JPM continues to see semi trade extending its longevity supported by:

• Agentic AI benefits gradually materialize in real life applications across industries.

• Emerging adoptions in physical AI, like robotics, autonomous driving.

Both driving CSPs and enterprises to invest more.

JPM continues to like AI beneficiaries in Asian tech supply chains, naming TSMC, ASE, Alchip, Unimicron, SEC, and SK Hynix.

Citi with similar comment noting NVIDIA suggested the tightest constraints are not in GPU silicon alone, but broader AI system deployment capacity. Key bottlenecks now include HBM supply, advanced packaging, optical interconnects, liquid cooling systems, rack integration, and power infrastructure.

FundAI made these observations more than a month ago here.

Taiyo Yuden: +11% on Friday after sell-side roundtable on Thursday.

GS, MS and Citi published after the meeting with Buy (¥7,100), EW (¥4,700) and Sell (¥2,900). The stock ended the week at ¥9,100.

• MS says Taiyo Yuden’s MLCC market share (11.3%) is third place behind Murata (40.8%) and Semco (22.5%)

• AI server MLCC sales grew ~30-35% YoY in FY3/26 and guided to accelerate to +80% YoY in BOTH FY3/27 and FY3/28 (a ~3.2x jump in two years).

  • However, servers were only ~6% of total company sales in FY3/26 and GS adds that MLCCs for servers were just over 9% of total MLCC sales.

• Taiyo Yuden forecasts 32% CAGR for AI server MLCC demand 2026-2030 (4x over five years), which the company itself frames as conservative.

• No MLCC price hikes this fiscal year, but FY3/28 is a possibility. GS and Citi tie say the company will hike if there is ~95% capacity utilization.

• Mix shift toward smaller, higher-capacity parts is the ASP driver. Continued migration to small-case high-cap products is lifting ASP and carries high margins (GS gives the specifics: 47µF moving from 2012/1608 to 1005, 100µF from 2012 to 1608 in 2H 2026).

• Capacity expansion held at ~10%/year. Management cautious about going higher given past overbuild.

• Management acknowledged attempts to reduce MLCC count/capacitance per board via vertical power delivery and component-embedded substrates. The company (and Citi) view this risk as small.

None of the analysts appear prepared to chase and up pay up for this.

We hear investors lean towards Yageo and Murata, but happy to listen to counterarguments.

Mitsui Kinzoku: +6% after medium-term plan briefing (FY2025-27) after close on Thursday

Mitsui Kinzoku holds a near-monopoly — ~100% global share — in carrier/ultra-thin copper foil, sold under the MicroThin brand. It is a required ingredient for building the ultra-fine circuit boards used in advanced chip packaging and high-speed optical transceivers. Mitsui also sells a high-speed foil (”VSP”) for AI servers and switches.

Read more

As always, if you can’t meet in person, always happy to hear from you in DMs about what you are interested in hearing more about.

This is the usual weekly roundup of what we have been discussing online, in the chat, DMs or on X.

AT&S: AMD and Intel supplier with capacity expansion funded by LTAs, with more on the way.

MLCCs: Fintwit now looking upstream

GL Techno: Industrials specialist Monolith Capital now a fan

Memory, PCBs, MLCCs

In case you missed it, this was being reposted on X yesterday, the point being the % increases of the various line items.

MS:

Based on our latest estimates, we estimate MLCC content for VR200 to be ~US$4.3K. This would present a quite meaningful increase vs. GB300 at only ~US$1.5K, and could explain why high-end AI server MLCC demand is currently so strong and causing all the ODMs to aggressively trying to secure and build as much inventory as possible, ahead of the Rubin rack ramp from 2H26 on wards. Our checks indicate that the MLCC content per computing board and switch board is increasing quite meaningfully, with the computing board MLCC content showing a larger content increase. In addition, the newly introduced BlueField and ConnectX modules will also contribute to more MLCC content per rack.

Feedback from chat on why this is the wrong way to look at it:

Read more

Lots of questions about whether the boat has been missed so I thought to summarise the bull case being circulated by key opinion leaders and reproduce some highlights from the chat.

