Hook
On the morning of March 14, the on-chain analytics account Serenity posted a public portfolio snapshot: a 49.4% drawdown from peak, yet the caption read, “Remains optimistic about the artificial intelligence bottlenecks and restricted domains.” The numbers were staggering—a 4,502% return in 2026, now halved. But the assets in question weren't AI stocks; they were blockchain infrastructure tokens tied to hardware: AXTI (a semiconductor substrate supplier), SIVE (a photonics chip developer), and AAOI (an optical transceiver manufacturer). This wasn't a crypto-native portfolio of DeFi blue chips or memecoins. It was a concentrated bet on the physical layer of decentralized compute networks—a thesis that the next wave of blockchain adoption will be bottlenecked not by software but by the physics of data centers.
Context
Serenity's strategy exemplifies a growing archetype in crypto investing: the “hardware maximalist” who believes that the scalability of all decentralized networks—from Ethereum to AI co-processors—depends on overcoming manufacturing constraints in optics, materials, and packaging. The three tokens in question are not blockchain projects themselves; they are publicly traded companies that produce components critical for high-speed optical interconnects, which underpin the infrastructure that AI models and blockchain validators alike rely on.
AXT Inc. (AXTI) dominates the market for indium phosphide (InP) substrates, the foundational material for laser chips used in 800G and 1.6T optical modules—the very modules that connect GPU clusters in both AI training and proof-of-stake validator networks. Sivers Semiconductors (SIVE) develops silicon photonic chips for low-latency interconnects, promising to reduce power consumption in cross-shard communication for sharded blockchains. Applied Optoelectronics (AAOI) manufactures high-speed transceivers, which are the physical pipes that carry data between nodes.
As blockchain networks like Ethereum, Solana, and Celestia push toward higher throughput and data availability sampling, the demand for such components skyrockets. Serenity’s thesis is that by 2027 H2, the combination of AI inference scaling and blockchain data availability demands will trigger a hardware procurement supercycle, lifting these small-cap manufacturers into profitability.
Core: The Infrastructure Bottleneck Thesis
Let me be blunt: this is not a typical crypto strategy. Most traders chase narrative-driven tokens or yield farming. Serenity is betting on a physical bottleneck—and my own audit experience tells me that bottleneck is real but misjudged in timing.
The core insight lies in the physics of data. Every blockchain transaction, every smart contract call, every zero-knowledge proof generation consumes not only computation but also bandwidth—optical bandwidth. As we move from monolithic chains to modular stacks, the communication between execution layers, consensus layers, and DA layers becomes the critical path. Ethereum’s Danksharding, for instance, requires each blob to be distributed across thousands of validators. The network’s capacity is ultimately limited by the speed at which validators can propagate signed attestations—and that speed depends on the underlying optical infrastructure.
Serenity’s bet is that by 2027, the industry will hit a hard ceiling where software optimizations alone cannot squeeze more performance. At that point, hardware providers like AAOI and AXTI will see an inflection in orders. His 49.4% drawdown, then, is not a failure of the thesis but a volatility tax—a classic mark-to-market hit during a period of macro uncertainty and rotation out of growth stocks.
From my experience auditing DeFi protocols in 2020, I learned that infrastructure plays are always early. Compound’s governance token took 18 months from launch to flip from being a governance token to a financial asset. Similarly, the hardware cycle for optical components is 12-24 months from design win to volume shipment. The market is currently pricing in the risk of that delay. Serenity’s cost base is so low—given his 4,502% run in 2026—that he can withstand the volatility. But for a new entrant today, the risk-reward is treacherous.
Contrarian: The Case Against the Bottleneck Thesis
Here’s where my ENTP brain kicks in: the bottleneck thesis may be right in direction but wrong in magnitude. The drawdown might be a signal that the market is losing faith in the timeline, not just grinding through a correction.
First, the optical transceiver market is notoriously competitive. AAOI competes with Zhongji Innolight (listed on Shenzhen) and Coherent, both of which have larger balance sheets and scale advantages. If they choose to cut prices to capture market share, AAOI’s margins—already under pressure—could evaporate, pushing its revenue inflection point into 2028 or beyond. Serenity’s optimism assumes a rational oligopoly, but history shows that hardware markets often become bloodbaths.
Second, the technology itself is being disrupted. Co-packaged optics (CPO) promises to integrate the optical engine directly onto the ASIC substrate, eliminating the need for pluggable transceivers. If CPO becomes mainstream by 2028, companies like AAOI that rely on pluggable form factors may face obsolescence. Serenity’s bet on SIVE (which does silicon photonics) hedges this risk, but SIVE’s revenue is still heavily dependent on military contracts, not data center AI.
Third, the blockchain demand thesis is speculative. Yes, Ethereum’s Danksharding requires validators to handle many blobs. But the bandwidth requirements for a single validator are modest (a few megabits per second). The real bottleneck is on the proposer side, which is already solved by centralized relays. Unless blockchain adoption explodes beyond 100 million daily active users, the hardware demand from crypto alone may not move the needle.
Takeaway: The Art of the Inflection Bet
Serenity’s approach is a high-conviction, low-certainty bet on the physical layer of decentralization. It mirrors the early days of Bitcoin mining when a single ASIC supplier could make or break a miner. Today, the bottleneck is optical; tomorrow, it could be power or cooling.
For the institutional bridging I do now, I see this as a teachable moment: the market is rewarding patience but punishing over-leverage. Serenity can afford to be philosophical because his entry price is a fortress. But if you’re reading this as a retail investor, remember that true ownership begins where the server ends—and right now, that server’s optical cables are taut. They could snap or sing.
Debate is the compiler for better consensus. Let’s watch the 2027 H2 earnings calls closely.