YT YouTube2Text

YouTube2Text → Transcripts → 【曲博科技教室 EP324】川湖、鋒霈、英特爾、友達、Lumentum、Coherent、群創、昱鐳應材、兆通光電、Corning、大立光

【曲博科技教室 EP324】川湖、鋒霈、英特爾、友達、Lumentum、Coherent、群創、昱鐳應材、兆通光電、Corning、大立光 — Transcript

by 曲博科技教室 Dr. J Class · 3,732 words · 612 segments · language zh-TW · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hello everyone, welcome to the Tech
0:02Classroom. I’m Hsu Po. It’s
0:04Thursday night, our time for tech Q&A.
0:07If you enjoy our show, remember to like
0:09, subscribe, share, and turn on
0:10notifications. First, we must address
0:13the issue of fraud. I will never
0:15discuss individual stock prices
0:16privately with anyone on platforms like
0:18YouTube, Facebook, or LINE, nor will I
0:20ever recommend specific stock
0:22investments. If you receive any
0:23messages from strangers asking about
0:25your stock trades or recommending
0:27individual stocks, that is definitely a
0:29scam group. Please don’t hesitate to
0:31report them. Please spread the word,
0:33and thank you all for your assistance.
0:35Now, first, a quick commercial break.
0:38Even though it's a bit late, I wanted
0:40to mention that this university has an
0:42Industrial Education Department. This
0:45is basically a department related to
0:47the industry. Their master's and
0:49doctoral programs for admissions via
0:51application—I’m sorry to say the
0:53registration deadline has already
0:55passed, as it was October 7th. Today is
0:58October 8th, so we missed it by just
1:01one day. Because our live stream was
1:03one day late, and there wasn't enough
1:05time to mention it earlier because the
1:07window was only three days. However,
1:10for the next stage, aside from the
1:12application-based entry, there is still
1:14the option of taking the entrance exam.
1:17So, if anyone is interested, you can
1:19consider that. Since this Normal
1:21University, as you know, is a school
1:24that trains teachers, the department
1:26within it that is closest to the
1:28industry is where talent from the
1:30industrial and vocational sectors can
1:32go. I feel it's a department that is
1:35quite close to the industry, and some
1:37of the courses there have a higher
1:39relevance to the industry. Also,
1:41speaking of this university, it
1:43reminded me of a news story I saw by
1:45chance today, so I thought I’d chat
1:47about it with you. It says here that a
1:49graduate from Taipei First Girls High
1:50School and National Taiwan University
1:52—likely a woman—collapsed after
1:54just one month of working as a junior
1:56high school teacher and cried for a
1:57long time when she got home. What
1:59happened? She said she stepped onto
2:01campus full of passion for education,
2:02but she never expected to be driven to
2:04a breakdown by students after only one
2:06month. This is a very interesting
2:07phenomenon, and I’ve been thinking
2:09about it. She said that some students
2:11not only make noise in class and don't
2:13turn in homework, but even
2:14intentionally talk back and provoke her
2:16. So, that made her feel exhausted
2:18every single day. Honestly, my
2:21personality is such that seeing this is
2:23, frankly, very tiring. I actually only
2:27want to teach and share knowledge when
2:29I'm in school; I don't want to manage
2:31people. So, that is why I never wanted
2:34to teach at a junior or senior high
2:36school in the first place. The
2:38advantage of universities and graduate
2:40schools is that everyone is an adult,
2:42so these situations are less likely to
2:44happen. The problem is that middle and
2:47high schools still need teachers;
2:49someone has to teach them. So, I can
2:52only say it comes down to personality;
2:55if you have the patience and can
2:57tolerate that kind of environment, I
2:59think middle and high schools are still
3:01great places to teach. But if you're
3:06like me and really dislike managing
3:08people, then middle and high school
3:10teaching will indeed be more difficult.
