A Week of Serenity's Public Posts, Lined Up: This Week the Evidence Was All Written Into Contracts

Contents

1. 2 November: Someone Simply Bought the Bottleneck
2. 3 November: The Next Day, Two Opposite Prices at Once
3. 3 to 6 November: The Contract Terms Start Talking
4. 6 November: The Rules That Took Effect on 9 November
5. 7 November: He Closes the Frame on Bottlenecks Versus Pricing Power
6. Getting This Far, the Field I Care About Most
7. The Other Place I Have No Answer: Being Right Still Means Waiting
8. One Thing Worth Taking With You

Words pay no debts, give her deeds.

— Shakespeare, Troilus and Cressida, Act III, Scene ii

Laid out in order, the public posts Serenity put on X across the six days from 2 to 7 November have a shape: nothing that pushed the argument forward this week came from a share price. It all came from contracts. A networking chip company bought an upstream laser maker at 4.5 of its own shares for every target share, in a deal that will not close until 2028. An optical components maker disclosed a second long-term agreement more than five times the size of the first, with lasers written into the supply scope. A compute landlord signed a five-year contract carrying ninety percent of the accelerator purchase as prepayment. The source layer for this piece is a single one: posts on his public timeline that anyone can open. Nothing from behind a subscription wall made it in here. The figures fall into three source types, public filings, public earnings calls and public news, plus a handful that travelled one step further through newspaper and brokerage accounts, which rate a level lower on evidence strength, and I say so in those passages. The limits go up front: he is a heavily positioned long, he says this drawdown has been deep, and he only added leverage into the fall on 5 November, so the material he selects leans toward the side where demand is still there.

2 November: Someone Simply Bought the Bottleneck

The week opened with an acquisition he had not expected. A networking chip company announced it was buying a laser and optical components maker, at 4.5 buyer shares for each target share, with closing projected for 2028. His first reaction that day was that he had assumed the buyer would pick a smaller target, since the buyer looked like it was walking the vertical integration path and what it lacked was light sources.

The company being bought was already sold out through 2029, and had announced a roughly five billion dollar co-packaged optics laser contract that same week. Co-packaged optics, in plain terms, means putting the optical engine inside the same package as the switch chip and skipping the pluggable module. He put gross margin on that contract above sixty percent, and ran one more number alongside it: a single site could carry five billion dollars of annual revenue. I find that figure more useful than the deal itself, because it converts “lasers are the bottleneck” into what the profit looks like once the thing scales.

Then came the step that set up the rest of the week: the remaining independent laser suppliers got more expensive that day. Two reasons. The first is acquisition value — other buyers eyeing vertical integration now know the path works, and know the pool of targets just shrank by one. The second is more practical: nobody competing head-on with the buyer wants their light source parked long-term inside a rival’s subsidiary, so qualifying a second supplier moved from something you could do to something you had better do quickly.

He also named a structural side effect, which I only thought about after reading him. With the deal hanging out to 2028, the target’s share price gets pinned to the exchange ratio for that whole stretch, and independent price discovery disappears. Anyone holding that stock spends the next year or more holding a shadow of the buyer’s price instead of a verdict on the company itself. A reader raised the same point that day: the ratio times the buyer’s price at the time still looked higher than the target’s market price, except it floats, and it falls when the buyer falls.

China already had cases of this shape. Someone in his thread added two: an electronics manufacturer that bought an optical module maker, and a systems company that invested in several domestic laser makers. The driving motive there is supply autonomy and less dependence on Western supply chains, and that same post concluded the strategic impact of this deal on China’s main module makers would be limited.

3 November: The Next Day, Two Opposite Prices at Once

The following day’s material split into two sides, and the two sides pointed opposite ways.

On one side, second sources were being repriced. He relayed a brokerage view — softer evidence, somebody’s inference rather than company disclosure — that every customer competing with the buyer will want an independent second light source, and that whatever the sold-out supplier cannot take will flow to whoever holds merchant capacity. One supplier named there plans capacity of a hundred million lasers a year, targeting roughly two to one external versus internal fab. The equipment end carried the same signal: an optical networking equipment maker had said plainly on last month’s call that lasers were its bottleneck, and that same day another brokerage listed it as a candidate for diversifying supply.

On the other side, valuations were collapsing. The same day he posted three numbers from a silicon photonics company: market capitalisation around 829.7 million dollars, net cash around 788 million, book value around 847.5 million. Priced below book, priced near liquidation. A few months earlier institutions had put 400 million dollars in at twenty-one dollars a share, with warrants struck at 26.25; the stock now sits near 4.8, needing more than a triple to return to that financing price. He gave the condition a name meaning fear of getting obliterated, set against the fear of missing out everyone already knows.

