저자: 미상 · 외국인
요약 보기
AI 학습에서 추론·사설 AI로 넘어가는 '산업화' 국면이라며, AMD(서버 CPU), 브로드컴(네트워킹·커스텀 실리콘), 델(AI 서버 조립)을 래리 핑크 발언과 연계해 유망 수혜주로 제시. AMD 데이터센터 매출 57% 성장, 브로드컴 AI 반도체 매출 84억 달러, 델 AI 서버 매출 342% 성장과 430억 달러 백로그 등 실적 수치를 근거로 듦. 투자 조언이 아니라는 면책 문구 후 포지션 설정을 권유.
다듬은 글
Before we proceed further, just a small announcement. I'll soon share and not financial advice. If you want might risk appetite. If you're sitting on the sidelines feeling like you've missed the AI gold rush , you're looking at the wrong map. The era of speculative chip flipping is over, but the era of industrialization of AI has just begun.
While the general public is still chasing the headlines of last year, institutional smart money has shifted its focus from the training phase to the high-stakes world of inference and private AI. In 2026, the game is no longer who can build the smartest model. It's about who can own the hardware that runs it locally, securely, and at scale.
Recently, BlackRock CEO Larry Fink signaled the end of this speculative era, declaring that we are witnessing the birth of a trillion-dollar asset class. As Fink put it, "A new asset class will buy the futures of compute , fundamentally redefining how global capital flows into physical infrastructure.
This isn't just a tech cycle. This is the creation of a sustainable economic base , where processing power has become the new global currency , as essential as oil or gold. " The statistics supporting this change are staggering. Hyperscalers like Microsoft, Meta, and Google are on track to deploy nearly $700 billion in AI infrastructure this year alone.
To dominate this cycle , you must own the pillars of the AI Factory. Today, we're analyzing three specific stocks that are the primary beneficiaries of this industrial shift. Provides an alternative high-performance engine that is breaking the GPU monopoly. The second acts as the nervous system controlling the networking and custom silicon that connects these clusters together.
The third builds chassis and integration systems, which are the actual server racks that transform raw power into operational intelligence. Let's delve deeper into the data. Let's start with the powerhouse that is providing the high-performance engine that is fundamentally breaking the single-source barrier in the data center , AMD.
For the past 2 years , the market was obsessed with GPUs. But, as we enter this phase of industrialization , Advanced Micro Devices has unlocked a huge secondary growth engine in the server CPU sector. Their recent Q1 2026 results were a big warning to doubters. Revenue reached $10.25 billion , a jump of 38% , but the real story was the data center segment , which grew 57% to $5.8 billion.
This demonstration confirms that the compute future recently revealed by Larry Fink is being built on AMD's EPYC architecture. What is really changing the landscape is the rise of agentic AI. We're moving away from the era where one CPU managed eight GPUs. In the new world of high-speed inference and localized personal AI, this ratio is moving towards one-to-one.
This structural shift has led management to revise the total market (TAM) for server CPUs to over $120 billion by 2030 , and some independent analysts estimate that number could reach $170 billion. AMD is uniquely positioned to achieve this because their next-generation Turin and Venice processors are specifically built for the intensive multitasking demanded by AI agents.
With 256 physical cores and 512 threads, these chips allow a single core to handle two tasks simultaneously , delivering twice the throughput per socket compared to leading ARM-based solutions. The financial profile here is as aggressive as the technology. While others are struggling with supply , AMD is displaying tremendous pricing power and average selling prices are expected to increase by 17% over the next few quarters.
This efficiency resulted in record free cash flow of $2.6 billion in the last quarter alone, representing strong margins of 25% and effectively reducing their balance sheet risk. As the industry moves toward personal AI and localized hardware , the path to earnings expansion toward $30 per share is becoming increasingly clear.
By providing the core engine for AI Factory, AMD is no longer just a challenger. It is the cornerstone of the new industrial infrastructure. As inference cycles accelerate, AMD's role in the rack is no longer optional. For any hyperscaler that wants to expand without relying on a single supplier , this is a mechanical necessity.
Moving from engines to the actual nervous system of the AI world, we reveal the second giant on our list. AVGO Broadcom is the company that serves as the invisible architect of the entire AI economy, dominating the high-speed networking and custom silicon needed to keep massive data clusters alive.