AI server demand for a tiny but critical component — the capacitor — is inflecting hard, and supply can't keep up:

• GPU power consumption is exploding across generations, and more power means proportionally more capacitors required per rack.

• New rack power architectures require more expensive, harder-to-manufacture capacitors, lifting average selling prices even before counting unit growth.

Read more

We feature one such idea contributed via email today.

This is said to be one of the cheapest ways to have copper clad laminate (CCL) exposure. You can own the majority of a CCL company through a holdco, with the holdco’s four other divisions covering almost the entire market cap of the holdco.

The tightness of the CCL market, and its attractiveness to investors seeking AI exposure, is evident from this story translated in Jukan’s tweet:

Subscribe now

Read more

Thank you to all the very generous people who have met me online and offline over the last month. This is an interesting new direction for the newsletter, as it is starting to offer readers a sense of what investors in Asia are thinking.

The content of this post will be pinned to the main page and updated from time to time. As always, if you would like more colour, please reach out.

Subscribe now

Memory

For the average retail investor, memory is just one of the more accessible ways to get exposure to the AI trade.

We have a bias towards NAND, but they will all do well.

Generally, in our conversations with investors, many believe multiples will move up as long term supply agreements are signed and earnings will grow as the shortage worsens.

The desperation for supply is evident from this article:

SK Hynix flooded flooded with unprecedented offers from big tech firms to secure chip supplies (RT)

The offers, unprecedented in the global memory chip industry, underscore ‌the severity of the component's scarcity around the world, as chipmakers struggle to keep pace with surging demand amid an artificial intelligence boom.

The firm’s customers have been proposing a range of offers to the South Korean chipmaker, including investing in dedicated memory production lines, six people said.

Another proposal involved customers financing equipment purchases such as ASML’s (ASML.AS), opens new tab extreme ultraviolet lithography machines, which are used ​to print circuits onto silicon wafers and are worth hundreds of millions of dollars, three of the people said.

But the chipmaker, flush with cash, is cautious about taking financial commitments ​from customers, as such deals could hold it hostage to specific buyers and require it to supply chips at lower prices in exchange for ⁠securing longer-term and more stable revenue guarantees, two people said.

There are more exotic ways to play this, such as through SK Square. We discuss these in the chat but you can also reach out to Origo.

Legacy NAND

Macronix is a relatively differentiated idea. As a retail investor you may believe an idea has already been discovered because it has been written up on X and Substack, but I have quickly learned that this is only scratching the surface of discovery.

For example, someone told me last week that they have many analysts covering memory and have only just discovered Macronix.

This is what Macronix reported yesterday. I can offer you some resources if you would like to read in.

Source: Company documents.

ABF Substrate

This is another investor favourite at the moment and is seen as one of the most severe shortages next to memory.

Jukan occasionally pops up in the Substack chat and he offered this chart yesterday:

As a price insensitive product you would expect a 27% supply shortfall to cause a significant price increase. Coupled with LTAs being signed, you get a similar dynamic to memory here. Further, ABF supply is constrained by an upstream input called glass fiber, specifically the high-end variety known as T-glass.

Ibiden, Unimicron and AT&S are the three companies I follow. Again, they will all do well. I don’t currently have an informed opinion on which will do better.

We have some notes on what is driving this shortage, available on request.

Subscribe now

Taiwan

These are some bullet points on Taiwanese companies flagged to me recently:

• Generally, the TPU complex has done very well over the last month and Mediatek has likely been one of the best performers. This one has been raised a few times in the chat and in DMs.

• E&R was flagged as an Intel supply chain idea by some very credible people. Recent report from Taiwan also suggests equipment orders from Intel.

• Prosperity Dielectrics, a MLCC play, has doubled since it was first mentioned in our chat.

• Aspeed

Read more

Toto closed at ¥5,425 yesterday and hit the +¥1,000 limit (+18.4%) early this morning.

Since Toto manufactures consumable electrostatic chucks and now reports it as a standalone segment, the company provides a read through for WFE consumable companies.

From the earnings release:

As replacement demand remained firm due to improved utilization rates at the existing factories of semiconductor device manufacturers, sales of the TOTO Group ceramic products used in semiconductor manufacturing equipment also increased compared with the previous year.

