Former Wall Street FICC Exec Herman Jin: AI Is an Industrial Revolution, Semiconductors Are Token Factories, and Long-Term Holders Beat Traders

Herman Jin, a former FICC executive director at a Wall Street investment bank, joined Binance Square's Inside The Blockchain 100 to share his frameworks for understanding the AI cycle, semiconductor investing, macro risks, and why retail investors may have a structural edge over institutions — if they resist the urge to trade. Jin's core thesis: AI is not a technology theme but an industrial revolution-level shift, and the investors who win this cycle will be those who understand how computing power converts into productivity rather than those chasing individual stocks or short-term entry points. Institutional Advantage Is Order Flow, Not Inside Information Jin's trading background shaped his most counterintuitive observation about institutional versus retail investing. After years on a rates desk within the FICC division — Fixed Income, Currencies, and Commodities — Jin said the real edge institutions have is not access to privileged information but visibility into order flow and positioning direction. "Our perspective was more about flow — who is buying and who is selling in the market — because we are dealers," Jin said. The core role of investment bank market makers is providing liquidity, not market manipulation. And information about positioning can be more important than any single news event. "Information on market positioning, or the direction of market positioning, can be more important than the information you have." On crypto specifically, Jin said he began trading Bitcoin seriously during the previous bull market, when fewer market makers meant more arbitrage opportunities across funding rates, coin-margined contracts, and USDT-margined contracts. He acknowledged he is more comfortable trading than holding Bitcoin long-term — his primary long-term beta exposure remains in equities including Google and other technology names. AI vs the Internet Cycle — Why This Time Is Structurally Different Jin's AI framework begins with a distinction that most investors miss: the internet revolution created platforms, while AI directly improves productivity. That difference makes all the difference for long-term economic impact. Long-term economic growth depends on two inputs — population growth and productivity improvement. As population growth has slowed globally, major economies over the past two decades have relied increasingly on liquidity expansion to sustain growth. AI matters because it may genuinely enhance efficiency at scale. "Starting from 2024, AI is an industrial revolution-level change," Jin said. "AI is the first time semiconductors, or technological progress, have been directly turned into productivity — because it becomes tokens." The Token Factory Framework — How to Think About AI Infrastructure Jin's most original contribution to the session was the Token Factory concept for understanding the AI supply chain. Computing centers produce tokens like factories. Those tokens can write code, create presentations, conduct research, and generate content — making them a new type of productive resource that can be consumed and priced like any other input. Under this framework, hardware iterations — advanced nodes, CoWoS packaging, CPO, memory, interconnects, and power supply — are no longer merely technology upgrades. They directly affect token production capacity and cost curves, making them the semiconductor equivalent of factory equipment upgrades. An improvement in interconnect efficiency is not a niche technical development — it is an increase in productive capacity. For evaluating which technologies are genuinely being commercialized versus which are speculative, Jin advised retail investors to focus on the roadmaps of major players. Nvidia, Google TPU, and the Broadcom ecosystem solutions reflect the real direction of the industry because their decisions are validated over years and are unlikely to change abruptly. When data center architecture, interconnect solutions, and power supply systems are chosen by hyperscalers, those choices define the supply chain for the cycle. Semiconductor Supply Chain — Memory, Compute, Optical, and Power Jin breaks AI infrastructure into four categories: memory, compute, optical, and power. GPUs provide parallel computing capacity. Memory handles data retention. Optical and copper interconnects improve cluster efficiency. Power systems support high-power data centers as cluster density increases. As AI clusters continue scaling, the importance of interconnects and power will rise relative to compute. In the past, interconnects accounted for a single-digit share of overall spending — a share Jin believes is structurally too low as ten-thousand-GPU clusters, 800V power systems, and CPO solutions advance. His analogy: compute chips are like brain cells, but without interconnects — the equivalent of gray matter — even the smartest chips cannot form an effective cluster. The investment framework therefore requires understanding CAPEX and ROI simultaneously. The capital expenditure and return on investment of major cloud providers are at the core of the current AI investment narrative. The market's key forward questions are: How much will cloud providers invest next year? Is their free cash flow sufficient? If financing is needed, how will financing costs and equity dilution affect valuations? Recent semiconductor volatility does not signal the end of the cycle — it signals the market waiting for hyperscalers to provide CAPEX and ROI answers. On Chinese versus US semiconductor companies, Jin noted both benefit from the same fundamental driver — strong demand from US cloud providers and AI data centers. Some Chinese names carry valuation premiums reflecting domestic substitution expectations and local liquidity dynamics. But the underlying demand still flows from global AI CAPEX. "We need to know where this wind is coming from." The Yen Carry Trade — The Macro Risk That Can Shock Everything On macro risks, Jin flagged the Bank of Japan, the yen, and the USD/JPY carry trade as the most significant structural threat to risk assets globally. Japan's long-standing low interest rates made the yen a global liquidity source — large amounts of capital have borrowed yen cheaply and deployed into dollar assets. A rapid yen appreciation or BOJ intervention could trigger a carry trade unwind creating short-term shocks across risk assets, including crypto. The July 2024 precedent is illustrative: a single BOJ rate decision triggered a yen carry unwind that sent Bitcoin from $65,000 to $50,000 in a week. Jin sees this risk as ongoing rather than resolved, with Japan's 10-year JGB yield now at a 30-year high of 2.85% narrowing the carry spread that makes the trade viable. However, Jin believes that against the backdrop of an AI industrial revolution, macro factors may become progressively less decisive. Market attention is increasingly shifting toward model revenue, cloud provider CAPEX, and semiconductor earnings realization — fundamental signals that are more durable than the liquidity flows that drove the prior cycle. Why 99% of Traders Have No Alpha — and How Retail Can Win Anyway Jin's most direct message for retail investors was also his most contrarian: stop trading, start holding. Intraday volatility is driven by machines and quantitative systems. Manual trading cannot generate sustainable alpha against high-frequency competition. "I believe 99% of people have no alpha — manual trading is very hard to generate alpha with." The area where retail investors can genuinely beat institutions is not short-term execution but long-term trend identification and the patience to stay with it. Institutions are constrained by NAV reporting, risk control frameworks, and drawdown limits that force them to cut positions at exactly the wrong moments. Retail investors face none of those constraints — giving them a structural advantage over a longer cycle if they identify the right direction, size positions appropriately, and withstand volatility without leverage forcing them out. "If you are skiing down a long slope, leverage can make you trip over a single rock," Jin said. If investors genuinely believe AI is a long runway — the way Jin frames it as an industrial revolution rather than a theme — the correct response is to manage positions, reduce unnecessary turnover, and avoid being forced out by short-term volatility. Holding through the institutional constraints that cause smart money to underperform over full cycles is the specific edge that patient retail capital has always had.