AI infrastructure surged throughout Q1; in Q2, who can still sustain the "high valuation"?

Q1 The market lifted the entire AI infrastructure chain together, and by Q2, it will start reconciling segment by segment, company by company.

Written by: DaiDai, MSX MaiTong

Previously, MSX released a comprehensive overview forecast titled “Rising oil prices, stubborn interest rates, the Seven Sisters collapsing: What should you watch for in Q2 U.S. stocks’ excess returns?” which systematically sorted out the overall market main themes for Q2. Following this framework downward, you’ll find that the Q1 U.S. AI market rally was no longer just about “whether the leading computing power stocks rise or not.”

In the view of MSX Research Institute, the increase isn’t limited to a few GPU-related companies but encompasses the entire data center: expanding server rooms, increasing bandwidth, supplementing power supply, adopting liquid cooling, and pushing capacity forward.

Therefore, in Q2, the main theme remains unchanged, but the market rhythm will shift.

Networks, optical, storage, power, and equipment manufacturing still stay on the main line, but the subsequent trend won’t be a simple valuation lift along the entire “AI infrastructure” chain. Instead, it will focus more on whether orders, deliveries, profits, and capex can meet expectations.

From the performance range perspective, since Q2, the most resilient stocks are AEHR.M and AAOI.M; year-to-date, the gains of AAOI.M, SNDK.M, AEHR.M, and LITE.M are more prominent, clearly showing internal differentiation within the AI infrastructure chain.

1. Q1 The purchase wasn’t of a single company but of the entire data center expansion

The most obvious point of this Q1 rally is that funds, after buying in, no longer focus solely on individual chips. Chips are still the entry point, but the real capital expenditure driver is the entire data center. As long as hyperscalers continue to increase capex, money will follow the question of “what’s missing for AI deployment” downward.

Looking at the asset pool on the MSX platform, this main theme already has a relatively clear trading mapping.

The first to strengthen are networks. AI clusters are not a single-machine computing race but a contest of interconnection, bandwidth, and latency. Switch leader ANET.M naturally stands at the forefront, while CRDO.M, MRVL.M focus on key links around interconnection and accelerators, and AVGO.M, which also covers chips and connectivity, will be collectively lifted by the market; CIEN.M, although more on the network infrastructure periphery, will also be involved as long as cross-data center transmission and data transfer increase. Once clusters grow large, networks are no longer supporting roles but key expenditure items.

Next is optical. As AI training and inference move toward higher density, optical modules and related components become more prominent. LITE.M, COHR.M, AAOI.M, FN.M are repeatedly grouped into the same line in Q1, with straightforward logic: increasing bandwidth, upgrading interconnection, and pushing specifications higher. Previously, the market didn’t rush to pick winners or losers but first lifted the valuation of “specification upgrades.” Who is at 800G, who at 1.6T, and who can capture higher-end shipments—all these differences were temporarily set aside.

Storage is also reclassified. MU.M, WDC.M, STX.M, SNDK.M were previously viewed more as cyclical stocks, but in the Q1 market, funds have started to push them toward the “AI infrastructure chain.” The reason is simple: larger models, more data, more frequent training mean memory and storage are no longer just old logic for PCs and phones. At least at the trading level, the market is now willing to revalue these companies within the “data center expansion” framework.

Further down are power and thermal management. VRT.M, a typical beneficiary of “power supply + cabinets + supporting facilities,” and GEV.M, more focused on power equipment and grid infrastructure, may not be the hottest stocks every day in Q1, but they are also lifted by funds. After all, computing power can be packed into data centers, but where does the electricity come from? How is heat dissipated? Can delivery keep pace? These all need accounting. As data center density increases, none of these issues can be avoided.

Equipment manufacturing hasn’t fallen behind either. LRCX.M, KLAC.M, AMAT.M, MKSI.M, TER.M, TSEM.M, AEHR.M, this line, in Q1, was more about expectations that “capacity expansion will continue downward.” Advanced process, storage, packaging and testing, yield management, automation, and testing verification—once market participants include them in the “AI capacity expansion next phase” framework, valuations will move upward accordingly.

Thus, the Q1 rally wasn’t driven by one or two companies but by the entire chain being lifted together. The market first confirmed one thing: data centers need expansion. The subsequent network, optical, storage, power, thermal management, and equipment manufacturing sectors were all brought into focus.

2. In Q2, the main themes remain, but valuations start to reconcile segment by segment

The issue in Q2 isn’t that the AI narrative suddenly disappeared but that the initial “whole chain valuation lift” has already been traversed. Further gains are possible, but they won’t be as orderly as before.

In Q1, the network and optical chains benefited from expectations of high bandwidth, interconnection upgrades, and cluster expansion. By Q2, these expectations won’t vanish but will be broken down and calculated separately. Companies like ANET.M, with clearer customer bases and steadier product cycles, will be scrutinized for whether order momentum can continue; firms like CRDO.M, MRVL.M, AVGO.M, which are positioned around interconnection and computing power, will be examined for customer structure, shipment rhythm, and revenue recognition; and more infrastructure-oriented names like CIEN.M will focus on new orders, delivery cycles, and project progress, making it unlikely to continue enjoying the entire chain’s premium valuation.