Positive read through for:

• GL Techno +6%

• UMS Integration (SGX closed today)

• Metasurface Technologies (HKEX closed today)

  • In particular, the replacement demand for vacuum chambers is key to the investment case for this company. I’m not sure the extent to which replacement demand is important for UMS Integration.

  • On a separate note, Metasurface has start referring to more than one semiconductor OEM customer in its annual report. The market will soon learn they supply more than just Applied Materials.

Remarkably, Toto’s electrostatic chucks business appears to form <10% of sales but generates 40% of profits. Guidance was strong and the company has plans to boost production capacity at three plants (BBG).

Tokyo Electron also gave strong guidance. TEL guided net sales of ¥1.57T for H1 FY3/27 — a 33% YoY increase and above market expectations of ¥1.42T — with OP of ¥431B, up 42% and ahead of consensus ¥405.9B.

In comments earlier this week, Tokyo Electron’s Chief Executive Officer Toshiki Kawai said there’s potential for more spending on NAND production capacity, and he has not yet factored in that upside to the company’s forecast.

Korea: Home to a cheaper TTMI?

Following up on our last note about Korea, using Funda AI we compared Korean’s version of TTMI to… TTMI.

This Korean company is thought to have Microsoft, Google, Nvidia, Cisco and Arista as customers, and is a purer AI play with a lower valuation:

Read more

Two caveats: (1) These names have run alot this year and (2) some of this content is quite dated but should be a good place to kick off some reading.

We also pull some names from the FundaAI platform.

Read more

Read more

Quite incredible to see ready answers to such a wide range of topics including:

• Frontier model updates (Mythos, Gemini 4).

• Anthropic's compute constraints and the opening it creates for OpenAI.

• Emerging CPU bottleneck.

• Cybersecurity.

• Memory/NAND pricing.

• Optics supply chain.

Business momentum has been great too as we sign up more clients to the institutional offering, which offers more AI skills, tools and features on the platform, more in-depth reports, and regular meetings with analysts. Reach out to sales@funda.ai for inquiries.

Rigaku limit up today as Onto takes 27% stake and announces that new jointly developed offering has already been selected by two key customers. (PR)

Rigaku is now up around 50% since we developed more conviction back in March.

Oppenheimer reiterates PT of $350 and has this to say:

Macronix: Dipped 25% this week on UMC eMMC rumours

Various people chatting with us have been discrediting the rumour that UMC is looking to compete with Macronix in legacy NAND products.

This was echoed in Digitimes yesterday:

MS is also constantly putting out bullish notes on Macronix, such as the one here.

Various semiconductor companies marked safe after earthquake in Japan (Nikkei)

Kioxia PR

TEL PR

No PR from Shin-Etsu or Mitsubishi Gas Chemical yet, but see Trendforce graphic below.

Victory Giant: Closes 50% above IPO price

Our friend and chat lurker, Acid Investments, has a take here.

In terms of where other peers trade at, the Taiwanese names are going absolute gangbusters. Nanya for example trades at a nose-bleeding 44.2x FY27 p/e and 20x FY28 p/e; Unimicrom is trading at 31.8x FY27 and 19x FY28 p/e. The Japanese names, Ibiden and MEC are also all trading in the 20s for FY28 p/e.

My guess it we let the initial quick traders wash out first before some long-onlys (who were reluctant to be locked up) come in and take this higher, as the the valuation discrepancy becomes apparent.

Victory Giant is a global leader in advanced printed circuit board and a key Nvidia supplier. See Victory Giant’s mentions in Nvidia’s roadmap below.

Read more

The idea was knocked around for a bit and by popular demand, Iqbal is publishing his take on FPS below.

Iqbal is an occasional contributor to this newsletter and is responsible for uncovering D-Box Technologies as well as helping me write up ideas like Metasurface Technologies, AT&S, CTR Holdings, and Suncall Corporation. Which means to say he has a pretty good nose for this work.

As always, by accessing this content you acknowledge our terms and conditions. Also, to be clear, Iqbal is not associated with FundaAI and what follows is written in his voice.

Subscribe now

Read more

In my second week working with FundaAI, we have published a note on TPUs, now Funda’s most read work on Substack.

