, reliability, hiring, quality systems, and customer qualification.

This explains why Sandeep Kumar’s role as COO matters. He appears to be central to the next phase: high-volume production, manufacturing scale, supply chain, and execution in Asia.

In other words, POET is no longer just trying to prove the technology. It is trying to industrialize it.

7. The balance sheet is now a weaponPOET talked about deploying approximately $50M into capital equipment purchases in H2 2026.

For a company that has long emphasized capital-efficient manufacturing compared with traditional optical approaches, this is meaningful.

The way I read it:

POET is not abandoning the capital-efficient model.
It is using that model aggressively.

If POET’s process really requires much less capex than conventional optical manufacturing, then $50M is not a small maintenance spend. It could represent a serious manufacturing acceleration.

Suresh also discussed component shortages, equipment lead times, isolator shortages, epoxy constraints, and competitors potentially missing deliveries due to lack of balance sheet strength.

This part matters because it shows POET thinking like an operating company, not just a development company.

Secure supply.
Place deposits.
Lock in strategic suppliers.
Build ahead of demand.
Create distance from competitors.

That is new language for POET.

8. The acquisition strategy sounded more important than many may realizeSuresh discussed acquisitions and partnerships in two verticals:

External light sources
High-speed communications components that can be integrated onto the POET Interposer

This is one of the most important strategic points.

POET does not appear to be thinking only about selling individual optical engines. It appears to be thinking about controlling more of the stack.

The key phrase:

“The integration itself is the innovation.”

That is the POET thesis in one sentence.

If POET can integrate key optical/electrical components directly onto its patented interposer platform, the value is not just in the component. The value is in the manufacturable integration architecture.

That is where the moat could come from.

9. Blazar may be bigger than just another productBlazar was described as a multi-channel external laser built on POET’s patented Optical Interposer.

The interesting part is not only the specs. The interesting part is the architecture:

separation of gain chip from wavelength selection
wavelength control on the interposer
potential avoidance of DFB-array yield degradation
support for CWDM and DWDM spacing
external cavity architecture
MOPA architecture
over 300 mW per channel
wafer-level chip-scale package
semiconductor manufacturing techniques applied to photonics

This is not just “POET has a laser product.”

It sounds more like POET is trying to enter the laser/light-source stack through the same integration logic that underpins the Optical Interposer.

And that matters because the laser market may become one of the biggest bottlenecks in AI optical infrastructure.

10. OCI alignment is a major pointSuresh connected Blazar to the Optical Compute Interconnect / Optical Compute Infrastructure direction and the new MSA backed by major AI infrastructure players.

His argument was basically that the market is moving toward specifications that Blazar was already built to meet.

That is a very strong claim.

It does not mean POET automatically wins.
It does not mean qualification is done.
It does not mean revenue is immediate.

But strategically, it matters.

If standards and customer requirements move toward POET’s architecture, then POET’s years of early work may finally become an advantage instead of a burden.

That is the “market coming to us” idea.

11. “SoC for photonics” is the bigger visionAnother important section was the comparison to system-on-chip architectures in electronics.

Suresh described POET’s platform as integrating:

multiplexers
demultiplexers
waveguides
micro-mirrors
V-grooves
lenses
electrical and optical components
wafer-scale assembly
elimination of wire bonds
reduced parasitics
semiconductor-style manufacturing

He called it a true system-on-chip architecture for photonics.

That is the big vision.

Not a module.
Not a laser.
Not an engine.

A photonics integration platform.

That is why the platform language matters so much.

12. Customer retention may be built into the architectureOne of the most interesting claims was that POET’s interposer platform can scale from 800G through 3.2T and beyond using virtually the same integration architecture.

If true, that creates potential customer retention value.

Once a customer qualifies POET’s platform, future speed transitions may become easier than switching architectures completely.

That is the kind of platform stickiness investors should pay attention to.

The question is whether this works in real commercial qualification cycles.

But the logic is powerful.

13. The Q&A confirmed the execution focusThe Q&A may have been short and controlled, but it still confirmed some important points.

Suresh said several products are qualified for production or going into production imminently.

He said Lumilens products are still in development and will require qualification by Lumilens and its customers.

He said other customers are ramping in H2 2026.

Most importantly, he said POET does not expect to announce many new customers over the next 12 months, with a couple of possible major exceptions.

That is important.

It means investors should not expect a constant stream of customer announcements.

The focus is now:

complete development
qualify products
ramp production
deliver against current customers
convert demand into shippable product

His wording around the “final chasm” was very clear.

Demand has been created.
Now POET must fulfill orders.

That is the phase transition.

14. This AGM did not remove risk. It clarified the risk.My read is not that POET is now guaranteed to win.

The risks are still real:

manufacturing scale
yield
customer qualification
supply chain constraints
equipment lead times
hiring
Malaysia execution
competition from larger players
timing delays
revenue conversion

But the risk has changed.

For years, the biggest risk was whether POET would ever get real commercial traction.

After this AGM, the biggest risk appears to be whether POET can scale fast enough to satisfy demand.

That is a very different investment setup.

15. My conclusionThis AGM sounded like a company crossing from technology validation into platform industrialization.

The phrases that matter most to me are:

Real customers.
Real manufacturing.
Production orders.
Volume shipments in H2 2026.
Tens of thousands of optical engines.
1M engines/year capacity today.
Potential 1M engines/month demand exiting 2027.
$50M capex deployment.
External light source strategy.
Integration itself is the innovation.
SoC for photonics.
Same architecture from 800G to 3.2T and beyond.
The market is coming to us.
Our moment is here.

That is not old POET language.

That is platform-company language.

Now comes the only thing that matters:

Execution.

If POET ships in H2 2026, ramps Malaysia, converts design wins into revenue, and keeps customers moving through 800G, 1.6T, Blazar, NPO and CPO, then this AGM may be remembered as the point where the story changed.

Not financial advice. Just my read as a long-term POET investor who has watched many AGMs and thinks this one sounded fundamentally different.

[toccodolce] [youtube video]

toccodolce — 6/29/2026 2:51:08 PM

[Selectric II] Sylvia Jablonski of Defiance ETFs was just on Making Money With Charles Payne (F...

Selectric II — 6/5/2026 2:32:15 PM

Sylvia Jablonski of Defiance ETFs was just on Making Money With Charles Payne (FBN). Discussed her photonics plays COHR, LITE, and said she's looking at POET. Payne likes it, too.

[toccodolce] It was what Jensen said that $MRVL was the next $1 trillion dollar company. The...

toccodolce — 6/2/2026 10:28:52 AM

It was what Jensen said that $MRVL was the next $1 trillion dollar company. The next arena is connectivity switching from copper to photonics.

Tocco

[scrooge] Benzinga... POET Technologies Inc (NASDAQ:POET) rose 11.37% to $15.47 in pre-m...

scrooge — 6/2/2026 10:18:02 AM

Benzinga...