Optical chains are more likely to diverge. Because even with the same emphasis on specification upgrades, the underlying customers, capacity, yield, and pricing pressures differ. LITE.M, COHR.M, AAOI.M, FN.M may rise together initially, but in Q2, the market will start to evaluate segment by segment: whose customers are more stable, whose shipments are smoother, and whose profits can be retained. The previous phase relied on “everyone buying along the optical chain,” but now it’s about who can keep the books balanced.

Storage will be more complicated. In Q1, the market was willing to assign higher “AI-related” valuations to MU.M, WDC.M, STX.M, SNDK.M, but in Q2, the focus shifts to how hard the demand really is, whether profit recovery can keep pace, or if it’s just another cyclical rebound with a different label. As long as prices, bit shipments, and profits move upward together, valuations will continue; if earnings reports aren’t solid enough, a sharp pullback will follow. The biggest risk for storage is being initially bought as AI infrastructure and then quickly reclassified as cyclical stocks.

Power, thermal management, and equipment manufacturing are more straightforward in Q2. They may not be the hottest topics daily, but during reconciliation, delivery, capacity expansion, order visibility, and gross margin trends become clearer. VRT.M’s order and delivery rhythm and GEV.M’s broader power equipment cycle will be compared against hyperscaler expansion plans. On the equipment manufacturing side, LRCX.M, KLAC.M, AMAT.M, MKSI.M, TER.M, TSEM.M, AEHR.M will be watched: whether AI capacity expansion can continue downward, and whether testing, validation, and final delivery can keep pace, or if they remain only at the upstream investment stage.

Initially, the market bought the entire sector collectively; later, it will need to evaluate company by company. The main theme persists, but valuations will no longer move in unison.

3. In Q2, the five key focus areas for funds are about anchoring high expectations to financial reports

First, look at orders. The network chain is usually the first to be checked: Are orders still piling up? Is delivery accelerating? Are customer procurement patterns shifting from trial orders to regular ones? Companies like ANET.M, as typical beneficiaries, will be repeatedly scrutinized for whether order strength can continue; firms like CRDO.M, MRVL.M, AVGO.M, positioned around interconnection and accelerators, will be examined for visible projects, clearer shipment guidance, and order flow.

Next, customer concentration. Optical chains are especially sensitive to this. The more concentrated the customers, the faster Q1 tends to rise; in Q2, if one major customer slows down, the stock price often reacts ahead of earnings. For stocks like LITE.M, COHR.M, AAOI.M, FN.M, the market cares more about who the customers are, their share, and the shipment rhythm, rather than just smoothing out fluctuations with “bandwidth upgrades.”

Third, capacity ramp-up. In Q1, many companies emphasized “strong demand,” but in Q2, funds focus more on “can you produce and deliver?” Optical looks at ramp-up, power and thermal management focus on delivery, and equipment manufacturing on order-to-shipment cycles. VRT.M, which is delivery-focused, will be re-priced by the market: as long as deliveries are steady, expectations won’t easily fall apart. The same applies to LRCX.M, KLAC.M, AMAT.M, MKSI.M, TER.M, TSEM.M, AEHR.M—if demand truly declines, it ultimately shows up in production lines, testing, and delivery.

Fourth, gross margin. When valuations were lifted in Q1, the market was willing to overlook profit margins temporarily; in Q2, it becomes more critical. MU.M, WDC.M, STX.M, SNDK.M are typical: strong demand is one thing, but whether profits can keep up is another. Prices may rise, but costs also increase, or competition compresses margins, causing valuations to drop first. Upgrades in specifications don’t automatically translate into higher profits; those who can turn higher-end shipments into higher gross margins are more stable.

Finally, hyperscaler capital expenditure. In Q1, the market assumed “big companies will keep investing”; in Q2, attention shifts to “where are they investing, what are the priorities, and are budgets being shifted?” As long as capex remains high, this chain remains active; but if the structure of capex changes, the beneficiaries’ order will shift quickly. As inference shifts toward reasoning, the rhythm of networks and optical will change; data center expansion pace shifts, and the order of power, thermal management, and equipment manufacturing will also adjust accordingly.

Overall, when capturing sector rotation opportunities, the core isn’t about prediction but about acting swiftly when signals appear. Currently, the market is focused on aerospace and defense, and MSX’s ongoing commercial aerospace trading week offers a good entry point. For traders interested in participating in the 10k USDT prize pool incentive and the sector, it’s also a practical opportunity to lower friction costs.

Conclusion

In Q1, the entire chain rose together; in Q2, the market will start reconciling segment by segment—networks, optical, storage, power, equipment manufacturing—no longer moving in unison. Whose orders land first, whose deliveries follow, whose profits materialize—those are the stocks more likely to stabilize their valuations.

Mapped onto tradable assets on the MSX platform, this main theme already has a relatively complete observation framework. MSX will continue to track the evolution of this theme and the rhythm of related assets.