Two more independent data points landed the same day. A burn-in test equipment maker beat on earnings, and he quoted the customers directly: ship as quickly as you can. Orders were arriving faster than revenue could be recognised, with acceleration from multiple silicon photonics customers. A niche memory maker held its call with gross margin above seventy-seven percent and operating margin near sixty, a thick pile of net cash plus inventory, and management saying DDR2 and DDR3 supply will not rise meaningfully in 2027 and prices should hold.

I stopped for a moment here. Orders accelerating, margins at highs, prices collapsing, all on one day. It is tempting to pick a side at that point, and he did not; he just posted both.

3 to 6 November: The Contract Terms Start Talking

These three days are the most valuable stretch of the week to me, because they show an inference walking from a procurement record to earnings confirmation.

The 3 November post covered an indium phosphide substrate contract between a US optical components maker and a Chinese germanium producer. Indium phosphide is the substrate material laser dies are built on, the very top of this chain. The shape of the deal: roughly 194.4 million dollars for 2027 alone, five-year term, about 970 million total before price adjustments, 180,000 wafers. He ran it backwards: that wafer count is far larger than the lasers its own modules would consume, so its laser capacity expansion runs bigger than what it has disclosed. That inference carries risk, since yields and die mix both move the conversion, and he hedged it himself with “seems”.

Three days later, earnings caught the inference. On 6 November the company disclosed a one to one and a half billion dollar long-term agreement with a hyperscaler customer, with a second in its final stages at more than five times the size, which works out somewhere between five and seven and a half billion. The key line sits here: that second agreement explicitly includes lasers. The company is moving from using them to selling them. Management’s own words were that demand is exceptional and unprecedented, and they were running out of superlatives.

One more item the same day looks underrated to me: a compute landlord signed a five-year contract carrying ninety percent prepayment of the accelerator capital spend. Market worry for weeks has centred on whether these companies can finance themselves, and a ninety percent prepayment term solves a large part of that question in advance.

The thread closed on 7 November. He noted that a memory maker has been stressing that its supply agreements differ from historical versions, with contractual price floors and binding commitment clauses written in. A reader drew an analogy to a power semiconductor company’s mutually agreed contract revisions years ago, and his answer was that the extrapolation does not hold: those were revisions both sides agreed to, and this round of agreements is a different instrument at the clause level.

Line those four up and they are four ways of writing the same thing. Prepayment percentage, term length, price floor, which items enter the supply scope — none of those fields move when prices collapse, because they were signed shut before the collapse. Everything I learned about demand durability this week came out of those boxes.

6 November: The Rules That Took Effect on 9 November

That evening he posted a batch of items with nothing to do with the industry, which will nonetheless change how the next few weeks read.

Taiwan is cutting the cap on borrowed-share short selling relative to order volume from thirty percent to three, a ninety percent reduction, and raising minimum collateral for margin short selling from ninety percent to one hundred and thirty. It also authorised a stabilisation fund of around 15.3 billion dollars to buy in the market. South Korea is waiving the 140 percent margin-loan collateral maintenance requirement for three months to reduce forced selling. China introduced a support package of its own. The first two take effect on 9 November, the day this piece goes out.

What I take from it differs a little from his read. These measures act on the mechanics of selling pressure, and change nothing on the demand side. The effect is that prices explain fundamentals less well for the next few weeks: longs will read a bounce as confirmation, shorts will read the same bounce as official support, and neither side gets new information. If you want to update a view in this stretch, the contract fields are where to look, and the quote is not.

Two information-quality episodes landed the same week, and they pair neatly. First, a report that a foundry was planning a 3D DRAM line at a customer’s request, denied by a spokesperson the following day — had it held, it would have made the foundry a competitor to the three big memory makers, which is a high-energy narrative. Second, an old 2023 search engine support contract expired without renewal, which headlines turned into one cloud provider getting beaten by another, when the actual reason was that older generation accelerators did not fit that customer’s use case. He called it frustrating to see market-moving headlines flying in both directions, amplifying volatility while adding no information. I agree, and I would add one thing: both episodes share a very short inference chain, jumping from an unconfirmed item straight to a conclusion, with no clauses, no quantities and no dates in between.

7 November: He Closes the Frame on Bottlenecks Versus Pricing Power

On the last day of the week he wrote two passages worth copying down.