3:13You still need passion for teaching;
3:16honestly, that is a very important
3:18concept. Next, someone asked about the
3:21server rail manufacturer King Slide
3:23hitting consecutive limit downs. A few
3:25days ago, as you all know, after King
3:28Slide hit a record high, it plummeted;
3:30this phenomenon is very interesting. I
3:32think the current situation is that
3:34King Slide's performance is excellent,
3:36which is a fact. So, where is the
3:39problem? The problem lies in its P/E
3:42ratio. If you calculate its EPS, it
3:46might fall between 200 and 300 this
3:49year, let's assume 250. Then if you
3:53calculate King Slide's stock price,
3:55you'll find its P/E ratio is around 50
3:58times or more. It's quite obvious, King
4:01Slide makes rails after all. Although
4:05they are very impressive with many
4:07patents and are a creative, unique
4:10company, the question is whether a P/E
4:13ratio over 50 is reasonable. In other
4:16words, a recent phenomenon in the
4:19capital market is that many stocks have
4:21very high P/E ratios. As I've said,
4:25because artificial intelligence is
4:27profitable, it's reasonable for these
4:30companies 'stock prices to be high. But
4:35the current situation is that many
4:37AI-related stocks have irrationally
4:40high P/E ratios, and I think that is
4:43the real risk. Therefore, I believe
4:46this correction for King Slide
4:48essentially corrects its excessive P/E
4:50ratio. The stock market is still in an
4:55uptrend, and the entire market hasn't
4:57seen a major correction yet, so I think
5:00some companies with high P/E ratios
5:02carry risks, mainly because we are at a
5:05relative high point. So, I still
5:09suggest that everyone keep an eye on
5:11whether the P/E ratios of the stocks
5:14you hold are reasonable. When the
5:16market keeps rising and everyone is
5:19chasing it frantically, no one thinks
5:21about this. But when the stock market
5:25has climbed to a very high level, you
5:28might need to start paying attention,
5:30because an irrationally high P/E ratio
5:33is something that I think warrants a
5:35review. Conversely, for some unpopular
5:39stocks that people aren't interested in
5:41, but which have very low P/E ratios,
5:44perhaps when the entire market pulls
5:46back, that might be an opportunity.
5:48Okay, next, this viewer asked about
5:51Feng Pei Environmental Technology. This
5:54company is actually in the water
5:56treatment business, so as you can see,
5:58the first is the wastewater treatment
6:01system; in factories and semiconductor
6:03plants, there is a lot of wastewater
6:05that requires advanced oxidation
6:07treatment and biological sludge
6:09conditioning and reduction. In other
6:13words, it extracts the components
6:15contained within the sewage. This
6:18includes high, medium, and
6:19low-concentration ammonia removal
6:21technology, nitrate-containing
6:23wastewater treatment technology,
6:25inorganic wastewater treatment, and
6:27biological treatment technology, which
6:29uses biochemical reactions to process
6:31wastewater. Advanced ammonia nitrogen
6:33removal technology; so, everyone, this
6:36company is all about water treatment.
6:38And of course, if you do water
6:39treatment, you have the opportunity to
6:41engage in the circular economy, because
6:43photoresist, after being used and
6:45cleaned, inevitably becomes wastewater.
6:47So, how do you recover this photoresist
6:50? That’s where photoresist recovery
6:52equipment comes in, as well as ammonium
6:54sulfate crystallization and drying, and
6:55precious metal recovery, which is also
6:57very important. During cleaning in
6:59semiconductor plants, many precious
7:01metals are washed into the wastewater,
7:03so they need to be recovered. Turning
7:06hydrofluoric acid waste into cryolite.
