원문 보기
7.6 trillion dollars. This is the headline figure that Wall Street typically uses to estimate total global spending on AI infrastructure over the next 5 years, making it the largest physical capital construction in human history. However, every important indicator proves that this figure is already systematically underestimated.
The top four cloud hyperscalers spent approximately $245 billion in capital expenditures in 2024, exceeding $360 billion in 2025, and are now on track to exceed $500 billion annually. Combined with government AI initiatives around the world, corporate investment is completely disrupting traditional institutional models.
This relentless wave of capital directly coincides with leading tech analyst Dan Ives’s predicted $4 trillion AI infrastructure supercycle. Ives highlights the huge multiplier effect in the market, where every $1 spent on basic computing hardware triggers an additional $8–10 in direct spending in the broader technology ecosystem.
However, the main obstacle to growth is not a lack of capital. This is the speed of physical implementation. Hyperscalers are severely limited by power grid capacity, optical data transmission limits, and chip production queues. Instead of scattering capital among dozens of speculative stocks, the smart money focuses on four critical monopolies that control bottlenecks and collect the inevitable “tribute” from every dollar spent.
A giant in clean energy, a pioneer in optical interconnects, an architect of specialized semiconductors, and the undisputed computing engine of the entire AI revolution. First on the list to solve this huge infrastructure problem is Constellation Energy, ticker CEG—a nuclear power behemoth that is rapidly transforming into an indispensable energy platform for artificial intelligence.
AI data centers require vast amounts of uninterrupted, carbon-free baseload electricity. This is a critical resource that the current power grid simply cannot provide in a timely manner . Operating the largest fleet of nuclear reactors in the United States, CEG possesses an asset base that is virtually impossible to replicate.
Following the strategic acquisition of Calpine's large natural gas and geothermal assets, the company has expanded its capabilities to deliver clean and sustainable energy. It positions itself as the primary energy provider for hyperscalers seeking to get connected to the grid as quickly as possible.
The commercial impact of this positioning is already reflected in financial results. The latest financial results showed impressive quarterly dynamics: revenue grew by 23% to $7.5 billion , and adjusted earnings per share increased significantly. Management recently raised its full-year guidance to $11.50–$12.50, driven by strong commercial margins, portfolio optimization, and share buybacks.
Importantly, CEG secured 920 MW of long-term contract capacity with investment-grade customers at significant price premiums. These long-term agreements provide predictable and well-defined profits that are not dependent on overly speculative forecasts of future demand. In addition to current operations, important factors such as the planned restart of the Crane nuclear power unit and tax incentives for nuclear power generation are laying a solid foundation for future growth.
Although the stock has a higher multiple compared to other utilities, the market still underestimates its ability to monetize deficits , expand commercial profits, and capitalize on massive data center expansion. Providing power to these data centers is only half the battle. Data transfer between massive chip clusters brings us to the second bottleneck winner—Coherent Corp., ticker COHR.
This is a powerhouse player in optical infrastructure whose biggest manufacturing bottleneck has officially become its strongest bullish catalyst. The company simply does not have time to produce high-speed optical components to meet the demand of hyperscalers . As ultra-fast AI clusters become more complex, light, not electricity, is needed to transfer data between GPUs.
This transition transforms optical transceivers from ordinary networking equipment into an indispensable, highly cost-effective computing layer. Coherent's latest operating results vividly illustrate this huge wave of demand. Fourth-quarter revenue reached a record $2.05 billion thanks to a 66% increase in data center revenue, while adjusted earnings per share rose 74% to $1.74.
The demand forecast is extraordinary: capacity for the next financial year is effectively sold out, and long-term agreements with customers are in place until the end of the decade. To meet this unprecedented order backlog, the company is switching to 6-inch indium phosphide production, which yields approximately four times the output at half the cost.
This structural shift has helped push gross margins above 40.2%, opening up huge opportunities for leverage as production ramps up. In addition to high-speed transceivers, strategic partnerships, including the supply of high-power continuous-wave lasers for shared optics architectures, solidify Coherent's status as a must-have for next-generation AI connections.
With quarterly revenue set to exceed $3 billion over the next two years and the consensus EPS forecast to reach $18.47 by fiscal 2029, Coherent's current valuation multiple will decline rapidly as profitability increases. While optical channels carry data, designing specialized silicon for these workloads brings us to a third dominant player—Broadcom Inc., ticker AVGO.
A titan in specialized silicon and networking technology, dominating the hardware layer behind Big Tech's move to its own AI chips. While general-purpose GPUs continue to power the early stages of training, leading hyperscalers and AI labs are aggressively implementing application-specific accelerators (ASICs) to optimize costs and improve efficiency.
Controlling about 70% of this specialized chip market and simultaneously leading in high-speed data center networks, AVGO is at the center of the world's largest cloud infrastructure expansion. Broadcom's latest financial results demonstrate unprecedented acceleration and operating leverage. Consolidated revenue rose 86% year- over-year to a record $29.6 billion, while adjusted operating profit jumped 92% to $20.1 billion.
The momentum was led by the semiconductor division, where sales of specialized AI chips and networking equipment rose 221% to $16.7 billion . In addition to significant revenue growth, the business is generating exceptional cash flow, delivering a record $13.7 billion in free cash flow in a single quarter.
Which amounts to an extraordinary free cash flow margin of 46%. What makes the AVGO platform truly attractive to long-term investors is the multi-year visibility of revenue. Management stunned the market with a forecast that AI semiconductor revenue would reach $115 billion in fiscal 2027 and $230 billion in fiscal 2028 .
With guaranteed manufacturing capacity and long-term implementation commitments with leading technology giants, this massive growth trajectory demonstrates that the market is significantly underestimating Broadcom's long-term earnings potential. Finally, the stack is the undisputed sovereign platform driving the modern AI revolution—Nvidia Corporation.
Ticker: NVDA. While skeptics continue to look for signs of a slowdown, the computing giant just delivered another record performance, doubling quarterly revenue to $96.2 billion, thanks to an extraordinary $89 billion in data center sales alone. Demand for their equipment is so unstoppable that management took the rare step of releasing an early full-year forecast, predicting about 70% growth in fiscal 2028, clearly confirming that the scale of growth is limited purely by production capacity, not customer demand.
What sets Nvidia apart is that they no longer just sell individual accelerators. They design complete integrated computing systems. Thanks to a rapidly expanding ecosystem that encompasses central processing units, high-speed networks, advanced optics, and proprietary software frameworks , revenue per gigawatt of data center capacity is growing rapidly: from $18 billion with Hopper to $40 billion with Vera Rubin.
The company is aggressively expanding its reach, acquiring open-source developer platforms like Hugging Face for $12.9 billion, and expanding beyond public cloud hyperscalers into the enterprise , industrial, and sovereign AI markets. Trading at a surprisingly modest multiple of 15 times fiscal 2028 earnings relative to its nearly double-digit growth trajectory , Nvidia remains the main driver of .
Rather than chasing dozens of speculative software applications, concentrating capital in these four core monopoly bottlenecks—Constellation, Coherent, Broadcom, and Nvidia—allows us to capture an inevitable share of every dollar invested in global AI development. Take positions early, ignore the short-term noise, and ride this historic infrastructure wave.
Comments 0