I have noticed the views there echoed in a few other places, like this academic paper:

"The dominant constraints on AI systems are shifting from raw compute capability to data movement, connectivity, energy efficiency, system-level integration, and cost effectiveness (e.g., $/token)."

"Connectivity is as critical as computation: Performance scaling increasingly depends on interconnect bandwidth, latency, and topology, requiring connectivity–compute co-design rather than treating networking as a secondary concern."

Fabian also made a similar comment on this back in December:

For years the semiconductor industry measured progress in FLOPs. More compute, faster chips, bigger clusters. That framing is now incomplete. The real constraint in modern AI infrastructure is not how fast individual GPUs compute. It is how fast thousands of them can share data with each other.

Kioxia: YMTC has three factories on the way that will more than double production (RT)

Funda also published a piece on Kioxia last week.

Kioxia has reportedly started operations at the second Kitakami building (K2) in September 2025, installing equipment and expanding cleanroom areas to further boost production.

We’ll be looking out for what they have to say about capital returns on their investor briefing in June.

Sentiment has recently been weighed down by reports of YMTC’s expansion plans:

• Expansion: Each of the three new plants will produce 100,000 wafers per month.

• Current Capacity: YMTC currently produces a combined 200,000 wafers per month from its first two fabs.

• Self-Sufficiency: Over 50% of the equipment for the third factory (Wuhan) is sourced from domestic Chinese companies. The third factory has been completed and is currently installing equipment.

• DRAM Entry: YMTC is also pivoting into DRAM; all three new plants will allocate some capacity to memory chips. The exact amount will depend on the company’s progress in developing those chips.

Another Chinese article continued to weigh on the stock today:

“Yangtze Memory Technologies Co., Ltd.’s revenue in the first quarter of this year exceeded 20 billion yuan, more than doubling year-on-year. Its NAND (flash memory) chip production has exceeded 10% of the global market share, approaching the third largest in the world,” said a core person in the memory chip industry chain. “The profits will be even more explosive in the future.”

Chinese imports of US chip tools from Singapore and Malaysia hit record (Nikkei Asia)

According to Nikkei Asia, Chinese imports of US chip tools from Singapore and Malaysia in 2025 have hit record highs.

• Singapore: $5.7 billion (up 17% YoY).

• Malaysia: $3.4 billion (more than double the 2024 figure).

Lam Research is significantly expanding its manufacturing footprint in Malaysia. We believe UMS Integration’s facility (serving Lam) is currently at 50% utilization. The stock gained roughly 15% this week.

Some good news for Lam and Tokyo Electron too in the revised MATCH bill reported by Reuters here.

Hansol Group makes a small acquisition to enter the DRAM probe card market

Maeil Business:

Hansol is acquiring an 83.37% stake in Will Technology for ₩177.2 billion KRW (~US$120.6 million), implying a total valuation of US$144.6 million.

• Will Technology is the #1 provider of probe cards for Samsung’s Galaxy S series APs and CIS (image sensor chips).

• Hansol plans to use this acquisition to enter the DRAM probe card market and expand into China.

• Valuation: We estimate this implies a P/E of ~24.4x and a P/S of ~3.15x. This might be seen as a rich valuation for a new entrant compared to established Japanese peers like JEM (6855).

Special thanks to one of our readers for flagging this.

Macronix thesis playing out elsewhere in Taiwan

In brief, the Macronix (2337.TW) thesis is that it is emerging as a highly profitable monopoly in the eMMC memory market as the "Big Four" memory giants abandon this market and reallocate their capacity toward high-margin HBM and high-end NAND flash.

There appears to be a similar play in another industry:

Read more

Please keep the DMs and emails coming - we like hearing from you.

In particular, if you have any questions about the Funda AI platform, I am happy to address them. This week, we have had two examples of readers surfacing new ideas from the Funda AI chatbox.

AMAT: UBS says management talking about incremental SAM capture opportunities from the industry’s shift to more vertical structures.

Gate-All-Around (the vertical structures in question) was described by management as a “monster” node.

This is one of the transitions driving the adoption of x-ray inspection, i.e. Rigaku.