POET Technologies Inc (NASDAQ:POET) rose 11.37% to $15.47 in pre-market trading, catching direct tech tailwinds as demand for high-speed optical interlinks scales alongside massive data center hardware buildouts...

...

[Heywood40] Good call! POET is a solid double since you counseled investors to have patienc...

Heywood40 — 5/27/2026 2:39:05 PM

Good call!

POET is a solid double since you counseled investors to have patience with it...

Where do you see it going from here?

[nicewatch] Last I checked this was still a POET thread and not meant for your personal trad...

nicewatch — 5/26/2026 9:30:07 PM

Last I checked this was still a POET thread and not meant for your personal trading diary. Please take it elsewhere, or start your own thread for your purpose. Thank you.

[zax] Well, that (PONY) fizzled after open. I think with the Robotaxi hype, another p...

zax — 5/26/2026 10:37:31 AM

Well, that (PONY) fizzled after open. I think with the Robotaxi hype, another pop seems likely. But these Chinese issues are a different animal, a lot of brutal competition going on there, BYDDY, BIDU, trading at big discounts relative to industry peers.

[zax] Another obvious setup: PONY. +15% pre-market, and I'm quite sure it will keep r...

zax — 5/26/2026 8:34:41 AM

Another obvious setup: PONY. +15% pre-market, and I'm quite sure it will keep running at open.

Currently $10.30

[toccodolce] [X] 🚨 $EUV : Finally — The first U.S.-listed Optics ETF is here. Corgi Lithog...

toccodolce — 5/22/2026 9:30:43 AM

[X]

🚨 $EUV : Finally — The first U.S.-listed Optics ETF is here.

Corgi Lithography & Semiconductor Photonics ETF — actively managed, 41 holdings, built to capture the photonics value chain powering AI infrastructure.

The top 10 reads like a who’s who of the AI optical interconnect… pic.twitter.com/R8FNL3jo2i
— Sandeep Anand (@SanCompounding) May 14, 2026

[/X]

[X]

Good timing.

We are reaching the point where optics is evolving from a “supporting AI component” into its own standalone investment theme.

The bottleneck discussion is increasingly moving beyond compute and memory toward optical interconnect, bandwidth and power efficiency. pic.twitter.com/zoTLH9kqaA
— Emma Stock Notes (@EmmaStockNotes) May 15, 2026

[/X]

[X]

If you re-rate $MU and $SNDK to 17x forward PE$MU is $1,700 $SNDK is $2,650

It's called the "Embarrassment Case."
The semiconductor sector average is 34x.

17x is only half. $DRAM

Half the credit for companies building the Cognitive Capacity of AIs, the KV Cache.
— Trade Whisperer (@TradexWhisperer) May 21, 2026

[/X]

[toccodolce] [X] Take This Short Quiz How well do you know our company? Here is the first st...

toccodolce — 5/21/2026 7:25:27 PM

[X]

Take This Short Quiz
How well do you know our company? Here is the first step in providing some info for which we know many of you have longed.

Comment with your best guesses. Short on time? Don’t worry — we’ve got you covered. We’ll flex with the answers when it’s most… pic.twitter.com/gXQXVEigW1
— POET Technologies (@POETtech) May 17, 2026

[/X]

Latest $POET thread from RDDT

WiseAddiction
OP• 1d ago

"So we now know that the ramp is taking shape and we are receiving 535 wafers from Silterra a month. If we extrapolate out at 400 OI per wafer that is over 200k OI per month!! That sounds like a pretty good ramp to start. As we grow here these numbers are going to be dwarfed and this ship is about to hit the Moon!! "

[zax] RGTI +24%Message 35513216

zax — 5/21/2026 10:25:33 AM

[zax] Now -38%. A very high beta MOMO.

zax — 5/18/2026 12:53:34 PM

[zax] Wish my last short held up on it's stop... we are -22% from there. Seems the pu...

zax — 5/15/2026 3:32:35 PM

Wish my last short held up on it's stop... we are -22% from there.

Seems the pull back was predictable, but minimizing risk to capitalize on both the short term ups and downs of a high beta is tricky indeed.

Maybe I'll ride the next wave too...

[toccodolce] [X] $POET is raising $400M from one institutional investor at $21/unit, above i...

toccodolce — 5/15/2026 9:10:07 AM

[X]

$POET is raising $400M from one institutional investor at $21/unit, above its prior close of $20.57. Proceeds will fund manufacturing expansion, R&D, acquisitions and working capital. pic.twitter.com/qxFUkAJlg4
— Wall St Engine (@wallstengine) May 15, 2026

[/X]

[X]

$POET .... POET Technologies secures financing of $400 mln in registered direct offering of common shares and warrant (20.57)
Co announces it has entered into a definitive agreement to issue and sell 19,047,620 common shares and a warrant exercisable for 19,047,620 Common Shares… pic.twitter.com/6dBzmpJVKa
— Marty Chargin (@MartyChargin) May 15, 2026

[/X]

POET Technologies Secures Financing of US$400 Million in a Registered Direct Offering of Common Shares and Warrant

POET Technologies Inc.
Fri, May 15, 2026 at 8:30 AM EDT

Additional changes in company executive leadership also announced

TORONTO, May 15, 2026 (GLOBE NEWSWIRE) -- POET Technologies Inc. (NASDAQ: POET) (the “Corporation”), the designer and developer of Photonic Integrated Circuits (PICs), light sources and optical modules for the AI and data center markets, today announced it has entered into a definitive agreement to issue and sell 19,047,620 common shares (the “Common Shares”) and a warrant exercisable for 19,047,620 Common Shares (the “Warrant” and together with the Common Shares, the “Offered Securities”) to a single institutional investor in a non-brokered registered direct offering. The combined price of one Common Share and the accompanying Warrant in respect of one Common Share will be US$21.00, raising aggregate gross proceeds of US$400,000,020. The Warrant will be exercisable at an exercise price of US$26.15 per Common Share for a period of three years from the date of issuance. The closing of the financing is expected to take place on or about Monday, May 18, 2026, subject to the satisfaction of customary closing conditions.

The combined price for the Offered Securities represents a premium over the closing price of the Corporation’s common stock as quoted on the NASDAQ Capital Market of US$20.57 on Thursday, May 14, 2026. The Corporation intends to use the net proceeds from this investment for the expansion of its manufacturing infrastructure, support for corporate development activities, including targeted acquisitions, scaling up of R&D, acceleration of the light source business, expanding operations, and general working capital.

This offering is being made pursuant to an automatic shelf registration statement on Form F-3 that was filed with the U.S. Securities and Exchange Commission (the “SEC”) and became automatically effective on January 22, 2026. A prospectus supplement relating to the offering will be filed with the SEC and will be available on the SEC’s website located at www.sec.gov. Electronic copies of the prospectus supplement may be obtained, when available, by contacting the Corporation by phone at (416) 368-9411 or by email at legal.poet.tech.