7:08Everyone knows that hydrofluoric acid
7:10is extremely dangerous; it's called "
7:12bone-dissolving water," and it's
7:14terrifying. If the wastewater isn't
7:16treated properly, it's very difficult
7:18to manage, so this hydrofluoric acid
7:20must be treated. Urban mining means
7:24recovering useful materials from
7:26wastewater or waste gas; regular mining
7:29involves going into the mountains,
7:31while urban mining is essentially
7:34recycling. Basically, I think such a
7:39company certainly meets ESG
7:41environmental requirements, so it
7:43definitely has practical opportunities,
7:46but the key remains what I said earlier
7:48: is the P/E ratio reasonable? In terms
7:52of business, there are indeed quite a
7:54few competitors in wastewater treatment
7:56. But because almost the whole world
7:59has been building plants over the past
8:01two years. And as long as a plant is
8:03built, there will be a need for a large
8:05amount of wastewater treatment
8:07equipment, so this company should have
8:09opportunities. It mainly depends on
8:11whether the price-to-earnings ratio is
8:13reasonable. Speaking of today, frankly,
8:15I am not optimistic about using
8:17MicroLED for multi-mode fiber. This
8:19refers to using MicroLED for optical
8:22communication; as you all know,
8:23long-distance communication has always
8:26relied on lasers. Compared to lasers,
8:28the first problem with MicroLED is that
8:31its light is not pure. Therefore,
8:33impure light entering a fiber optic
8:35cable will cause dispersion, known as
8:38material dispersion, which limits the
8:40transmission distance. Another point is
8:44that MicroLED is not suitable for
8:46high-speed transmission. While lasers
8:50can typically reach 200 Gbps, MicroLED
8:54can likely only reach 2 Gbps, a 100-
8:57fold difference. In other words, using
9:01MicroLED for optical communication
9:04requires a MicroLED array. That is, if
9:07one unit only provides 2 Gbps, you
9:10would need 100 units to reach 200 Gbps,
9:14coupled with the use of multi-mode
9:17fiber. Because of this material
9:19dispersion, its transmission range is
9:22naturally shorter, meaning fiber optic
9:24transmission still relies more reliably
9:27on lasers. Honestly, lasers are indeed
9:30more reliable, but I believe it is
9:32still too early to declare MicroLED a
9:35failure, as many companies worldwide
9:37are currently experimenting with it.
9:42Between edge-emitting lasers and
9:45MicroLED, there is another type called
9:48Vertical-Cavity Surface-Emitting Laser,
9:50or VCSEL. VCSEL could also become an
9:55option for future optical communication
9:58. However, VCSELs are more expensive,
10:01just like edge-emitting lasers. One
10:05possible approach is to reduce the
10:07number of N-type and P-type thin film
10:10cycles in the VCSEL laser; in other
10:13words, don't coat as many layers and
10:16use a simpler resonant cavity to
10:18replace the VCSEL—that is, a
10:20Micro-LED. But by using a simpler
10:23resonant cavity as a replacement, you
10:25can reduce costs. I call this component
10:28"Micro-VCSEL." This is a name I coined
10:31myself, and now some in the industry
10:33have started using it. I'm not sure who
10:36said it first, but looking at its
10:38structure, I felt it was just a
10:40simplified VCSEL, so I called it
10:42Micro-VCSEL. There is an opportunity
10:44there, so I think it is too early to
10:46tell. We might need to wait another two
10:49years. Wait two or three years to see
10:51how the overall market develops. Next,
10:53a viewer asked about previous news
10:55mentioning Intel and "early delivery."
10:56It might be related to glass substrates
10:59. It's possible, as Intel has been
11:01developing glass substrate packaging.
11:03Perhaps they are looking for a foundry
11:04to manufacture it for them. Every few
11:06days, there’s news that Intel has
11:09partnered with AUO to add microLEDs
11:11onto CPUs. I personally think the
11:13second possibility is higher, because
11:15as we just mentioned, microLEDs can be
11:17used for optical communication. Intel
11:20must have this idea too, so by using
11:22packaging to place these microLEDs
11:24directly onto the CPU, electrical
11:26signals can be converted directly into
11:29optical signals. In other words, the
11:31output from the chip will be optical
11:33signals in the future. So, this is
11:36possible. Frankly speaking, both are
11:38possible, but I think the latter is
11:40more likely. Because for this, they
11:42definitely have to find someone. As for
11:45whether they sought out AUO for the
11:46former, you all know Intel has its own
11:48factories and does glass packaging, so
11:50they might not need to look elsewhere.
11:52Of course, it's possible they might
11:54look for a foundry to cut costs and
11:55license the technology to AUO; that’s
11:57not entirely impossible. But I think
11:59the relative probability is lower.
12:01However, for the latter—adding
12:03microLEDs to CPUs—they definitely
12:04have to find a partner, because Intel
12:06wouldn't manufacture microLEDs
12:07themselves. Next, what key technologies
12:09do companies like Coherent or Lumentum
12:11hold? Can Chinese manufacturers really
12:13not catch up? It seems like some
12:15companies domestically can also produce
12:17100G or 200G EMLs. Alright, let me
12:19explain briefly. The most competitive
12:21aspect of companies like Lumentum and
12:23Coherent is their vertical integration.