UBS’s PT of $430 is unchanged.

Micronics: Profiled by Bernstein today

Bernstein moving in on our Japanese small cap turf!

This report is part of a series of reports on Japan semi small cap companies, which we don’t cover but occupy unique positions in the supply chain and potentially benefit from the technology shifts. Investors may find them interesting as the small caps generally trade at lower multiples than large caps, but the growth profile is not necessarily inferior.

Can we replace Micronics with JEM in the whole document:

We note four factors shaping Micronics’s revenue outlook: (1) increased HBM contribution driving up probe card content per wafer, (2) DRAM capacity expansion across all three major players, (3) Samsung HBM exposure as a concentrated but large revenue driver, and (4) Micronics’s own manufacturing capacity expansion as the primary historical constraint.

… we believe the underlying probe card demand potentially can grow growth 32% / 27% in CY26/27, respectively, accelerating from 27% growth in 2025.

Below the paywall:

• JEM’s consensus numbers (which are likely too low) compared with peers.

• MS updates us on one of their top picks in Taiwan.

• New bottleneck: a game of whack-a-mole.

• Micro/small caps in Singapore.

• Some articles on Digitimes that caught my eye.

Subscribe now

Read more

This year, I decided that you would be better served by me meeting the industry in APAC, instead of staying anonymous.

Inevitably, I met the founders of FundaAI.

I came away impressed by the depth of the team and their determination to become a respected and trusted research firm. They are also intent on delivering an ambitious AI powered research OS.

From day one, every decision about this newsletter has been centered on one question: will this earn me one true fan today?

That question led me to accept FundaAI’s offer to join them while continuing to run this newsletter.

Collyer Bridge posts will be pushed to FundaAI’s institutional clients. You will continue to receive at least one email a week here as well as notes on small cap APAC names not covered by FundaAI.

I am excited about this opportunity for a number of reasons:

1. As mentioned yesterday, independent voices on Substack are going to play a bigger role in finance. FundaAI is already a respected voice in just over a year of publishing, and I want to help accelerate the growth of the firm.

2. My research work for FundaAI will focus on mid-cap APAC names.

3. There will be strong synergies. FundaAI’s research will inevitably surface strong small cap value chain ideas that you will hear about here.

4. You will continue to receive an APAC roundup at least once a week. Many of you have told me you find these helpful. Personally, I’m starting to find that talking to more people is helping me with high signal, differentiated content.

5. Posts on this newsletter represent my personal, independent views. As mentioned, they will also be received by FundaAI’s clients, which is my subtle way of saying that equal care will go into both endeavors.

6. Importantly, I believe we can only realise these synergies because FundaAI has a generous and collegiate culture, and a strong commitment to ethical research and compliance.

Since we started seven months ago, I have iterated and pushed myself in service of this growing audience. The drive to improve and the openness to experiences and ideas has served us well so far, and should continue to yield results.

Paid subscribers here grew 150% in Q1 of 2026.

Engagement is at an all time high.

We have a once-in-a-generation AI buildout to tackle. You, me, FundaAI, and everyone publishing on this platform. LFG.

Get 20% off forever

I started advocating the idea that your Substack writer could serve as your personal analyst, concierge, or private investigator.

It was a big part of the early pitch for becoming a paid subscriber: I will go do things for you in a faraway foreign country!

I thought it is worth repeating that idea.

We have 512 paid subscribers today, but I think I have had fewer than 20 significant requests for things. Off the top of my head these included:

1. Visit a data center under construction in Singapore and identify the sub-contractors.

2. Visit a factory in Johor and a building in Yokohama (obviously will only agree to this in special circumstances).

3. Send me material on X subject (only a handful of people regularly make this request).

4. Send me a summary of your call with Y (only one person has ever asked for this).

5. What did AYZ say about Maruwa?

6. Connect me with Z.

The power users of this Substack ask me questions all the time. There is also a similar dynamic in the FundaAI chat.

I wonder if I am opening a Pandora’s box by repeating the invitation to do so. You can reach me on DMs or at research@collyerbridge.com.

Of course I can’t answer every question, but it costs paid subscribers nothing to ask.