This press release shall not constitute an offer to sell or the solicitation of an offer to buy, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation, or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.

Executive Leadership Changes

As noted in the Company’s press release dated May 12, 2026, POET previously announced the appointment of Sandeep Kumar as Chief Operating Officer. In addition, Thomas Mika, the Company’s longstanding Chief Financial Officer, has informed the Company’s board of directors (the “Board”) of his desire to retire sometime this year from this role, after ten years of service to the Company. The Board has acknowledged Mr. Mika’s decision and is commencing a search for a suitable successor.

[VisionsOfSugarplums] POET Technologies Reports First Quarter 2026 Financial Results finviz.com Nota...

VisionsOfSugarplums — 5/14/2026 10:06:49 PM

POET Technologies Reports First Quarter 2026 Financial Results

finviz.com

Notable Business Highlights:

Announced a strategic collaboration with LITEON Technology, one of the world’s leading providers of optoelectronic semiconductor components and high-power optical systems. The partnership aims to co-develop next-generation optical communication modules built on POET’s patented optical interposer technology and integration platform.

Announced the joint development of a 1.6T 2×DR4 optical transceiver module designed for next-generation AI clusters and hyperscale data center networks with Lessengers.

Announced its intention to move the Company’s headquarters to and domicile the Company in the U.S. so that it will no longer be a foreign corporation, which would eliminate the possibility of the Company being classified as a PFIC in future years.

Entered into a supply agreement that establishes a strategic joint development and commercial technology partnership to advance a new class of wafer-level photonic integration for frontier AI infrastructure. The supply agreement between Lumilens and the Company establishes a commercial framework to support the joint development program, and, as part of that framework, Lumilens has placed an initial purchase order with the Company for the manufacturing of EOI-based engines valued at $50 million. This purchase order represents the first phase of a broader supplier relationship that could scale to $500+ million in cumulative purchases from the Company over five years.

[zax] Read my posts.

zax — 5/14/2026 4:58:13 PM

[da_cheif™] wheres ur stop

da_cheif™ — 5/14/2026 4:39:39 PM

[zax] Not if they stop you from getting on the wrong side of a runaway train. Poet no...

zax — 5/14/2026 4:34:12 PM

Not if they stop you from getting on the wrong side of a runaway train. Poet now printing $21.50 after hours.

[da_cheif™] stops will cost u money

da_cheif™ — 5/14/2026 4:17:06 PM

[zax] Look at all that short covering at close. Its gone up another 5% in the last 10...

zax — 5/14/2026 3:59:57 PM

Look at all that short covering at close. Its gone up another 5% in the last 10 minutes of trading.

[zax] I put in the stop when it fell fairly below my entry, to lock it in. I do like ...

zax — 5/14/2026 3:37:41 PM

I put in the stop when it fell fairly below my entry, to lock it in. I do like to be sure I'm not going to get on the losing side of a runaway train quickly. Long money is much easier to make than short, and much less nerve racking.

[w0z] Yikes your stop was 3 cents below what you paid? Talk about a hair trigger! I'...

w0z — 5/14/2026 3:25:20 PM

Yikes your stop was 3 cents below what you paid? Talk about a hair trigger! I'm just sitting on a 7 figure gain in LEAPs (mostly $10 Jan 27) just waiting and watching the fireworks. January is a long way off so doing nothing.

[zax] Got in again at $20.04 and got stopped out at $20.01 again already, LOL.

zax — 5/14/2026 2:33:56 PM

[benwood] I would consider reloading on the gap fill.. we'll see! A lot of froth here. T...

benwood — 5/14/2026 2:19:50 PM

I would consider reloading on the gap fill.. we'll see! A lot of froth here.

The other good news here is the class action filing by whatever that firm was surely is moot now.

[zax] It was short term opportunistic after 40% run on the day. I was stopped out a l...

zax — 5/14/2026 12:40:16 PM

It was short term opportunistic after 40% run on the day. I was stopped out a little better than even.

I've no idea where this one will settle out, I just know it has outrageously high beta.

I made a big gain on the long side earlier this month.

[da_cheif™] shares to short........so ur calling a sell now

da_cheif™ — 5/14/2026 12:34:27 PM

[toccodolce] [X] 今天 $POET 的订单让他们一度涨超30%， POET Technologies 今日宣布与 Lumilens 达成合作，初始采购订单 5000 万...

toccodolce — 5/14/2026 12:32:08 PM

[X]

今天 $POET 的订单让他们一度涨超30%， POET Technologies 今日宣布与 Lumilens 达成合作，初始采购订单 5000 万美元，五年累计可达 5 亿美元。产品路线图覆盖 800G、1.6T 可插拔收发器，再到近封装光学和 CPO，工程样品预计 2026 年底，量产对接超大规模云厂商在 2027 年。这一次合作能覆盖… pic.twitter.com/I5kdfHtTbW
— mark (@cherryPayment) May 14, 2026

[/X]

Translation:

Today's $POET orders pushed them up over 30% at one point. POET Technologies announced today a partnership with Lumilens, with an initial purchase order of $50 million, potentially accumulating up to $500 million over five years. The product roadmap covers 800G and 1.6T pluggable transceivers, extending to near-package optics and CPO, with engineering samples expected by the end of 2026, and mass production aligning with hyperscale cloud vendors in 2027. This collaboration can encompass all upstream and downstream aspects of $poet, with specific details as follows:

1. $sive will receive indirect positive impacts, as $sive and $poet share a very strong long-term partnership, with core POET being a buyer of Sivers lasers. So, POET securing larger orders ? requiring more optical engines ? theoretically needing more Sivers InP lasers as light sources. This logical chain holds up. Therefore, as an upstream laser supplier, order pressure will be transmitted to $sive.

2. We can build a model, for example, assuming laser costs account for 15% of module costs, the laser procurement value corresponding to a $500 million order is roughly estimated at $75-125 million. In comparison, Sivers' current full-year revenue is only 360 million Swedish kronor (about $35 million), so this one client's potential order could equate to **23 times** its existing revenue.

Why do I calculate it this way?

In the cost breakdown of an optical engine module, it's roughly:

1. Photonic integrated chips (POET's EOI): about 40-50%

2. Lasers/light sources: about 15-25%

3. Packaging, electronic components, testing: about 30-40%

So my 15% estimate is very reasonable.

3. Looking at $poet's overall roadmap: 800G ? 1.6T pluggable ? near-package optics ? CPO. Among these, Sivers' core competitiveness lies in the rising failure rates under 1.6T architecture, where InP solutions become a hard requirement. This is a rigid need that's hard to change.

