Bull markets like to pretend every red candle has a reason. On August 9, a trader who goes by Serenity published a note that cuts through that theater. He remains bullish on storage — Micron, Samsung — and argues the market rotates between supply bottlenecks rather than discovering new facts. Then he pivoted to photonics: AXTI, LITE, COHR. His point? The only thing that changed since July is price. Not the two-year laser backlog. Not the indium phosphide substrate shortage. Not the demand imbalance AAOI flagged on its latest earnings call. Price.
If you’ve spent any time watching Ethereum blob fees, that paragraph should sound familiar. It is the exact shape of every modular-DA drawdown we’ve seen this cycle: the bottleneck stays, the price gets marked down, and the market invents a new narrative to justify the marking.
The Note
Let me translate the equity trade before I translate it back to crypto. July was brutal for photonics. AXTI and LITE got sold down hard on macro noise and AI-capex jitters. The usual Twitter chorus said optical demand was rolling over, the AI trade was crowded, and the winners would be software, not silicon. Serenity is not having it. The market already knew in July that COHR and LITE had sold out their laser products for the next two years. It had already heard about the demand imbalance from AAOI’s earnings call. There was no new negative information. There was only a liquidation event. Serenity says this week the focus is back on photonics names like AXTI and LITE. Compared with July’s decline, the main change is price; the scattered narratives and business updates are just noise.
Then there is storage. Micron signed sixteen supply agreements and delivered an excellent forecast. Retail was euphoric. Now the same retail is capitulating. Serenity calls the ratio of operating profit to market cap "extremely unreasonable" and expects the demand imbalance to get more severe next year. Same company, same agreements, same multi-year demand. The only variables that moved are valuation and narrative.
Now replace "photonics" with "modular DA." Replace "laser backlog" with "blob consumption." Replace "indium phosphide substrates" with "Ethereum blobspace target." Replace "Micron supply agreements" with "L2 rollup integrations." The story is uncomfortably close. In April, modular DA was the market’s favorite bottleneck. By June, the same assets were "infrastructure without revenue." The code didn’t change. Blob utilization didn’t disappear. The fee market didn’t suddenly flood with supply. The market just found a new wall to run into.
The Bottleneck
Here is where my audit background kicks in. In late 2017, during the ICO boom, I wrote a Python script to parse newly deployed contracts on Ethereum mainnet. I found the integer overflow in Bancor before the official audit firms published it. The code was broken before the fix, and it was still broken during the panic. Only the narrative changed. That experience rewired how I read markets: the code doesn’t lie, but the market’s attention is selective.
My methodology has been the same ever since. I open the explorer first, read the contract second, and read Twitter last. That ordering is non-negotiable. During the Celsius collapse, I tracked treasury movements and found $230 million moving to a Huobi wallet before official reports existed. The hack rumor was false; the panic was real. We didn’t need a new narrative in June 2022. We needed a timeline. The same is true now. We didn’t need a macro excuse for July’s photonics drawdown. We needed a risk-limit explanation.
Look at the physical facts behind COHR and LITE. If laser products are sold out for two years, that is not a price target; it is a booked order. The bottleneck upstream is indium phosphide substrate production, which cannot be scaled on demand. It requires crystal growth, polishing, epitaxy, and months of customer qualification. You can’t spin that up in a quarter. The bottleneck in optical transceivers and indium phosphide substrates has not changed. It may have intensified. The price moved because someone had to sell, not because a fab suddenly materialized.
Storage is similar. DRAM and HBM supply agreements signed in 2024 become physical shipments in 2025 and 2026. Wafer starts were cut during the 2023 downturn, and capacity is not a switch you flip. Retail sees a stock falling and assumes demand is falling. The operating-profit-to-market-cap ratio says the opposite. The market is pricing today’s sentiment, not the next four quarters of committed demand.