12:25They handle everything from the
12:27upstream raw materials; if you look at
12:29them, they both produce their own
12:30materials. They do it all, from the raw
12:34material crystal growth and wafer
12:36slicing to the laser manufacturing
12:38process and even the backend packaging
12:41modules. So, they are vertically
12:43integrated. Of course, some of their
12:45downstream packaging modules are
12:47outsourced. I think they prefer to
12:49outsource these lower-end processes,
12:51and that’s what manufacturers in
12:52Taiwan and mainland China are currently
12:54doing—handling that outsourced
12:56packaging work. That is the current
12:58situation. So, simply put, the upstream
13:01and midstream sectors with high
13:03technical barriers are essentially
13:06controlled by them, and that is their
13:08key technology. So, things like
13:11material production, crystal growth,
13:14and laser manufacturing—these
13:16high-end, difficult technologies. Does
13:19Taiwan produce 100G or 200G EMLs? As
13:23far as I know, some people are doing it
13:25, but they probably haven't reached
13:27100G or 200G yet. I’ve seen some
13:29doing 20G, but it doesn't seem to be
13:32EML either. The EML you're referring to
13:34should be Electro-absorption Modulated
13:37Laser, I guess, that's what you mean.
13:40Because EML has two meanings: one is
13:43Externally Modulated Laser, and the
13:46other is Electro-absorption Modulated
13:49Laser. There are Taiwanese
13:52manufacturers making external
13:54modulation lasers, but to my knowledge,
13:56none have reached 100G. In other words,
13:58this technology still lags behind the
14:00industry leaders. Electro-absorption
14:02modulated lasers are likely not being
14:04produced in Taiwan; it seems that area
14:06is protected by patents. A netizen
14:07asked about Innolux's position in
14:09Chip-on-Panel; it's basically just a
14:11temporary move. Currently, Innolux is
14:14primarily focused on glass-based
14:16packaging. What they are doing now is
14:19panel-level packaging, which involves
14:22using a square panel as a temporary
14:24carrier, bonding chips and building a
14:27redistribution layer, and finally
14:29removing the glass and dicing it. This
14:33is not CoWoS packaging; at TSMC, it is
14:36called fan-out wafer-level packaging.
14:40If it's a square glass, it's
14:42panel-level packaging. Fan-out
14:45panel-level packaging, or Chip-on-Panel
14:48, basically uses glass as a temporary
14:51carrier to stack chips. If Innolux
14:55assists TSMC with foundry services,
14:57they would likely handle this part, but
15:00as far as I know, Innolux's glass
15:02packaging hasn't reached that level yet
15:05. Chip-on-Panel essentially requires
15:09very fine line widths. The fan-out
15:13panel-level packaging Innolux currently
15:15does has wider lines, so there is still
15:18a gap, but if they do foundry work for
15:21TSMC in the future, it would likely be
15:23for Chip-on-Panel. Everyone please note
15:28, TGV stands for Through-Glass Via. TGV
15:31is not used for temporary carriers; TGV
15:36has two uses. One is as the core layer
15:40for substrate boards; you know, ABF
15:43substrates have a core support. That
15:47currently uses glass fiber fabric, but
15:50there's a chance to switch to glass in
15:52the future. To switch to glass, you
15:55need vias; that's TGV. Another is the
15:58substrate manufacturers, like Unimicron
16:01, Kinsus, and Nanya, who are already
16:04working on this. This TGV process is
16:06being handled by companies like
16:08Unimicron, Kinsus, and Nanya because it
16:10involves the core layer of the
16:12substrate. The equipment has specific
16:14suppliers, so that's the current status
16:16. I believe Innolux might want to get
16:19into it, but their progress is
16:20currently uncertain. The third type is
16:23using glass interposers to replace
16:25silicon interposers. This part is a
16:28longer-term prospect, for after 2030,
16:30because I hear the difficulty is even
16:33higher, so that's the current situation
16:35. Next, a netizen asked if Leadyo makes
16:38OLED materials? Yeah, organic
16:41light-emitting diode materials. You
16:43know, there are actually no OLED panels
16:46in Taiwan, because there are almost no
16:48companies in Taiwan making OLEDs; this
16:51sector is almost monopolized by Korea.
