CITIC Securities: Computing Power Sector Has Not Yet Reached Medium-Term Major Shift Criteria, Mainly Due to Short-Term Rebalancing Demand 中信建投：算力板块当前尚未达到中期大级别切换标准，主要是短期再平衡需求

Let's start with the facts. Computing power, especially AI computing power, is indeed the engine of the future world. Without it, large language models, autonomous driving, and smart manufacturing are all castles in the air. Absolute fundamental strength? I wholeheartedly agree. But the problem is that the entire market has seen this clearly, and what happened? Capital rushed in like sharks smelling blood, sending stock prices sky-high and packing the trading structure tighter than a subway car during morning rush hour. At that point, any little breeze — a regulatory document, U.S. tech stocks catching a cold, or even a major investor quietly taking profits — can trigger a wave of panic selling. This isn't investing; it's collective hysteria.

Tighter regulation? It's about time. Hot money in the market is frantically chasing concepts, inflating bubbles faster than balloons. If no one hits the brakes, it's always the late-to-the-game retail investors who get hurt. Profit-taking? Perfectly normal — you'd be a fool not to take profits when you've made enough. Adjustments in the U.S. and South Korea's AI supply chains? That only exposes our vulnerability: how many so-called "domestic computing power" companies still have their core technology and supply chains in someone else's hands? The moment there's external turbulence, their legs turn to jelly. So those who wail and doubt the trend at the first sign of short-term volatility have no idea about the underlying logic of the game. The trend hasn't disappeared — the market is catching its breath and shaking out the weak hands.

China Securities Co. says it's not yet time for a major mid-term rotation. I'm half-skeptical. Short-term rebalancing needs? In plain terms, it means things rose too fast and need to pull back to digest valuations. But what's the standard for a "major rotation"? They define it as a slowdown in computing power's own growth rate, or a significant recovery in other sectors' fundamentals. Is the former anywhere close? Look at the momentum of the global AI arms race — a slowdown in growth? Maybe in the next life. The latter is even more amusing. Other sectors like coal and industrial metals are just the last gasp of the old economy — how can they compare to the future that computing power represents? Shifting money from computing power to those sectors would be like abandoning electric vehicles to invest in horse-drawn carriage repair shops. What a brilliant thought process.

The most laughable is that "tech falls, defensives rise" rotation narrative. This is purely an excuse for short-term speculators. Real investors focus on the long arc of industrial revolution, not on staring at intraday charts doing Brownian motion. Today tech stocks plunge, so you sell and chase defensives; tomorrow tech rebounds, so you jump back in? All that churning, and transaction fees alone will eat away most of your capital. Market sentiment is like a pendulum, always swinging between excessive optimism and excessive pessimism, and most people make their worst decisions precisely when the pendulum reaches its extremes.

Wait for overseas tech stocks to stabilize before positioning in computing power? This is another typical brokerage talking point — it sounds like advice but is really just passing the buck. "Overseas stabilization" is a vague concept; by the time you've figured it out, the ship will have long sailed. The right time to position yourself is often when everyone is most fearful and the trade is most crowded — not when some clear signal appears. Those who keep saying "I'll buy on the pullback" often end up watching prices rise without them. As for watching coal and industrial metals as "other sources of economic vitality" — I nearly laughed out loud. That's as absurd as advising you in 2023 to stock up on pagers. Under the megatrends of energy transition and digitalization, the so-called "vitality" of traditional cyclical industries is nothing more than ripples from short-term supply-demand mismatches — how can that compare to the tidal wave of productivity revolution that computing power represents?

At the end of the day, the volatility in the computing power sector amplifies every market pathology: herding behavior, excessive trading, and short-sighted speculation. The fundamental strength is an open card, but the crowded trade is the real danger. Of those elbowing their way in, how many truly understand GPU architectures and the evolution of AI algorithms? Most are probably just mindlessly following the upward curve of stock prices. In such an atmosphere, it would be strange if volatility didn't intensify. Regulatory intervention, in a sense, is protecting the market from self-destruction.

Looking at it independently, I believe the current correction is not a bad thing. It makes prices more rational, cools down frenzied capital, and gives companies with genuine technological moats a chance to prove themselves. Once the foam is squeezed out, real gold will shine. But the process will be painful, especially for those using leverage or trading on tips. The market always rewards patience and insight, while punishing greed and ignorance.

So don't be fooled by China Securities Co.'s "inflection point signals." All that talk about shifts in valuation baselines and fundamental convergence is just after-the-fact rationalization — Monday morning quarterbacking. What you need to do is recognize that computing power is a high-conviction track for the next decade, then tune out the short-term noise, dig deep into the supply chain, and find the great companies that can emerge from the crowded field. As for which sector capital rotates into today or which concept surges tomorrow — let it go. Investing is not a 100-meter dash; it's a marathon. On a crowded track, the winners are those who survive and keep running to the finish line. The future of computing power is bright, but the road to that future is full of bumps and pitfalls. Buckle your seatbelt, grip the steering wheel, and don't get thrown off the ride.

