Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
I left him 10% of my holdings hoping for a Hail Mary for him in trust. He is a teen.
Enjoy your scotch that’s some good stuff.
Cheers
Yes I said all that and I hope I am proven to be wrong. I could not believe the management would allow us to almost go bankrupt.
I hope they do find some sustainable success but I would keep my eye on the foundries and their Pic plans.
My grandson is having fun with his eventual Poet shares reaching all time highs.
I was out last August on unrelated management and stock price issues.
Good luck!
You are well read and up to date as usual! 😂
It remains to be seen if Poet will become a significant player in the photonic packaging space.
Cheers
Hi KCC,
Most silicon interposers used in the electronics industry have just used TSV’s to connect electricity to various chips on the substrate.
It’s when silicon interposers start to combine TSV’s with photonic capabilities that it gets more complex.
Here is an example from a paper showing silicon micro ring photonics being used In a photonic interposer application.
Poet as you know uses a different design.
Which one is cheaper and the most effective.
? I sure don’t know.
Cheers
https://cea.hal.science/cea-04225225/document
Enjoy that scotch!! It’s real! All that matters!
This other dreaming of winning it big on penny tech stocks was always a very long shot.
There is nothing Poet has that others are not doing just a different method. Whatever Poet says they have not signed up the big transceiver OEMs. Ok Adtran, Luxshare and a couple of tier 2
Players. No more interest in their light bar products.
Sorry not much hope Jmo.
I said it before they are done.
They misssd the window. All the foundries have solutions.
Sorry to say but they are now a salary scam for management and the employees.
It’s fulfilled the CEO’s dream but it’s a complete failure for retail investors.
The can is definitely kicked down the road again.
The bar is open 😂
KCC,
Enjoy the scotch!
I sold 90% of my shares at the end of last August due to a serious health issue.
I’ll keep the rest for my grandson.
Live each day large !!
Any idea on who’s laser Poet is using on the 200g per lane demo at the OFC?
The lumentum 200 g with a EML attached would be amazing Jmo.
TFLN?
How does an investor like Mazan back this company over and over?? Is he a preferred share holder? Does he exist? Or more Poet BS.
5 years ago I thought this stock was the real deal.
It’s been a death spiral after the split. Ajo
Macom has their own patent for cable equalization.
I don’t know if it infringes Spectra7 patents.
https://patents.justia.com/inventor/naga-rajesh-doppalapudi
Thanks for sharing KCC!
I see Astera Labs is bringing a Smart Cable Module to the Market.
It’s a version of an electrical cable to extend copper to 7 meters.
They also just went public.
Lots to choose from jmho.
There is still built in the need to use optical transceivers. Especially if longer distances are involved in connecting the servers.
DAC’s ACC’s and optical transceivers can be used. Just saying. From your link.
1) GB200 NVL72 inter-cabinet connection (external cable): large data centers often require a large number of cabinets for parallel computing, if the cabinets need to be networked externally, they are connected through the TOR switch with a DAC/AOC cable (as shown in Figure 10). For a large number of cabinets, the external interconnection needs to be installed in the cabinet above the cabling equipment for orderly connection, cable length is often long, copper cable is more than 2 to 4 meters after the connection requirements can not be met, so the long-distance interconnection of the main use of fiber-optic cables to connect, copper cables in this link can not completely replace the fiber-optic cables.
From OFC a general update.
The OIF released a document on the next generation electrical interface two years ago.
Pg 32 deals with copper connections. ACC’s are discussed.
If Sev has the capability to do 224g per lane they would have announced it at OFC 2024. Jmo
Maybe later this year jmo
https://www.oiforum.com/wp-content/uploads/OIF-FD-CEI-224G-01.0.pdf#page31
It’s the DML tx engine based on 100g per lane. The 200g per lane EmL tx will take to the end of 2024 to get samples out.
Sounds like 800g transceivers will generate direct sales to Poet if I read the CEO’s comments correctly.
Still wonder how the China plus one strategy fits into this complete transceiver solution plan.
Don’t forget nvidia uses Macom linear equalizers in the custom linear pluggable transceivers in the over 5 meter applications.
But we as we all know Macom is using Spectra7 patents to drive nvidia AI connectivity solutions
Should be $50 a share soon.
We are just early.
It’s going to be awesome!!!😎
This patent is assigned to Macom
ADAPTIVE CABLE EQUALIZER
Publication number: 20220263540
Abstract: A cable equalizer configured as part of a cable comprising a first stage, a second stage, and a third stage. The first stage comprises a first stage bias current circuit configured to generate a bias current and a pre-emphasis module configured to introduce pre-emphasis into a received signal to counter the effects of signal amplification. Also part of the first stage is a bias voltage circuit configured to provide a bias voltage to the first stage. The second stage comprises a buffer configured impedance match the first stage. The third stage comprises a third stage bias current circuit configured to generate a bias current and a tank equalizer circuit configured to perform frequency specific equalization on a second stage signal. An amplifier is configured to amplify the second stage signal to create an amplified signal, which is output from the cable equalizer by an output driver.