While many investors focus on the raw horsepower of standalone chips , Broadcom is the essential provider of the fabric that connects them. By mid-2026, the company has set a firm goal of reaching a staggering $100 billion in AI-related semiconductor revenue by 2027. This is n't just a guess, it's backed by deep, multi-year partnerships with six leading hyperscalers to develop custom AI accelerators , or XPUs, co-designed for specific cloud environments, like Google's TPUs and Meta's infrastructure.
Broadcom's recent performance proves this strategy is working in a big way. In the first quarter of 2026 alone, AI semiconductor revenue reached $8.4 billion , a triple- digit increase over the previous year. What's even more impressive is that as they expand this hardware , they're not compromising profitability.
Management maintained a consolidated gross margin of 77% , disproving Wall Street's common fear that hardware expansion would erode earnings. They achieve this by working at the cutting edge of the industry's technological boundaries, delivering high-speed SerDes and advanced packaging that others simply cannot replicate.
Moving beyond hardware, Broadcom is demonstrating its massive software prowess through VMware. Their latest release, VCF 9.1, is a game-changer for the production AI era we’re discussing. This allows enterprises to move away from the high costs of the public cloud and run complex inference and agentic applications on their own private secure infrastructure.
This software segment operates at an incredible operating margin of 78% , providing a high 'moot' recurring revenue stream that adds great stability to the business. The industry opportunity here is huge. Mega- cap hyperscalers are projected to spend nearly $700 billion on infrastructure this year, and Broadcom sits at the center of this toll road.
Whether it's the 5- year, $200 billion cloud deal between Anthropic and Google, which directly benefits Broadcom's TPU business , or the shift toward secure private clouds, the company's role as an indispensable provider of AI connectivity makes it a structural winner. If AI is the industrial revolution of our time, then Broadcom is the company that owns the tracks, switches, and signals.
To bring these components into a functional physical reality , the third essential stock we must discuss is Dell. In our framework of the AI Factory , Dell is the entity that takes high-performance engines from AMD and sophisticated networking from Broadcom and turns them into a working industrial-grade reality for the enterprise.
The figures here are nothing short of structural change. In the most recent quarter, Dell's AI- optimized server revenue grew 342% year-over-year, reaching a $9 billion run rate in just 3 months. This isn't just a jump. This is a complete shift in their business model from a cyclical hardware provider to a mission-critical pillar of the global AI infrastructure.
The most compelling part of the Dell thesis is the unprecedented visibility into future revenue. They entered the current fiscal year with a massive backlog of $43 billion in committed AI server orders. Unlike a speculative pipeline, this represents already-committed demand from hyperscalers and over 4,000 enterprise customers who are moving their proprietary data onto their hardware to run private AI.
Management is targeting $50 billion in AI server revenue alone for fiscal year 2027. And given the massive $ 64 billion order they recently received, it's likely they're being quite cautious. Dell's scale is their biggest security. Their deployment teams are 10 times larger than their nearest competitors , and Michael Dell's supply chain relationships give them a huge advantage in acquiring parts during this time of shortage.
Looking forward , the catalyst of agentic AI will only deepen this demand. As AI agents move from training to sustainable production workloads, the frequency of inference is increasing, requiring the same high-density infrastructure that Dell provides. Their latest 17G PowerEdge servers offer a 7:1 consolidation advantage over older models, allowing customers to replace seven older servers with one AI-optimized unit and cut energy costs by 65%.
Although rising DRAM and NAND prices pose a challenge to gross margins, Dell 's strong market position allows them to remain flexible on pricing. As enterprises around the world race to build their own local AI capabilities, Dell has gone from being a cyclical box manufacturer to becoming the structural foundation of the modern data center.
They are the tail end of the industrial AI revolution , and as long as companies value their data sovereignty , Dell's factory floor will be busy for years to come. In short, the transition from AI training to AI industrialization has become a trinity for the modern investor. AMD provides the optional compute engine, Broadcom builds the networking nervous system, and Dell assembles the ready-made AI factory.
Together, these three form the three important layers of Larry Fink's trillion-dollar asset class. We are no longer betting on which AI model will win the popularity race. We're betting on the infrastructure that makes each model possible. The data is crystal clear. The backlog is breaking records, and $700 billion in infrastructure construction has only just begun. Set your position accordingly. We'll see you in the next Deep Dive.
Comments 0