In crypto, the equivalent raw data is available on-chain. Blob demand is not a whisper; it is a public good. The post-Dencun Ethereum setup gives us a target of three blobs per slot and a maximum of six. The exact numbers will change with future forks, but the structure won’t. Blobspace is finite, target-sensitive, and auctioned to the highest bidder. If a meaningful fraction of the L2s that announced blob migration actually follows through, the target fills up faster than the marketing docs suggest. When it fills, rollup fees stop being a rounding error and become a pricing signal.
Crypto’s version of operating-profit-to-market-cap is fee revenue divided by fully diluted value. Most people ignore it in a bull market because the next narrative is always more exciting. But it is the closest thing we have to a fundamental multiple. If a DA layer posts 10,000 blobs a day and its fee revenue is a rounding error, the bullish narrative depends entirely on future demand. That is fine, as long as you admit you are paying for future demand. But if the market was buying that same future demand in April and selling the same future demand in June, the rotation tells you more about the seller than about the demand.
Arbitrage is just patience wearing a speed suit. July’s photonics panic was not a signal that the backlog vanished. It was a liquidity gap for anyone whose thesis did not depend on this month’s chart. The same applies to modular DA: if your thesis depends on blob demand, a token drawdown is a markdown, not a reset. You can wait for the next monthly post-Dencun utilization report.
In early 2021, I built a bot to exploit the latency gap between OpenSea’s API and direct Ethereum node queries. I bought NFTs below the perceived floor for a week. Everyone else was watching a delayed window and calling it the market. That is what July felt like for photonics. Floor prices are opinions; volume is the truth. The laser backlog is booked volume. The blob fees are settled volume. The narrative is just the opinion layer on top.
Every time I watch a sector rotation, I repeat my own mantra: smart contracts are smart; humans are the bug. The rollup contract does not panic when its token falls 40%. The utilization schedule does not reprice because someone got a margin call. Humans read a red chart and rewrite the thesis to match the chart. The bug is not in the code. It is in the attention algorithm.
The Blind Spot
Here is the part Serenity did not need to spell out. The rotation between storage and photonics is not a rotation between sectors. It is a rotation between labels on the same physical demand. Micron’s HBM and COHR’s lasers go into the same AI server. LITE’s optical modules and Samsung’s memory sit in the same data-center aisle. The sectors are invented after the price moves, not before them.
Crypto does this even harder. Modular DA, alt-L1, Bitcoin Layer2, DePIN, AI-agent infrastructure — these categories do not represent distinct value chains. They are attention buckets. The same smart contract can be called a Bitcoin L2 after a website redesign. I have watched the Bitcoin L2 parade with a skeptical eye for a simple reason: most of those projects are Ethereum-style execution layers that wrap BTC and rebrand. The real Bitcoin community does not acknowledge them; a pitch deck does. That doesn’t make them worthless. It makes them narrative-first products, which means the narrative rotation will hit them harder than the code.
The same logic powers the liquidity fragmentation pitch. Fragmentation is not a new problem that appeared because the market is growing. It is the natural state of any market that rotates attention faster than it audits code. A fresh L2 that pitches you a solution to fragmented liquidity is selling you the problem its token needs. Serenity’s point about storage applies directly: when a stock falls without a fundamental change, ask what story the new story is selling. The answer is almost always the same product with a different multiplier.
The Trade
Watch the bottleneck, not the chart. For photonics, watch booked laser orders and InP wafer starts. For storage, watch supply agreements and operating profit relative to market cap. For crypto, watch blob fee per byte, rollup utilization, and the ratio between committed L2 demand and Ethereum’s target. If those numbers stay high, then the demand imbalance Serenity expects next year will hit exactly the places the market just sold. The rotation from storage to photonics and back is not a sign that supply chains healed. It is a sign that the market has a short memory. A cheap blob is a subsidy, not a breakthrough. Within two years, the subsidy burns off and rollup gas doubles. Liquidity leaves fast, but the smart money stays. So when the next rotation comes, ask yourself: did the backlog disappear, or did you just stop looking at it?