4. As a core partner, $sive faces technical risks and certification costs if POET switches suppliers. Moreover, Sivers' InP100 platform is specifically optimized for CPO scenarios, with strong compatibility.

So I believe that $poet's this contract will directly benefit the company itself, but the substantial orders it brings will benefit $sive. If this order is confirmed, $sive could gain at least $40 million in revenue from this single order in 2028. Adding in the synchronous growth from existing clients like Jabil, Ayar Labs, etc., a 3-4x revenue increase by 2028 holds up numerically.

Very bullish! This might be my personal speculation and not necessarily accurate, but they do have a partnership, and an upstream-downstream relationship does exist. The only risk is that $500 million is an upper limit commitment rather than a guarantee, with actual fulfillment depending on how many hyperscale client orders Lumilens secures itself.

But no matter what ! bullish! YOLO!

[zax] The AI's I traded into, NBIS, CRWV, QS, WOLF are all doing well today. I got st...

zax — 5/14/2026 11:27:45 AM

The AI's I traded into, NBIS, CRWV, QS, WOLF are all doing well today. I got stopped a little better than breakeven on the POET short at $19.30. It was a great hedge; I'm surprised it came back to that level again.

[w0z] Lumilens is backed by leading venture capital investors. [graphic] [graphic]

w0z — 5/14/2026 10:01:47 AM

Lumilens is backed by leading venture capital investors.

[zax] OK, nearly +40% now. Did I call this, or not? I couldn't help myself, called F...

zax — 5/14/2026 9:41:03 AM

OK, nearly +40% now. Did I call this, or not?

I couldn't help myself, called Fidelity and got some shares to short.

[toccodolce] POET & Lumilens You probably have not heard of Lumilens yet. You will. In...

toccodolce — 5/14/2026 9:39:08 AM

POET & Lumilens

You probably have not heard of Lumilens yet. You will.

In San Jose, California, roughly forty miles south of San Francisco, a two-year-old photonic interconnect company has been doing the kind of quiet work that tends to precede a loud arrival. It has drawn capital from two of Silicon Valley’s most disciplined infrastructure investors, Mayfield and Spark Capital. It has assembled an engineering team out of Cisco, Juniper, Meta, Lumentum, and Coherent — the institutions whose own transitions defined the last three decades of data center networking. And this week, after most of that work happening behind closed doors, Lumilens chose its first publicly announced partner: POET Technologies $POET

That choice is the story.

A Company Built by People Who Have Done This Before

Every major networking era reaches a point when the old architecture starts to strain and the next one becomes inevitable. Lumilens founder and CEO Ankur Singla has built a distinguished career out of recognizing those moments early.

At Aruba, later acquired by HPE, he helped shape the enterprise Wi-Fi wave before wireless became the default way businesses connected people and devices. At Juniper Networks, he worked on Ethernet fabrics as data centers became the backbone of the cloud. With Contrail, acquired by Juniper, he founded one of the defining software-defined networking companies of the cloud era. With Volterra, acquired by F5, he moved into distributed cloud services before enterprises had fully absorbed what the edge would become.

The pattern is hard to miss: wired to wireless, hardware to software, core to edge. Each was a structural change in how networks were built. Singla was early to each.

Now, with Lumilens, he is making the same kind of bet again — this time on the connectivity architecture required for large-scale GPU clusters and AI data centers. Lumilens is building a connectivity platform for AI infrastructure that accelerates the transition from electrical to optical connectivity inside the data center. AI infrastructure is forcing that transition into focus, pushing bandwidth, power, and distance requirements beyond what copper interconnects can comfortably support.

Lumilens may be a young company. But the insight behind it is not young-company intuition. It is the accumulated pattern recognition of someone who has helped drive several networking waves before they broke.

Smart Money, Moving Quietly

The capital structure tells the same story in a different language. Mayfield and Spark Capital are not transactional investors. They are infrastructure-focused venture firms with multi-decade track records of underwriting long, hard development cycles in semiconductors, networking, and systems software. Their presence on Lumilens' cap table signals conviction in a five-to-ten-year arc, not a quarterly narrative.

That conviction has translated into one of the fastest unicorn ascents the infrastructure sector has seen in recent memory. Lumilens crossed the $1 billion valuation mark in under a year of founding — an unusual pace in any environment, and a particularly rare one in deep-tech hardware, where long development cycles, capital intensity, and manufacturing risk typically slow the path to mature valuations. The trajectory says something about both the team and the moment: investors are not waiting for the photonics transition to become obvious. They are pricing it now.

Young company, seasoned operators, patient capital, and a valuation curve that suggests the market is already adjusting to the architectural shift underway. The combination tells you how some of the people closest to the photonics transition are thinking about its impact and opportunity.

The Physics That Forces the Shift

Copper interconnects move data by pushing electrons through metal. For short links, copper remains efficient, mature, and cost-effective. But at the speeds AI infrastructure now demands — 800 gigabits per second today, 1.6 terabits and beyond on the roadmap — copper’s useful reach begins to shrink. The issue is not simply bandwidth. It is distance, density, and scale.

Large AI clusters are no longer confined to a single rack. As model sizes grow, hyperscalers need GPU clusters to scale horizontally across multiple racks while still behaving like one tightly coupled system. That requires moving enormous volumes of data between GPUs, switches, servers, and racks with low latency and high reliability. Copper can support short, high-speed links, but as speeds rise, it becomes harder to extend those links across racks without signal integrity challenges, bulkier cable plants, and tighter physical constraints.

Photonics addresses the problem by changing the reach equation. Light moving through optical fiber can carry high-bandwidth signals farther across the data center without the same distance penalties that limit copper at higher speeds. Cable plants become lighter, thinner, and easier to route. Most importantly, optics allow AI clusters to scale beyond copper’s physical reach envelope, connecting multiple racks into larger GPU fabrics without forcing every critical link to stay within short copper distances.

This is not a marginal upgrade. It is an architectural requirement. The next generation of AI infrastructure depends on larger, denser, more distributed GPU clusters — and those clusters need interconnects that can span racks without compromising bandwidth or system design. The industry is not moving toward photonics because copper disappears. It is moving toward photonics because copper alone cannot support the reach and scale hyperscalers now require.

That is the demand-side pressure pulling capital, talent, and partnerships into the photonics stack — and the pressure Lumilens and POET were founded to address.

The Solution: A Rare Dual-Architecture Play

Lumilens describes itself as building “a connectivity platform for AI infrastructure.” That phrase does more work than it first appears.

The company designs, develops, and manufactures near-package optics, co-packaged optics, and pluggable optical transceivers, supported by its own silicon photonics, mixed-signal ICs, electrical-optical interposers, and optical systems. The breadth is the point. Lumilens is not betting on a single optical form factor. It is building across the interconnect stack as AI data centers move through multiple architectural phases at once.