No, I will not visit the Strait of Hormuz. I would only agree to that if I had 10,000 paid subscribers.

Substack is complementing traditional sell-side research

On a different note, larger independent voices on Substack are becoming increasingly important in finance.

Citrini’s thought experiment was apparently escalated to the top of KKR.

TMTB post titles appear on Bloomberg.

Vik is too modest to advertise that he is quoted in JPM updates:

This trend is still in its infancy. By being a paid subscriber to any finance or tech Substack, you are already ahead of the curve.

And on that note, just because you read this far, here is a 20% coupon to the newsletter.

Get 20% off forever

We’ll like to see that trend accelerate. More on that soon!

Hat tip to Jukan who is the source for 3/4 of this update:

1. SLC and MLC price increases.

2. Applied–Besi hybrid bonding inline system already running at TSMC and being adopted by SK Hynix and Broadcom, and possibly Samsung.

3. Samsung evaluating new inspection equipment for hybrid bonding.

4. Metasurface results.

Subscribe now

TrendForce March 2026 NAND Flash Contract Price Update

▶️ March Pricing Highlights

SLC: ASPs across all products rose in unison, each up more than 20% MoM

MLC: Supply exit and inventory depletion drove a +49.34% surge in 64Gb, hitting a new all-time high

Demand diversification (AI devices, robotics, industrial equipment, automotive) and extreme supply constraints are acting simultaneously

Read more

Much of Japan went ex-div today. Among the companies we follow that includes Seikoh Giken, Innotech, GL Techno, JEM and Fujibo.

Optics and CCL remain investor favourites so we would flag that Seikoh Giken, Sumitomo Electric, Nittobo, and Resonac are around their 50-day EMA.

TSMC: Citi base case now NT$ 2,800; bull case NT$ 3,000

Citi’s 2027/28E EPS are 18%/28% higher than Bloomberg consensus.

Subscribe now

Citi’s capex estimates are US$67bn in 2027 and US$75bn or higher in 2028.

Surely Uncle Wei can be persuaded to go for US$70bn and US$80bn in 27/28.

It looks like the TSMC capex winners in the last month were All Ring Tech and GPTC:

And to a lesser degree Ta Liang:

Samsung Electronics: moving into silicon photonics in 2028

Below the paywall: Soitec short case and Malaysian stocks holding up surprisingly well this month.

Read more

Also positive for the Japanese players:

U.S. International Trade Commission investigation of SK Hynix and Kioxia (PR)

This appears to be a patent infringement complaint initiated by MonolithIC 3D Inc.

Could be a nothing burger designed to shake down SK Hynix and Kioxia. Seemed underreported on Twitter.

Chunghwa Precision Test: Breaks ground on $111 million investment, its third facility.

More bullish news for probe cards.

“Storm Media”:

General Manager Scott Huang (黃水可) highlighted that demand in the current market cycle is no longer growing linearly but expanding by multiples.

To meet this immediate surge, the company is leasing additional floor space and reorganizing production lines at its two existing facilities, aiming to double its medium-to-long-term capacity by late August or September. If these near-term upgrades proceed as scheduled, Huang expects this year’s growth to easily surpass last year’s figures.

While co-packaged optics represent a crucial future direction, Huang stressed that the underlying testing architecture must be completely redefined as the industry transitions from advanced packaging to mass production.

Consequently, future competition in advanced testing will hinge on accommodating new packaging formats and managing extreme power consumption. Chunghwa’s current probe cards can support testing at roughly 1,000 watts, with a next-generation target of 2,000 watts — a milestone that will require significant breakthroughs in thermal management.

Related Digitimes coverage:

Chunghwa Precision Test breaks ground on new Taiwan plant to meet AI chip demand

Taiwanese MEMS probe card makers to double capacity in 2026

Subscribe now

SemiAnalysis: “very material” reduction in previously forecasted 1.7mn Rubin shipments in 2026 due to Hynix production issues.

The HBM4 situation for Nvidia’s Rubin ramp is becoming worse than expected and therefore we see much greater downside risk to this year’s Rubin output.