16:53For Leicai, I think besides OLEDs, they
16:56should pick some other materials. They
16:58specifically mentioned electronic
17:01materials here, and if they only do
17:03OLEDs, I think it would be quite
17:05difficult, so they are doing high-end
17:07values in semiconductors and electronic
17:09materials, right? Right, they
17:11definitely have to do something else.
17:14Materials for printed circuit boards
17:16and substrates—I think there's an
17:18opportunity there. The most important
17:21key is the chairman of Xincai, Chairman
17:24Zhang Wen-xiong. He's supportive, and
17:27regardless, Xincai is a very successful
17:30example. As you know, Xincai used to be
17:33in a bad state; the company was almost
17:35unable to continue operations, but
17:37after Chairman Wen-xiong stepped in, he
17:39turned the company around, so he is
17:42very capable. He has also established a
17:44very good relationship of trust with
17:47TSMC. Therefore, with him leading Lilei
17:49Yingcai, I think there is an
17:51opportunity. Although he has a
17:53financial background, he has a great
17:56sense for business and has been dealing
17:59with TSMC for so many years, so he is
18:01very experienced and knows clearly what
18:04TSMC needs and how to do it. As long as
18:07he finds the right technical team to
18:09assist him, I don't think it would be
18:11difficult to find such a team given his
18:13current position in the industry. So, I
18:16think this company has a chance, but if
18:19you want to invest, I personally think
18:21that since they are just starting out,
18:24the risk is relatively high. So, if you
18:27are going to invest now, you should
18:29look at whether the price is reasonable
18:31; don't just follow the crowd and chase
18:33those high, unreasonable prices, or you
18:36will face risks. Now, let's talk about
18:38Zhaotong Optoelectronics; I haven't
18:41heard of this one. Let me take a look;
18:43it makes epitaxial wafers for GaAs and
18:46InP substrates, which are III-V
18:48materials, basically similar to Lianya.
18:51It's somewhat similar; they make
18:53epitaxial wafers, which means growing
18:55epitaxial layers on wafers. These are
18:57single-crystal thin films, so they make
19:00surface-emitting GaAs VCSELs, right? As
19:02we just mentioned, the epitaxy for
19:04Vertical-Cavity Surface-Emitting Lasers
19:06. They don't make the lasers themselves
19:08; they make the epitaxial wafers—that
19:11is, growing thin films, many layers,
19:13for GaAs LEDs. This looks like it's
19:15making something similar to Lien Ya,
19:17though it seems to have fewer product
19:19types, with both 4-inch and 6-inch
19:21being gallium arsenide. And there's
19:23gallium phosphide too, right? This
19:25gallium phosphide is for edge-emitting
19:27lasers. DFB laser is the epitaxial
19:30wafer for the laser. So I think this
19:36company is worth watching depending on
19:38their performance; I think there's an
19:40opportunity, as there is currently a
19:42shortage after all. There's a chance, I
19:45think there's a chance, and this market
19:47is currently in demand. Next, a netizen
19:49asked about the news that Intel is
19:51working with AUO on MicroLED, just to
19:53make it emit light. Yes, yes, it's just
19:56using LEDs. It serves as CPU status
19:58indicators, such as for boot failure or
20:01other conditions, rather than a
20:03MicroLED CPU; it's likely just for CPU
20:05lighting. I think if it were just for
20:07status lights, it wouldn't be worth
20:09much; how many can you put on a single
20:11CPU? That's not very useful. I think it
20:14goes beyond that; they probably want to
20:17do this—well, it's not called CPO,
20:19but it basically converts signals from
20:21the chip into optical signals, and a
20:24CPU can do that too. It's a type of...