Type: Application
Filed: February 12, 2021
Publication date: August 18, 2022
Inventors: Quazi Ikram, Naga Rajesh Doppalapudi, Saman Jafarlou
https://patents.justia.com/inventor/naga-rajesh-doppalapudi
226 Gbps per Lane Active Copper Cable Equalizers: MACOM will host a live demonstration of its four channel, 226 Gbps per lane, linear equalizers supporting active copper cable applications up to 1.6TB. MACOM’s new linear equalizer product line can extend a typical copper cable’s performance to 2.5 meters at 226 Gbps data rates, allowing hyperscale cloud and Internet Service Providers to consider active copper cables (ACC) as an alternative to direct attach copper (DAC) cables or active electrical cables (AEC) in high data rate applications.
https://ir.macom.com/news-releases/news-release-details/macom-showcase-200g-lane-products-optical-fiber-communication/
Lots of information that may help understanding the various interconnect solutions.
https://www.naddod.com/blog/comparing-dac-aoc-cables-vs-dsp-lpo-optical-modules-for-data-center-network-interconnects
Yes he did some great investigative work to dig it up.
We can only hope at this time the Blackwell architecture will use the ACC’s as well as passive DAC’s.
What’s coming.
“We expect shipments of 8x100G optical transceivers to exceed 5 million units in 2024 and first 8x200G modules delivered to customers by the end of the year,” commented Dr. Vladimir Kozlov, founder and CEO of LightCounting Market Research. “Google and Nvidia will be the first adopters of 200G per lane optics for interconnecting GPUs and TPUs in AI Clusters – the hottest area of the market now. Broadcom is once again among the first suppliers of components enabling the next generation of optical transceivers.”
https://www.broadcom.com/company/news/product-releases/61941?utm_source=social&utm_medium=linkedin&utm_campaign=ai-oc
“Enterprises continue to demand larger AI clusters, elevating the importance of cutting-edge optical interconnects,” said Craig Thompson, vice president of LinkX products at NVIDIA. “NVIDIA is at the forefront of photonics innovation, and Broadcom has been an important optical-component partner, matching the pace and scale required as we advance our HPC and AI optical-interconnect technology.”
I stopped asking over a year ago as the CFO told me I ask too sensitive questions.
I believe Poet is in a very risky state. But my friends call me captain obvious 😂
Broadcom and their Ethernet switch tomahawk 5 with their custom switch Serdes extends passive DAC’s reach up to 4 meters.
I’m sure Nvidia and their Serdes in their switches would do the same Jmo.
We still don’t know the lengths of the copper interconnects in the Blackwell setup.
From Broadcom,
“While the IEEE specification for 100G PAM4 corresponds to a reach of roughly 2 meters for 26AWG DAC, many SerDes on the market struggle to achieve even that. Tomahawk 5, in contrast, provides excellent performance for twice the IEEE standard, up to 4 meters of passive DAC. This means that all devices within a data center rack, and in some cases even between racks, can be interconnected with passive copper. For customers that need shorter cables, finer-gauge DAC can be employed, which improves the cost, weight, and bend radius of the cables.
- Jensen says the algo for server GPU design should be to build the biggest chip litho allows and then connect two of them together. Beyond that, it’s best to use NVLink copper cabling to connect as many GPUs in a system as possible, modulated by power and cost (GB200 clusters support up to 576 GPUs connected via NVLink). Qualifies this by saying optics are still needed beyond that and that optics/retimer demand remains high
A few notes from $NVDA's analyst Q&A:
— Eric Jhonsa (@EricJhonsa) March 19, 2024
- Jensen: Most of our customers have known about Blackwell for some time. Yet Hopper demand hasn’t changed.
- Colette: Blackwell coming to market “late this year.” Suggests supply constraints are likely early on (fits with prior comments).
-…
The other thing that comes to mind is who is making the lightbar products? They are built on the interposer technology at Silterra but they are not referred to in the agreement??
I’m hoping I’m right! lol otherwise Poet gets no direct revenues from the transceiver business in the telecom or data centers going forward.
Up to 3.2t it’s based on 400g optical engines.
The only revenue opportunity is in light sources.
If a quote comes in to Poet from a really big transceiver player outside of China, Superphotonics may say that’s a schedule n company.
No direct sales to Poet but it goes to the joint venture.
It maybe why Poet wants a China plus one strategy where poet makes a subsidiary and makes the complete transceiver outside China including the optical engines.
All with Superphotonics approval.