算力板块最近的过山车行情，简直比好莱坞大片还刺激。中信建投那帮分析师，这次总算说对了一句话：核心矛盾就是行业基本面绝对优势撞上交易结构绝对拥挤。但这帮券商的话，听听就好，他们自己往往就是拥挤交易的一部分。

先说点实在的。算力，尤其是AI算力，确实是未来世界的发动机。没它，什么大模型、自动驾驶、智能制造，全是空中楼阁。基本面优势？我举双手赞成。可问题在于，全市场都看清了这一点，结果呢？资金像闻到血腥味的鲨鱼一样一拥而上，把股价推得老高，交易结构挤得跟早高峰的地铁似的。这时候，任何一点风吹草动——监管发个文件、美国科技股打个喷嚏、甚至某个大户悄悄止盈——都能引发一片恐慌性抛售。这不是投资，这是集体癔症。

监管加强？早该管管了。市场热钱疯狂追逐概念，泡沫吹得比气球还快，再不踩刹车，受伤的永远是后知后觉的散户。资金止盈？太正常了，赚够了不跑是傻子。美韩AI产业链调整？这更暴露了我们的软肋：多少所谓“国产算力”企业，核心技术和供应链还捏在别人手里？外部一有动静，这边立马腿软。所以，那些看到短期波动就哭爹喊娘、怀疑趋势的人，根本没搞懂游戏的底层逻辑。这不是趋势没了，是市场在喘口气，在洗掉那些不坚定的筹码。

中信建投说，还没到中期大级别切换的时候。这话我半信半疑。短期再平衡的需求？说白了就是涨太猛了，得跌一跌消化估值。但“大级别切换”的标准是什么？他们定义的是算力自身增速放缓，或其他行业基本面显著修复。前者现在远吗？看看全球AI军备竞赛的劲头，增速放缓？下辈子吧。后者更逗，其他行业像煤炭、工业金属，那是旧经济的回光返照，和算力代表的未来能比吗？把钱从算力切到这些板块，就好比抛弃电动车去投资马车修理铺，脑回路清奇。

最可笑的是那种“科技跌、防御涨”的轮动论调。这纯粹是给短线投机客找借口。真正的投资者，看的是产业革命的长线，而不是每天盯着盘面做布朗运动。今天科技股大跌，你卖了去追防御股，明天科技股反弹，你又杀回来？来回折腾，手续费都能把你本金啃掉一大半。市场的情绪就像钟摆，总在过度乐观和过度悲观之间摇摆，而大多数人，恰恰在钟摆摆到极端时做出最糟糕的决策。

中期看海外科技股企稳后再布局算力？这又是券商的典型话术，听着像建议，实则是推卸责任。海外企稳是个模糊概念，等你看明白了，黄花菜都凉了。真正该布局的时刻，往往是大家最恐惧、最拥挤的时候，而不是等什么明确信号。那些天天喊“等回调再买”的人，往往等来的是空涨和踏空。至于关注煤炭和工业金属作为“其他景气线索”，我差点笑出声。这就像在2023年建议你多囤BB机一样荒谬。能源转型和数字化大势下，传统周期性行业的所谓“景气”，不过是短期供需错配的浪花，哪能和算力这种生产力革命的浪潮相提并论？

说到底，算力板块的波动，放大了一切市场顽疾：羊群效应、过度交易、短视投机。基本面优势是明牌，但交易拥挤才是杀机。那些挤破头往里冲的，有多少人真正理解GPU架构、AI算法的演进路径？恐怕大多只是看到股价曲线向上就无脑跟风。这种氛围下，波动不加剧才怪。监管的出手，某种意义上是在保护市场免于自毁。

独立一点看，我认为当前的调整不是坏事。它让价格更理性，让狂热的资本冷却，给真正有技术护城河的企业一个证明自己的机会。泡沫挤掉后，真金才会闪光。但这个过程会很痛，尤其对那些用杠杆、听消息炒股的人。市场永远在奖励耐心和认知，惩罚贪婪和无知。

所以，别被中信建投的那些“拐点信号”忽悠了。什么估值中枢变动、基本面收敛，都是事后解释的马后炮。你要做的是，认清算力是未来十年的确定性赛道，然后忽略短期噪音，深入研究产业链，找到那些能在拥挤中跑出来的好公司。至于资金今天切向哪个板块、明天哪个概念暴涨，由它去吧。投资不是一场百米赛跑，而是一场马拉松，在拥挤的赛道里，活下来并坚持跑到终点的人，才是赢家。算力的未来光明，但通往光明的路上，满是颠簸和陷阱，系好安全带，握紧方向盘，别被甩下车。