Today, the dominant optical interconnect is the pluggable transceiver — a self-contained module that plugs into a switch faceplate or GPU tray. Pluggables remain essential to scale-out networks, where clusters expand across rows, racks, and data halls. They have served the industry well for more than a decade and will continue to matter. But as bandwidth climbs from 400G to 800G, 1.6T, and beyond, the limits of faceplate-based optics become harder to ignore.

That is pushing optics closer to the compute silicon. The first step is near-package optics, which brings optical engines closer to the GPU or switch ASIC while preserving some modularity. The next step is co-packaged optics, where optical engines sit directly alongside the ASIC package itself. Together, NPO and CPO represent the scale-up transition: the move toward tighter, denser, lower-latency connectivity inside the AI cluster.

Few companies in photonic interconnects can credibly play both sides of that transition — serving the scale-out architectures AI data centers rely on today while engineering the scale-up architectures they will require tomorrow. Lumilens is one of them.

That positioning is precisely why the $POET partnership makes strategic sense.

Why POET Is the Right Counterparty

POET Technologies has spent more than a decade developing its Optical Interposer and refining its manufacturability — a wafer-scale integration platform designed for the dense, co-packaged architectures the AI buildout now requires.

As Lumilens was advancing its own electrical-optical interposer, or EOI, it found in POET a partner that could accelerate its roadmap and strengthen execution against a technically demanding problem. Over the past year, that partnership has evolved into a highly manufacturable EOI solution that fast tracks Lumilens’ ability to deliver its unique silicon, mixed-signal ICs, and optical system IP into a tightly integrated platform for AI connectivity.

The result is not a simple component-level collaboration. It is a deeper effort to solve the packaging, system-design, and manufacturing challenges that emerge as optics move closer to GPUs and switch ASICs.

The pairing is strategically clean. POET extends its proven optical interposer and wafer-scale manufacturing, while Lumilens contributes the AI infrastructure connectivity roadmap, custom silicon, and systems architecture needed to turn that integration into products for scale-out and scale-up AI data centers.

This is the structural story underneath the headline. The integration challenge in photonics — combining silicon photonics, advanced packaging, electrical-optical interfaces, and systems-level engineering — is too broad for narrow point solutions in a market moving this quickly. Hyperscalers are evaluating combinations of platform technologies and systems specialists that can move faster together than either could alone.

The Market Beneath the Deal

The photonic interconnect market is valued at roughly $30 billion in 2026 and projected to exceed $100 billion by 2030, driven almost entirely by the AI data center buildout. Inside that market, the strategically important shift is the move from pluggable transceivers toward near-package and co-packaged optics that reside much closer to the GPU and switch ASIC. It is, in plain terms, the single most consequential transition in data center networking in a generation.

A market growing this fast is not an opportunity for any single winner. It is an opportunity for an ecosystem — and ecosystems are built through exactly the kind of complementary pairings that POET and Lumilens represent.

The question for the rest of the industry is no longer whether photonics will replace copper in the AI data center. That answer is settled. The question is which companies will build the platforms, partnerships, and manufacturing capacity to meet hyperscaler demand on the timeline the market requires. In San Jose, on a two-year-old unicorn backed by Mayfield, Spark Capital, and other venture investors, and now publicly aligned with a decade-old photonic platform on the Nasdaq, that question is starting to answer itself.

[toccodolce] [X] $poet 공급계약 떴으니 이제 떡상할까? 1. $aaoi는 이미 검증된 제조 시설과 기존 고객사를 보유한 상태에서 마이크로소프트(M...

toccodolce — 5/14/2026 9:27:32 AM

[X]

$poet 공급계약 떴으니 이제 떡상할까?

1. $aaoi는 이미 검증된 제조 시설과 기존 고객사를 보유한 상태에서 마이크로소프트(MS) 등과 대규모 공급 계약을 맺고 곧바로 매출을 일으키는 구조인 반면에,

POET & Lumilens는 현재는 초기 확정 주문($50M)을 받은 상태로,

실제 타임라인은 2026년…
— SoFire (@Sofigoodboy) May 14, 2026

[/X]

Translation:

$poet With the supply contract announced, will it moon now?

1. $aaoi already has verified manufacturing facilities and existing customers, and has signed large-scale supply contracts with Microsoft (MS) and others, immediately generating revenue.

In contrast, POET & Lumilens are currently at the stage of having received an initial confirmed order ($50M), with the actual timeline targeting sample release in the second half of 2026 ? full-scale mass production in 2027.

In other words, there's still a process left where, for about 1-2 years from now, they need to work hand-in-hand to secure the product's reliability.

2. The standout point in this news is that POET has given Lumilens the right to purchase about 23 million shares. This goes beyond a simple customer relationship—it's a combination of Lumilens's demand that "we'll buy your stuff in bulk, so let's share the profits if your stock price rises later," and POET's strategy to lock in a solid major customer.

- Ultimately, Lumilens only benefits if POET's technology succeeds, so the partnership intensity is much stickier than a typical purchase contract.

However, as in the recent case where Marvell canceled POET's existing order, just stamping the contract isn't the end— what matters is actually producing the product and passing the customer's quality checks.

[Selectric II] First I've heard of Lumilens. Have I missed something?

Selectric II — 5/14/2026 9:24:59 AM

[toccodolce] Message 35517424

toccodolce — 5/14/2026 9:16:07 AM

[zax] Nearly +30% pre-market. A positive analyst comment is all it took. :)

zax — 5/14/2026 9:01:19 AM

[toccodolce] I think they are an American company based in Santa Clara, CA. Tocco

toccodolce — 5/14/2026 8:58:57 AM

[Selectric II] Lumilens?

Selectric II — 5/14/2026 8:38:19 AM

[Selectric II] I haven't seen this new COO announcement posted: POET Technologies Announces Ap...

Selectric II — 5/14/2026 8:35:35 AM

I haven't seen this new COO announcement posted:

POET Technologies Announces Appointment of Sandeep Kumar as Chief Operating Officer

Print / Save as PDF

TORONTO, May 12, 2026 — POET Technologies Inc. ("POET" or the "Company") (NASDAQ: POET), a leader in highly integrated optical engines and light sources for AI networks, today announced the appointment of Sandeep Kumar as Chief Operating Officer reporting to the Chief Executive Officer. Dr. Kumar's appointment was effective as of Monday, May 11, 2026. Dr. Kumar was also named as an Officer of the Corporation.

