卡拉比

CoreWeave and IREN both rose by around eight percentage points, with the elasticity of AI computing power leasing and data centers fully stretched this wave; Cerebras rose by two to three points, Seagate Technology rose by almost four points, with AI training chips and storage hardware moving together.

From computing power leasing, AI chips to data storage, the entire AI infrastructure chain was being bought that day. Pure computing power leasing companies like CoreWeave have greater elasticity, are also more sensitive to expectations of AI capital expenditure, and their volatility during pullbacks will similarly be more intense.

Zhipu AI surged 13%, bucking the broader market's pullback that day. Interestingly, the catalyst in recent days should have been negative—the first batch of cornerstone lock-up expirations post-IPO happened to fall in this window. However, about 70% of the unlocking parties (many being national strategic capital and local industrial funds) publicly stated they would hold long term and not reduce their holdings. The market directly interpreted "the lock-up expiration didn't cause selling pressure" as a vote of confidence.

MiniMax also rose 12%, with no specific news of its own, purely following this wave of sector rotation back into China's large AI models—a rebound after the previous halving. Moreover, it faces an even larger batch of lock-up expirations next week, making the future movement of its stock price uncertain.

A report by SemiAnalysis on CPO and 800V DC power architecture indicates that they expect 800V DC power to be delayed until 2028, while CPO will be delayed until 2028-2029. They believe NPO and 400V DC solutions will immediately become the biggest beneficiaries. One company mentioned optimistically in the report is Amphenol. This report directly triggered a major capital migration last night: the optical communication/CPO chain ($Lumentum(LITE.US) $Coherent Corp.(COHR.US) $Applied Optoelectronics(AAOI.US) $Corning(GLW.US) ) collectively saw outflows; $Amphenol(APH.US) rose 7.29% against the broader market to close at $154.07, touching $154.92 during the session.

The report's impact lies in the fact that it provides engineering details rather than opinions: a 95% yield for a single optical engine sounds decent, but a chip needs to integrate 32 engines, and the system yield after multiplication drops to only 19.4%. Without solving this number, CPO cannot scale profitably. On the 800V DC side, cloud providers are collectively taking a wait-and-see approach to NVIDIA's single-ended solution. Meanwhile, the two "transitional routes" of NPO and ±400V happen to be Amphenol's stronghold — technology roadmap delays never mean demand disappears, it just means the money moves to a different pocket.

Of course, NVIDIA issued an urgent refutation the same day: "No delay, mass production and delivery on schedule in the second half of 2026," with Spectrum-X photonics already in mass production. However, I believe the 19.4% yield is mathematics, the refutation is a posture, and engineering problems won't disappear because of a press conference.

As expected, the space stocks surged and then plummeted: $AST SpaceMobile(ASTS.US) was still hitting a 52-week high the day before yesterday, but then dropped 4% after-hours yesterday and plunged -10% pre-market on 5/29, closing at 133.

Before considering whether to catch the falling knife, first figure out why it fell:

I looked through recent news and couldn't find any fatal fundamental negative news—it's more about "it's been hyped too high and needs a breather": In the past two weeks, the entire space sector was ignited by SpaceX's $1.75 trillion IPO expectations and NASA's lunar base plans, with ASTS and small rockets surging consecutively. This kind of purely sentiment and expectation-driven rally, once there's no new catalyst, profit-taking happens this fast. The negative isn't "bad news happened," it's "the good story is over, and the valuation is too expensive."

However, I think: ASTS has the real catalyst of a June satellite launch plan as a cushion. It looks interesting after a pullback, but right now it's "the knife is still falling." I won't catch it now; I'll wait for it to bottom out and see the market sentiment around the June launch.

What's even more critical is $Tradr 2X Long ASTS Daily ETF(ASTX.US), which fell 20.8% pre-market. Leveraged stocks are a double-edged sword in a "pump and dump": great when rising, but your principal can be halved faster when falling, plus there's decay. My conclusion is clear: If you really want to bet on an ASTS rebound, only touch the underlying stock, not the 2x leveraged ETF—catching a falling knife is dangerous enough; adding leverage makes the volatility terrifying 😰