20:26next, I was asked what I think of
20:29Corning. Corning, well, as everyone
20:31knows, Corning is the best at glass,
20:34making optical fibers and glass
20:36waveguides. These are all Corning's
20:38strengths, so this company is certainly
20:40very unique; after all, they are the
20:43leader. There are many processes in
20:45semiconductors that require glass, and
20:47even future panels and chip packaging
20:49will use glass. So I think Corning's
20:52situation should be quite good. With
20:55the demand for AI in recent years,
20:57Corning is a company to keep an eye on,
20:59so let me know if you have any more
21:01questions about them. They previously
21:04launched GlassBridge, right? It's a CPU
21:07solution that uses glass as an optical
21:10waveguide. So they are involved in this
21:13area as well. Next, I didn't attend
21:16Largan's investor conference, so you
21:18might have seen what they said—if you
21:20have questions, just ask. Let me put it
21:23simply: Largan's performance this time
21:26was actually quite good, but it was
21:29mainly due to lenses. Their CPO-related
21:32co-packaged optics don't have much
21:35revenue yet. Whether it can
21:37successfully pass TSMC's verification
21:39in the future is still something to
21:42observe. Everyone, I haven't received
21:44any news that they have passed
21:46verification, but the company is indeed
21:48very serious about transforming and
21:49investing, which is why the stock price
21:52has been hyped up so high recently. I
21:54still want to remind everyone that you
21:55should still look at the P/E ratio.
21:57Besides Corning, Formosa Plastics has
21:59also risen significantly recently; I
22:01heard it broke through 80 today. It's
22:03the same logic. Formosa Plastics
22:07originally focused on
22:08petrochemical-related plastics, but it
22:11transformed to start producing
22:13semiconductor-related chemicals, so the
22:16market has re-evaluated it higher.
22:19That’s about it. In short, it has to
22:22be reasonable; if the P/E ratio is
22:24reasonable, there’s generally an
22:26opportunity. As for Largan, it also
22:29depends on whether they pass TSMC's
22:31verification later on. Remember, the
22:34CPO product Largan is making is a fiber
22:36array unit, which several companies are
22:39working on. And as I’ve said before,
22:41this thing isn't as incredibly
22:43difficult as you might imagine. It’s
22:46just a simple structure; although
22:47alignment is difficult, it remains a
22:49simple structure. In other words,
22:51everyone wants to do it, so a bunch of
22:53manufacturers jumped in all at once;
22:55almost every company says they want to
22:57make fiber array units. Their gross
22:59margin is also very high. Think about
23:01it: everyone wants to make something
23:03this profitable, so once it's produced,
23:06the market will be very competitive.
23:08That’s a risk to keep in mind. Next,
23:10is there any research indicating how
23:12much electricity a data center could
23:15save by adopting MicroLED CPO glass
23:17substrates? The AI says it could save
23:19tens of thousands of kilowatt-hours a
23:21year, which is possible, but it depends
23:23on the scale of your data center. I
23:25believe the main power waste in
23:27training isn't in the optics; it’s
23:30primarily in the GPUs themselves.
23:33Ultimately, the power draw is the GPU;
23:36optics are secondary, though using
23:38lasers does consume more power. I might
23:41need to look that up, as I don't have
23:43actual data on the AI’s claim of tens
23:46of thousands of kilowatt-hours. It’s
23:48not impossible because it depends on
23:50the scale of the data center. For a
23:52large-scale data center, saving tens of
23:55thousands of kilowatt-hours a year in
23:57the optical section is possible. But
23:59the scale of the data center is the key
24:01point. Today happens to be the start of
24:03the holiday, so I guess everyone wants
24:05to take a break, right? That’s why
24:06there are fewer questions today. So
24:08let’s wrap up early and let everyone
24:11enjoy their holiday. Alright, that's it
24:13for today's live stream. Since it's the
24:17holiday, if you have any further
24:18questions, feel free to leave a comment
24:20below the video, and I'll explain them
24:22to you later. Thank you everyone, good
24:26night, bye.

About this transcript

This page contains the full transcript of 【曲博科技教室 EP324】川湖、鋒霈、英特爾、友達、Lumentum、Coherent、群創、昱鐳應材、兆通光電、Corning、大立光 by 曲博科技教室 Dr. J Class, generated from the public captions YouTube serves with the video. The transcript has 3,732 words across 612 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.