Dr. Sandeep Kumar is an experienced executive in the semiconductor industry. In his last position at Silicon Labs based in Austin, TX where he worked for over 18 years, he was Senior Vice President of Worldwide Operations. He has managed manufacturing teams, supply chain, planning, CAD, process engineering, package engineering, product and test engineering, quality assurance, failure analysis, as well as prototype production and reliability test labs. Sandeep has experience working closely with the design, marketing, sales teams, and customers addressing their geopolitical concerns. Prior to joining Silicon Labs, Sandeep was responsible for worldwide product and test engineering for the storage business at Agere Systems, Lucent Technologies and AT&T Bell Labs working closely with the internal factories in the United States and Asia. As a member of the Bell Labs technical staff, Sandeep was part of the group that designed ISDN U-interface chips, Speaker Phone Codec and Sigma-Delta A/D converters.

Sandeep has a bachelor's in electrical engineering from the Indian Institute of Technology in Bombay, India, a M.S. in Electrical Engineering from the University of Evansville in Indiana and a Ph.D. in Electrical Engineering from Lehigh University in Pennsylvania. Sandeep also attended and completed Wharton's Leadership Development Program at the University of Pennsylvania.

POET's Chairman & CEO, Dr. Suresh Venkatesan commented: "We are delighted to be adding to our C-Suite an experienced semiconductor executive. Sandeep is expected to play a major role in managing global operations in with an immediate focus on bringing proven talent into our manufacturing organization in Malaysia to prepare it for high-volume production."

RSU Grant
In connection with his appointment to the Board of Directors, Dr. Kumar was awarded a total of 410,397 RSUs that will vest at the rate of 1/3 of the total on the first, second and third anniversaries of the grant. The RSUs were granted subject to provisions of the Corporation's 2025 Omnibus Incentive Plan and related agreements between Dr. Kumar and the Corporation.

About POET Technologies Inc.
POET is a design and development company offering high-speed optical modules, optical engines and light source products to the artificial intelligence systems market and to hyperscale data centers. POET's photonic integration solutions are based on the POET Optical Interposer™, a novel, patented platform that allows the seamless integration of electronic and photonic devices into a single chip using advanced wafer-level semiconductor manufacturing techniques. POET's Optical Interposer-based products are lower cost, consume less power than comparable products, are smaller in size and are readily scalable to high production volumes. In addition to providing high-speed (800G, 1.6T and above) optical engines and optical modules for AI clusters and hyperscale data centers, POET has designed and produced novel light source products for chip-to-chip data communication within and between AI servers, the next frontier for solving bandwidth and latency problems in AI systems. POET's Optical Interposer platform also solves device integration challenges in 5G networks, machine-to-machine communication, self-contained "Edge" computing applications and sensing applications, such as LIDAR systems for autonomous vehicles. POET is headquartered in Toronto, Canada, with operations in California, USA, Shenzhen, China, Penang, Malaysia and Singapore. More information about POET is available on our website at www.poet-technologies.com.

Media Relations Contact:
Adrian Brijbassi
Adrian.brijbassi@poet.tech

Company Conta

poet-technologies.com

[da_cheif™] hit over 18 in pre market

da_cheif™ — 5/14/2026 8:34:16 AM

[toccodolce] [X] $POET and Lumilens Ink a Half-Billion Dollar AI Bet POET Technologies ($PO...

toccodolce — 5/14/2026 7:51:51 AM

[X]

$POET and Lumilens Ink a Half-Billion Dollar AI Bet

POET Technologies ($POET) has just secured a landmark deal that could transform it from a niche innovator into a foundational layer of the AI stack.$POET has signed a supply and joint development agreement with Lumilens, a… https://t.co/dE1rs4tOup pic.twitter.com/jglHRTrV9U
— Finn Stockinger (@FinnStockinger) May 14, 2026

[/X]

[X]

$POET

“In AI infrastructure, tiny bottlenecks with NICHE product can MARVEL giants like MRVL.” pic.twitter.com/shZK6fBvr3
— VenkatP (@VenkatP1359) May 14, 2026

[/X]

[toccodolce] POET Technologies and Lumilens Advance Wafer-Level Photonic Integration for Next...

toccodolce — 5/14/2026 7:38:55 AM

POET Technologies and Lumilens Advance Wafer-Level Photonic Integration for Next-Generation AI Optical Networks

POET Technologies and Lumilens Advance Wafer-Level Photonic Integration for Next-Generation AI Optical Networks

POET Technologies Inc.
Thu, May 14, 2026 at 7:00 AM EDT

Joint development and sale of high-speed optical modules based on the Electrical-Optical Interposer (EOI) -- a new paradigm for scale in the optical layer of AI compute

SAN JOSE, Calif., May 14, 2026 (GLOBE NEWSWIRE) -- POET Technologies Inc. ("POET" or the "Company") (NASDAQ: POET), a leader in highly integrated optical engines and light sources for AI networks, and Lumilens Inc. ("Lumilens"), an emerging leader in high-performance scale-up and scale-out optical interconnects for AI workloads, today announced they have entered into a supply agreement that establishes a strategic joint development and commercial technology partnership to advance a new class of wafer-level photonic integration for frontier AI infrastructure.

Modern AI computing was made possible by successive leaps in wafer-level integration: first 2.5D electrical interposers that brought GPUs and HBM into a single package, then hybrid bonding that enabled today's HBM stacks and 3D logic. The optical layer, now the defining bottleneck for scaling AI, has not made that leap. At the center of the POET/Lumilens joint development program is a new paradigm for integration and module fabrication – the Electrical-Optical Interposer (EOI) – combining alignment-free wafer-level optical engine production with next-generation optical chipsets and advanced manufacturing capabilities, all to meet the ever-growing demands for scale and performance in AI infrastructure.

POET and Lumilens have structured the joint EOI platform to deliver:
? Enhanced performance & density, enabling compact, high-bandwidth solutions for next-generation AI and data center architectures
? Active-alignment-free manufacturing, replacing the single largest cost, yield, and throughput constraint in optical engine production with wafer-scale processing
? Capital-efficient, high-volume manufacturing, allowing optical supply to scale with hyperscaler GPU fleet growth rather than against the labor-bound limits of conventional optical assembly

The supply agreement between Lumilens and POET establishes a commercial framework to support the joint development program, and, as part of that framework, Lumilens has placed an initial purchase order with POET for the manufacturing of EOI-based engines valued at $50 million. This purchase order represents the first phase of a broader supplier relationship that could scale to $500+ million in cumulative purchases from POET over five years.

In connection with the supply agreement, and to align both companies to the long-term value created through the partnership, POET has granted Lumilens a warrant to purchase up to 22,921,408 common shares. The warrant is immediately exercisable for 2,292,140 shares, with the remaining shares vesting and becoming exercisable in tranches based on cumulative payments by Lumilens toward future purchase orders totaling up to $500 million. The warrant is exercisable over nine years at an exercise price of $8.25 per share.

"GPU interconnects are emerging as the defining bottleneck for scaling AI, and addressing it requires rethinking the full optical stack — silicon, photonics, and packaging — together," said Ankur Singla, CEO and Founder of Lumilens. "This joint development partnership with POET combines Lumilens' next-generation chipset and advanced optical manufacturing capabilities with wafer-level photonic integration to build a clear path to the performance, density, and economics that frontier AI deployments will demand over the next decade."