The first concerns lasers themselves. High-power co-packaged optics lasers carry higher pricing premiums and higher gross margins than the lower-range continuous-wave lasers still stuck in a bottleneck, and fewer players can supply them, so a supplier that can is strategically more valuable. Then he shuts the gate himself: what matters in the end is net income. The premium, the scarcity and the strategic value all have to clear that one gate first.

In the second he switched industries for contrast. A gene synthesis company’s AI testing segment is growing more than a hundred percent year on year, which sounds as hot as the AI supply chain; yet the whole company’s revenue projection is a doubling by 2031, around fifteen percent a year. His closing line: not every bottleneck has the economic impact that hyperscaler capital spending delivers. What that line does is draw a border — two things can both be “stuck”, and one converts the tightness into its own net income while the other converts it into news.

His write-up of the compute landlord that same day is the structure I most want to borrow. He turned fundamentally positive on the company after an acquisition closed and earnings landed, with a stronger operating outlook and better access to financing. Then in the same passage he flagged that its six billion dollar at-the-market equity programme is largely used up with about 2.37 billion left, and that executive compensation tied to share price appreciation would align better with shareholders than compensation funded by dilution. Positive on the company and positive on shareholder returns, written as two separate sentences.

Getting This Far, the Field I Care About Most

If you hold a company that gets described as sitting on a bottleneck, where the shortage story has been running for a year while the stock halved in this drawdown, this week’s material will push you toward one question: where exactly did this go wrong?

I have no clean answer, but the week helped me break it into three questions whose order matters.

First, is the shortage in the company or in its layer? The difference lives in substitution cost. If a customer can solve it by qualifying a second supplier, the tightness belongs to the layer and the rent gets spread thin. If the customer has to change technical path to route around it, the tightness belongs to the company. This week handed over a ready demonstration: the moment the deal was announced, the buyer’s competitors went looking for second sources, and nobody went to redesign anything. Which says the tightness in the laser layer is partly “only these few have volume right now”, and only partly “only these few can build it”.

Second, has the tightness turned into price? This one has a mechanical answer: open the gross margin line. A shortage story running a year with a flat gross margin means the company captured order volume and not pricing power, and volume is the easiest thing in the world to dilute with new capacity. That niche memory maker’s seventy-seven percent gross margin and its management’s “supply will not rise meaningfully in 2027” are two faces of one thing, the first being the price evidence for the second.

Third, does the captured price stay captured? Upstream is raising prices too. That five-year, 180,000 wafer substrate contract states amounts before price adjustments, which means the upstream room to raise was reserved inside the contract. So on a single chain, a price increase shouted about in the middle layer can end up, one lap later, inside the net income of substrate and equipment makers. There is no quick answer here. What I do is check whether the company locked its own upstream into long-term agreements as well — anyone signing a five-year substrate contract at least knows where that risk lives.

The Other Place I Have No Answer: Being Right Still Means Waiting

In the 5 November post, he wrote that he had not been using much leverage, and that he added some into that day’s fall, with an explicit note that it was not advice to anyone else. Two days later, replying to a reader, he said the scale of the drawdown had been extreme and stressful, that he remained directionally long, and that he was hoping for a recovery in 2027.

Put those two together and what I see is that a framework and a position are two different objects. His framework can be right out to 2029 — that company’s capacity really is sold out through 2029. His position has to survive November 2026. And this week’s material spells out the shape of the waiting: the sold-out supplier’s shareholders wait until the 2028 close, the second long-term agreement was still in final stages at the time, and the first annual slice of the substrate contract lands in 2027. Not one of those dates falls this year.

He added an observation on listing timing that I read as the same skeleton from another angle: one AI company’s annualised revenue passed eighty billion dollars in October, above the market’s earlier expectation in the seventy billion range, and his suggestion was to delay going public while market liquidity is being drained, since a large listing in June already demonstrated once how a big deal pulls money out of growth names. Even with supply and demand both in place, timing stays its own separate risk field.

One Thing Worth Taking With You

Supply tightness does not become profit by itself. It has to clear two gates on the way: price, then cost.

Next time you see a company described as sitting on a bottleneck, pull up its latest quarterly gross margin and compare it with the same quarter last year, and write the difference in percentage points on a piece of paper. A year of shortage talk with that number unmoved, and the answer is in that column; if it did move, look one box further down and see whether its upstream raised prices right back. I picked up this habit from the phrase “before price adjustments” in this week’s substrate contract — those four words reminded me that upstream had already written its own increases into the paperwork.

For the first-hand version, look up Serenity’s account on X; the original threads from 2 to 7 November are all still there, including the inference chain posted the day of the deal and the details of those contracts.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.