I think revenue to Poet will be based on the data speed not the underlying optical engines that allows the data speed. Otherwise Poet will never get revenue from the data centers in the transceiver market. Jmho
On page 15 the agreement states Poet can sell 400g outside of China but they introduce schedule N. at the time of signing there were no named customers on that list. It’s found on the last page of the agreement. Whatever that means. It Could allow Superphotonics to cheery pick a big client outside China.
All 400g optical engines will be built by Superphotonics regardless of end users market except for the following, if an end customer does not want product produced in China, Superphotonics can outsource to a contract manufacturer or to a subsidiary of Poet.
I believe Adtran will demand that from Poet due to the BaBa requirement from the Us government.
No transceivers made in China Jmo.
If Poet is building the capability to handle all types of lasers including VCSEL’s then they can supply the multi mode type of transceivers as per the MSA.
https://terabit-bidi-msa.com/
https://agoracom.com/ir/POETTechnologies/forums/discussion/topics/799475-poet-linkedin-2/messages/2407730#message
I don’t think anyone knows what revenue is going to be booked in 2024.
I still can’t get over no revenues for 100-400g transceivers goes to Poet.
800g and above Poet will get revenues is my understanding. The lightbar product I believe generates revenue to Poet as well.
https://investorshub.advfn.com/boards/read_msg.aspx?message_id=173364425
Don’t forget Luxshare and Spectra7 are partners!!
Macom are into everything as well but each year they show different technologies.
https://www.prnewswire.com/news-releases/luxshare-tech-preview-for-ofc-2024-collaborating-with-ecosystem-partners-to-explore-cutting-edge-technologies-302087509.html
Don’t forget Luxshare and Spectra7 are partners!!
Macom are into everything as well but each year they show different technologies.
https://www.prnewswire.com/news-releases/luxshare-tech-preview-for-ofc-2024-collaborating-with-ecosystem-partners-to-explore-cutting-edge-technologies-302087509.html
I wonder if Luxshare could jam into their new facility a module making capability for Poets future plans?
They seem to be doing ok with their relationship with Apple Jmo.
https://theinvestor.vn/news/apple-supplier-luxshare-ict-looks-for-1000-new-workers-in-central-vietnam-d8762.amp
Hi Proto,
Although not specifically stated I’m assuming as the bandwidths get higher these companion redrivers chips can be added to the switch ASiCs
To compensate for the signal loss.
I’m not sure how far Broadcom can take these redrivers chips as the bandwidths increase over time.
That maybe where optical I/O like what Ayar Labs and others come in. The chip to chip and chip to remote servers from 2 nm to 2 kms. At terabit speeds.
Cheers
“With each doubling of the bandwidth, the distance that a PCI-Express signal can be driven down a copper wire is cut in half. The retimers are used the extend the length of the copper wire; the longer the wire, the more retimers you need. Because of latency issues and the fact that PCI-Express is used as an extension of the CPU bus, you try not to let those wires get too long. But if you want to stretch PCI-Express fabrics across several racks, or even an entire row of machinery – as we here at The Next Platform want the industry to do – then retimers are going to be needed more and more as the PCI-Express bandwidth goes up.
The switch ASICs from Broadcom like the Tomahawk 5 that use these next generation redrivers extend the copper interconnects.
As speeds/bandwith increase copper DACS will get shorter. The redrivers extend the reach of a DAC cable from 2 meters to 4 meters.
In my opinion Spectra7’s ACC’s can only benefit from the improvement in Broadcoms switches.
Maybe the company could comment. I’m not a network engineer just someone who likes reading about the technology.
https://www.nextplatform.com/2024/03/06/pushing-pci-express-switches-and-retimers-to-boost-server-bandwidth/amp/
Great catch!!
It is and Arista switch ASICs are based on Tomahawk designs.
There may have been a technological licence transfer agreement done at some point.
I don’t see Arista promoting ACC’s, I have found them promoting DAC’s and AOC’s and transceivers.
I’ll keep looking.
COBO never caught on and its founder Microsoft abandoned CPO copackaged optics.
But the initiative lives on under a different acronym.
And we wonder why Poet struggles to advance their technology.
It’s the wild Wild West.
https://www.yolegroup.com/industry-news/the-apcs-blueprint-for-silicon-photonics/
Poets optical interposer is a new and evolving proprietary technology that allows the electrical and photonics components to function on the same chiplet.
It’s still being proven as they introduce different components to the platform.
They never discuss how many electrical components the platform can handle or the maximum overall size it can be applied to.
Look at celestial AI drawings and there are many components involved to make the photonic fabric work. They say there is an interposer involved but not poets. We would have heard if poet was involved in that technology.
It means to me there are limits to what the platform can handle.
Also all the large foundries have or are developing different packaging technology doing the same thing poet has developed only using different techniques.
A buy out?? Let Poet keep proving their technology and build revenues. It’s a safer way to see if it will ever scale and maybe become a part of future standards. Right now it’s an interesting technology with no real major league demand.