“Our focus has always been on redefining the integration paradigm in photonics,” said Dr. Suresh Venkatesan, Chairman & CEO of POET Technologies. “This new EOI platform will allow us to jointly bring semiconductor-style manufacturing discipline to optical engines — delivering precision, scalability, and cost structure advantages that are essential for AI infrastructure at scale. Working jointly with Lumilens enables us to translate these capabilities into high-volume production and end-customer deployments for the next generation of AI data centers.”

Together, POET and Lumilens will deploy the joint platform across a multi-year roadmap: from 800G and 1.6T pluggable transceivers to next-gen high-density Near-Package Optics (NPO) and Co-Packaged Optics (CPO). Fulfilment of the purchase orders and associated revenues is subject to the successful development and ultimate qualification of the modules, as well as the successful scaling of manufacturing capability. Engineering samples from this joint development program are expected in late 2026, with production ramp aligned to hyperscaler customer deployments in 2027.

About Lumilens
Lumilens is building the connectivity platform for AI infrastructure, delivering scale-up and scale-out optical interconnect solutions designed for the performance, power, and bandwidth demands of next-generation AI workloads. The company designs, develops, and manufactures near-package optics, co-packaged optics, and pluggable optical transceivers for hyperscale AI data centers. With its own silicon photonics, mixed-signal ICs, electrical-optical interposers, and optical systems, Lumilens enables tighter integration, higher bandwidth density, lower power consumption, and greater flexibility across AI network architectures. Lumilens is backed by Mayfield, Spark Capital, and other leading venture capital investors. For more information, visit Lumilens.com.

[toccodolce] [X] $POET signed a supply and joint development deal with Lumilens for AI optic...

toccodolce — 5/14/2026 7:17:24 AM

[X]

$POET signed a supply and joint development deal with Lumilens for AI optics, including an initial $50M order for optical engines. The partnership could scale to $500M over five years, with samples expected in late 2026 and production ramp planned for 2027. pic.twitter.com/xcop2fz9v3
— Wall St Engine (@wallstengine) May 14, 2026

[/X]

[w0z] Yes the link to Lessengers hit agoracom.com last night at 12:45 AM. Since it's ...

w0z — 5/11/2026 1:27:09 PM

Yes the link to Lessengers hit agoracom.com last night at 12:45 AM. Since it's behind a Korean paywall I don't think the market has fully absorbed it yet. There's a lot of discussion on agoracom but it's spread over hours of garbage posts.

[toccodolce] [X] Bulk Silicon Beats SOI: Why POET’s Optical Interposer Has the Cost Edge 1/...

toccodolce — 5/11/2026 12:24:09 PM

[X]

Bulk Silicon Beats SOI: Why POET’s Optical Interposer Has the Cost Edge

1/ There's ongoing debate over what $POET 's Optical Interposer is actually built on. Much of the industry assumes some SOI variant is a must for silicon photonics — conventional wisdom says bulk silicon…
— Convequity (@convequity) May 3, 2026

[/X]

[X]

$POET Korean news came out and POET has invested in Lessengers. Asian market demand for 1.6T. Korean VCs like the look and are partaking too. News is behind a Korean paywall and has not hit the western media wires yet. It's legit. https://t.co/h32BAAUXTh
— Tom Albers (@TomAlbers264982) May 11, 2026

[/X]

[toccodolce] [X] $SIVE$LWLG$POET The AI infrastructure supply chain is evolving into a tight...

toccodolce — 5/11/2026 10:28:26 AM

[X]

$SIVE $LWLG $POET
The AI infrastructure supply chain is evolving into a tightly integrated photonics ecosystem, and the strategic alignment between Sivers Semiconductors, Lightwave Logic, POET Technologies, and SilTerra may become one of the most important architectures behind… pic.twitter.com/ndeGhXcIgM
— Pep Invest (@PepInvestStocks) May 9, 2026

[/X]

Rumor circulating on X that $QCOM wants to buy a stake in $POET.

[toccodolce] Latest from RDDT: As I received feedback, I promised I'd make this post reg...

toccodolce — 5/10/2026 12:28:29 PM

Latest from RDDT:

As I received feedback, I promised I'd make this post regarding the Marvell incident.

First, I must say, this is my POV. I'm not a professional analyst, so make your own conclusions.

I will separate this into four blocks and a brief conclusion.

BLOCK 1. THE ORIGINS AND THE RESHAPE

Celestial AI was a company founded in 2020 that developed a systems architecture for light that is called Photonic Fabric. This took them years.

To prove this lights architecture concept, Celestial AI needed partners to validate their idea (because this was like a hypothetical highway of light).

There were more, but basically the chain was:

Lasers (Lumentum/Sivers): Generate photons.
Optical Engine (POET): Organizes and couples light to the system.
Photonic IC (Imec/Celestial): Modulates light to transmit data.
HBM Memory (Samsung): Stores the data flowing through the tissue.
Control ASIC (Broadcom & TSMC): Manages all rack operations.

POET Technologies was key to validating the Photonic part. POET’s Optical Interposer solved the packaging nightmare by allowing lasers to be flip-chipped and aligned with high precision at a low cost. Without POET, Celestial AI would have struggled to build a prototype that was small enough and reliable enough to show to hyperscalers.

A major reason Celestial AI could claim their tech could work inside hot AI racks was POET’s Starlight engine. It allowed the lasers to be placed slightly away from the high-heat ASIC/GPU while still coupling the light efficiently. This separation was the proof that Photonic Fabric wouldn't melt in a real data center, this means, the technology would be manufacturable and thermally stable for AI data centers.

POET provided a plug-and-play light source. This allowed Celestial AI to focus entirely on their 16 Tbps modulation and routing architecture rather than spending years trying to become laser packaging experts themselves.

This is what the chain looks like after Marvell dumped POET and their adquisitions:

Lasers (Lumentum/Sivers).
Optical Engine (Marvell In-House): The 1.6T Light Engine that organizes and couples light into the system (replacing POET).
Modulators (Polariton/Marvell): Ultra-fast plasmonic modulators that convert electrical signals to light at 3.2T speeds with minimal power.
Photonic IC (Celestial AI/Marvell): The core Fabric architecture that manages optical data routing between chips and boards.
HBM Memory (Samsung/ SK Hynix/ Micron).
Control ASIC (Marvell Orion): The central brain (logic and switching) that orchestrates all rack operations and data traffic.

BLOCK 2. THE ISSUE

What's the deal here?

Marvell absorbed an IP (Celestial AI Photonic Fabric) that was proven to work with POET tech.

POET spent years working with Celestial AI to figure out exactly how the light needs to enter the Fabric (angles, power levels, thermal tolerances).