Read more

We also had our day trip to Yokohama pay off:

Innotech Corporation: Sale of HQ announced

Innotech Corporation is slightly underwater from when we wrote it up on 25 February. We took a basket sized position after reporting on the company’s HQ, one of the only small positions we kept going into the Iran war.

However, today the company announced the foreshadowed sale and leaseback of its HQ is going ahead.

As a consequence, this 35 billion yen company will recognize a 6.3 billion yen gain (being the sale price less book value and transaction costs).

The company announced a small 1.95% special dividend but we think the hope is that:

• a bigger special dividend will be coming in the next financial year (ending 31 March 2027), when the bulk of the gain will be recognised.

• the company will increase its buyback, currently sized at 3% of the company.

The amount of cash coming in the door is of course much larger than 6.3 billion yen, so we are looking forward to seeing where this goes.

Broadcom flags supply constraints, including lasers and optical transceiver PCBs (Reuters)

Broadcom quoted by today about the tight supply in PCBs and the emergence of LTAs in this market:

"Even though there are ‌multiple ⁠suppliers in the industry today ... there is definitely a supply constraint in the laser space,” he said, adding that printed circuit boards had also emerged as an "unexpected" bottleneck.

He cited PCBs used in optical transceivers, saying lead times have stretched from about six weeks to six ​months.

Both Taiwanese and Chinese ​PCB suppliers are ⁠facing capacity limits, contributing to the delays, Ramachandran said, without naming the suppliers.

He said that he is not overly concerned about the industry as new entrants ​and capacity expansion are expected to ease supply constraints over time.

Many ​customers are ⁠now entering long-term agreements with suppliers to secure capacity commitments for as long as three to four years, he added.

Zephyr likes FIC Global for optical transceiver PCBs:

Echoed by Serenity:

Zhen Ding Tech looked the most interesting when I was looking into Taiwan a few days ago. This table is from GS:

A small account on X has been pitching a US$4B Japanese PCB maker that will apparently double FY25 operating income by FY28. The tweet has 55 views and the stock is up ~20% since:

Subscribe now

Read more

We’re only holding on to our highest conviction positions as we ride out the exchange of tweets between Donald Trump and a cast of Iranian leaders.

In Hong Kong, the largest position is our favourite Applied Materials fabricator, which is pursuing a secondary listing in Singapore:

We also own a negative EV Singaporean builder, Kwan Yong Holdings, which got cheaper this week.

At HK$ 0.435/share, the company has more than 2.5x its market cap in cash. The company should also generate 72% of its market cap in profit from its order book alone. Last year, it declared a high single digit dividend.

And finally we have what is left of our value trap CTR Holdings, in the hope that they will disclose their work for Micron and declare a dividend.

Two new ideas for you in Hong Kong.

The first relates to the demand for server chassis:

Here Dan links to an article explaining that Vera Rubin is causing a three fold increase in demand for server racks.

Trendforce is also reporting that ASIC-based AI servers are forecast to account for 27.8% of total AI server shipments in 2026, rising to nearly 40% by 2030.

There is a server chassis maker benefitting from both trends and now trading below its recent placement price.

Subscribe now

Read more

This is a continuation of that note.

30 January 2026

TMTB’s coverage of Sandisk earnings. This one helped me (and our readers) put a $555 floor (which is also “LOL” in Thai) on Sandisk over the last two months, helping us trade in and out of this multiple times for 15-25% each time.

I think Funda AI’s comments yesterday on 2027 EPS will likely be as helpful as we navigate the weeks to come.

2 February 2026

A reader spent the time to independently “audit” Substack returns using Gemini to create this (crazy) 30-day return leaderboard.

There is absolutely no question that Global Tech Research was the highest alpha Substack around. Now offline due to some drama with Chinese investors - but we have archived many of his posts - please reach out if you are looking for one.

A little surprising that we are at number three. I would expect many of the picks here to take more than three months to play out.

Anyway, 2 February 2026 brought us AYZ’s incredible Mipox callout in this post:

This post was full of gems but for most readers of this newsletter, the highlights were:

• Calling out Seikoh Giken, which helped it go from 13k yen to 30k yen (which is where I sold most of my bag).

• Calling out Mipox, which proceeded to double in two weeks.

I have archived this post, which paid readers can access below the paywall.

Read more