From an investment perspective, I’m watching optical and power names like $Lumentum(LITE)$ , $COHERENT(COHR)$ , $Ciena(CIEN)$ , $Corning(GLW)$ , VRT and CEG. However, I wouldn’t chase them purely on the AI narrative, especially after the strong rerating in some names. Earnings growth and actual order growth will be much more important from here.
Personally, I lean toward power as the longer-term constraint, because new generation and grid capacity can take years to build. The key question for me is no longer just “who sells the GPUs?” but who gets paid when every additional GPU creates another infrastructure bottleneck?
AI Is Starting to Fight for “Power” and “Light”: Is the Next AI Infra Trade Moving Beyond GPUs?
@Tiger_comments：
Two AI infrastructure stories are worth watching together today. On one side, Google is locking in power. The company plans to invest at least €13 billion in AI infrastructure in Finland over the next two years and has signed its first long-term nuclear power agreement outside the U.S. Under the deal, Google can purchase up to 50% of the output from one unit at Finland’s Loviisa nuclear plant for 22 years. On the other side, Verizon is locking in fiber. Corning has signed a multibillion-dollar long-term supply agreement with Verizon to provide more than 80 million miles of high-density fiber and connectivity products from 2027 through 2032. At first glance, one story is about nuclear power and the other is about fiber. But they are really answering the same question: Once hyperscalers keep adding GPUs, what becomes the next bottleneck? Increasingly, the answer is:Power and data transmission. More GPUs mean power becomes a hard constraint Google’s nuclear deal matters because AI data centers need enormous amounts of stable, round-the-clock electricity. Hyperscalers are no longer only competing for chips. They are also starting to lock in: nuclear power grid capacity long-term electricity contracts cooling infrastructure That is why the phrase “AI ends in power” is becoming less of a slogan and more of an actual capex reality. If the next wave of AI infrastructure continues, electricity supply may become one of the most important constraints on data-center expansion. The other bottleneck is bandwidth More computing power also creates another problem: How do all those chips move data between each other fast enough? That is where optical connectivity becomes critical. The Corning-Verizon agreement is mainly about fiber and connectivity, not optical modules themselves, but the market is trading the same broader theme: AI clusters need dramatically more bandwidth. The upgrade path is already moving from: 400G → 800G → 1.6T And the faster AI clusters scale, the more important high-speed interconnect becomes.This is why names across optical networking and components continue to get attention. Why are power and optics moving at the same time? Because AI infrastructure is moving into its second layer of constraints. The first problem was simple: Not enough compute → buy more GPUs. Then the next problems appear: GPUs need memory. GPUs need networking. Data centers need power. Servers need cooling. So the AI capex chain increasingly looks like this: GPU → HBM → Optical / Networking → Power → Cooling The market is no longer only asking:“Who is the next NVIDIA?” It is also asking: After NVIDIA sells more GPUs, which industries are forced to spend more because of that growth? Tiger View Tiger thinks today’s Google nuclear deal and Verizon fiber agreement are more interesting when viewed together. They both confirm one thing: AI capex is not disappearing. It is spreading outward from compute into infrastructure. The next three signals matter most: Do Google, Amazon, Meta and Microsoft keep signing more long-term power deals? When does 1.6T become a true large-scale hyperscaler deployment cycle? Can power and optical companies grow revenue fast enough to justify what the market has already priced in? There is one important risk: A strong industry trend does not mean every stock is cheap. Many optical and power names have already rerated significantly. So the next leg will depend less on the narrative itself and more on whether orders and earnings can continue to beat expectations. For now, Tiger would separate AI Infra into two questions: Compute: Is demand still expanding? Power and optics: Are the bottlenecks getting tighter? If Big Tech keeps locking in electricity, fiber and 1.6T capacity, the next phase of the AI trade may increasingly sit outside the GPU itself. Related Stocks AI Power $Constellation Energy Corp(CEG)$ $Vistra Energy Corp.(VST)$ $GE Aerospace(GE)$ $Vertiv Holdings LLC(VRT)$ Watch: hyperscaler power demand, nuclear exposure, grid investment and data-center power infrastructure. Optical / Networking $Lumentum(LITE)$ $Coherent(COHR)$ $Ciena(CIEN)$ $Corning(GLW)$ Watch: 800G-to-1.6T upgrades, AI cluster interconnect demand and long-term fiber buildouts. Upstream AI Capex $Alphabet(GOOG)$ $Amazon.com(AMZN)$ Watch: whether hyperscaler spending continues to spread from chips into power and networking. Today’s Poll What is the next major AI infrastructure bottleneck? ① Power / Nuclear ② 1.6T Optical ③ Cooling / Data-center equipment ④ GPUs are still the core trade For market discussion only. This is not investment advice. Markets involve risk, and investment decisions should be made carefully.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
41
Report
Login to post
No comments yet