When Marvell bought Celestial AI, they bought all that joint research and contracts established between POET and Celestial AI. Marvell used POET as a free R&D lab to validate that the Fabric was commercially viable. Once Marvell saw the map, they realized they could try to build their own version by doing it themselves.

The NDA violation in April provided the perfect legal excuse to cancel the POET contracts without paying massive termination fees.

By dumping POET, Marvell was ripping a functional working system and replacing it with a different architecture yet to be proved.

Why? Because you can't just swap an interposer like a Lego brick. The original Photonic Fabric used POET's passive alignment and specific light-coupling angles.

Marvell is basically building a Photonic Fabric 2.0. Integrating their ecosystem. But Marvell is using a monolithic approach. This doesn't work with Photonic Fabric 1.0 (the original one).

By using Polariton’s plasmonic modulators, Marvell prints the light-handling parts directly onto the silicon at a much smaller scale. It doubles the speed (3.2T) while cutting the physical footprint.

Some analysts say the shift likely pushes full-scale production back by 12 to 18 months. This is why Marvell's 2026 updates focus on sampling and lab demos rather than the commercial shipping Celestial AI had originally promised for early 2026.

People don't realize Marvell is doing a big gamble. Marvell is betting that their plasmonic approach at the end of 2027-start of 2028 will be much better than the glass hybrid tech. They are sacrificing 2026-2027 revenue to try to own a big part of the future photonics pie.

(I assume that if you are a Marvel investor, you know what this means.)

Marvell didn't want to be dependent on a small company like POET, or the most critical component of their AI strategy (vertical integration) would collapse.

If you paid attention to the previous paragraphs, yes, you're right, plasmonics is an experimental tech. that has to work under a lot of heat and pressure (the solutions proposed are yet to be proved in the manufacturing stage).

Hyperscalers want reliable chips with thousands of hours of productive life cycle, Marvell won't be able to assure it in the short term regarding their photonics department.

POET Tech is proven. In 2026 other giants are doing another glass hybrid or mixed approach too.

Polariton’s Plasmonics is faster (3.2T), but it is much harder to manufacture at scale. It requires exotic materials (like specialized organic polymers) that aren't standard in silicon foundries.

BLOCK 3. THE RIGHT APPROACH?

Glass is a natural conductor of light. Silicon and plasmonics aren't.

Light travels through glass with almost zero resistance, it is a more efficient, natural medium for data. Plasmonics, which Marvell acquired through Polariton, forces light to interact with metal surfaces, which creates heat and signal loss.

By dumping the natural glass-based path of POET for the forced metallic path of Polariton, Marvell is making a massive high-risk bet, as I explained earlier.

The POET Way (Glass): Light stays in its natural state. It’s cool, stable, and proven. This is what allowed Celestial AI to validate the Fabric because it didn't overheat the chips.

The Marvell/Polariton Way (Plasmonics): It squeezes light into spaces smaller than its own wavelength using metal. This allows for 3.2T speeds in a tiny footprint, but it is fighting physics. The metal causes high insertion loss, meaning you need much more powerful lasers (Sivers/Lumentum) to punch the signal through.

POET's Interposer was built to be compatible with standard CMOS foundries.

Plasmonics requires integrating metals like gold or silver and specialized organic polymers directly onto high-end silicon.

This is a nightmare for mass production and can contaminate standard semiconductor lines. Marvell is betting they can solve a manufacturing problem that the rest of the industry has avoided for years.

By abandoning the POET/Celestial glass-based architecture, Marvell:

-Has to re-engineer the optical tissue to handle the heat and loss of the plasmonic modulators.

-Increasing Power Demand. Because plasmonics is lossy, the total system power might actually go up initially, potentially negating one of the original selling points of the Photonic Fabric.

Marvell’s leadership believes that silicon photonics (glass) will eventually hit a size wall. They are willing to break a working system today to try to be a company with a dominating system tomorrow based on Plasmonic.

4. PROS AND CONS

Pros

-Marvell's logic is that plasmonics allows them to squeeze light into spaces 100x smaller than glass allows (but take the 100x as a maybe, due to lasers/thermal management risks). Marvell is betting that glass is too big. They believe the future of AI is scale-in (everything inside one tiny chip), they think glass is simply too fat.

-While light moves fast in glass, modulating it (turning it on and off to create 1s and 0s) requires energy. To reach 3.2T or 6.4T speeds using glass, traditional modulators have to become very long (several millimeters) or use a lot of power to force the light to change.

-Plasmonics uses electrons on a metal surface, it can switch at terahertz speeds in a space of just a few microns. Marvell is betting that the energy they save in switching will make up for the energy they lose in friction (heat).

Cons

-Plasmonics generates localized hot spots on the chip because metal absorbs light. Glass stays cool.

-Plasmonic heat is a serious risk for the longevity of the chips in advanced stages. Energy efficiency is a huge con yet to be solved. Glass and solutions like POET tech promise a huge consumption reduction for the current generations.

-Worse conductor than light. Light and silicon simply aren't made to cooperate. It's a forced union.

-Contamination issues and organic polymers are yet to be tested in real loop cycles.

-If the glass becomes the standard, the big minds will find a way to reach higher speeds in a natural medium, they always do. Hybrid Bonding and 3D glass could be a more reasonable approach.

The industry is splitting right now. Which standard will prevail? Or will they be able to coexist?

CONCLUSIONS

I can be wrong, but when I see this case, I can't stop thinking about this.

I remember a part of the book Chip War. It said something like this:

Most people don't know, but a long time ago, near the 50s, the first transistors used germanium (Ge) instead of silicon (Si). Germanium was a faster medium to move electrons than Silicon. But, when the silicon approach took form, the germanium rapidly collapsed.

Why did silicon become the standard? Simple, the researchers discovered that silicon was a better medium (more stable and less susceptible to overheating) to move electrons than germanium.

The industry tends to prioritize stability and control over raw and unstable speed. I think Marvell is making the same mistake regarding the light. Time will tell.

Hope you enjoyed it.

[zax] I don't do technical. The setup seemed obvious to me. [graphic] You could t...

zax — 5/9/2026 6:43:23 PM

I don't do technical.

The setup seemed obvious to me.

You could throw a dart at the POET chart and buy during any rapid upturn, sell quickly and make a significant % gain.

RGTI was hit much harder in the market downturn because it was much more hyped. But it is drifting up again with the market. Another pop like the last one, would be a double in days. And the money is drifting back into AI after a brief market disillusionment / pullback. The bubble is far from over. Set your stops, take your losses, and it is easy money. What a wonderful time to be trading.

[zax] I think RGTI is setting up similarly.

zax — 5/8/2026 2:33:21 PM

[benwood] That's a pretty good dead cat bounce!

benwood — 5/8/2026 1:57:45 PM