저자: CathieWood · 외국인
요약 보기
GPU 중심 AI 훈련 투자 열풍은 정점에 가까워지고, 추론·자율 에이전트 AI로 전환되면서 CPU 수요와 CPU:GPU 비율이 1:1 방향으로 재편된다는 주장. 이에 따라 AMD, 인텔, ARM, Flex, 테슬라를 유망 종목으로 제시. AMD는 CPU 매출점유율 46.2%, 2030년 서버 CPU 시장 1,200억 달러 전망, 1분기 매출 103억 달러(+38%). 인텔은 데이터센터·AI 매출 +22%(51억 달러), 어드밴스드 패키징 40% 마진 주장. ARM은 자체 데이터센터 칩 출시 후 주가 140% 상승, 고객 수요 20억 달러 이상. Flex는 클라우드·에너지 부문 매출 성장률 2027년 75%, 2028년 80%+ 전망. 테슬라는 2분기 인도 48만 대, 매출총이익률 21.1%, 마이애미 무인 로보택시 운영, 마일당 25센트 목표 제시.
다듬은 글
This is an important decision. Is this a hoax? Is this 1999? I can definitely say that this is not true. As I mentioned, the seed was sown back then. Now we are ready, and we believe that this technological revolution will far surpass the industrial revolution. I think there is more competition. We want this. We need this.
But Serafina was interesting at OpenAI. I want to congratulate her, because I heard her talk, and she said that people are chasing GPUs. They will be very shocked by how generative AI activates processors, and how logical inference generally activates processors. There are now four to five GPUs per processor when it comes to powering AI. Lisa thinks it will become one-to-one in the future. I think that was a dormant factor.
For the past two years, global financial markets have been in a state of absolute hyper-fixation, pouring hundreds of billions of dollars into a single asset class: Nvidia GPUs, graphics processors. Active investors treated these chips as the undisputed gold standard of the digital age. But according to ARK Invest founder Cathie Wood, a macro strategy pioneer who defied Wall Street by predicting Tesla's 1,000% breakout long before it happened, the entire market is looking at the wrong equipment.
Wood's core investment philosophy has always focused on identifying revolutionary convergent technologies before they become mainstream. Today her firm manages billions in assets of revolutionary innovations, and she sounds an unmistakable alarm. The initial phase of GPU-based AI training during the gold rush is nearing its peak, and an aggressive structural shift is taking place.
The numbers confirm this large-scale transition. Currently, data centers deploy an average of four to five GPUs per central processing unit (CPU). However, as AI evolves from training massive models to performing real-time inference and running continuous autonomous agent-based AI workflows, this computing landscape is changing radically. Industry forecasts signal a rapid decline in the equipment-to-equipment ratio. This is triggering a back-to-the-future phenomenon, aggressively reviving outdated 1990s tech giants that the market had largely written off.
We set the stage for an analysis of five specific stocks that fit perfectly with this compelling macroeconomic forecast, companies about which Wood herself is incredibly bullish. We'll analyze architectural geniuses, silicon pioneers reclaiming server racks, manufacturing giants building physical infrastructure, and an integrated mobility giant poised to drive transportation costs down to just 25 cents per mile. Get ready to refocus your portfolio. These are five stocks that will become weapons for the next era.
The first stock on our list is Advanced Micro Devices, ticker AMD. The architectural giant is ideally equipped for this large-scale computing transition. AMD is experiencing an unprecedented CPU supercycle, capturing an extraordinary record 46.2% of CPU revenue share as the market undergoes a rapid structural shift toward real-time model generation and autonomous agent-based AI.
While the consensus remains focused on the GPU arms race, running complex AI agents requires large-scale data orchestration, tool invocation, and API management: workloads that are inherently processor-specific, not accelerator-specific. AMD's foundational EPYC server franchise exploits this very inflection point. In fact, next-generation infrastructure is moving away from historical host ratios toward a balanced one-to-one configuration of CPUs and accelerators.
Driven by this structural hardware evolution, AMD recently doubled its forecast for the total addressable server processor market to an impressive $120 billion by 2030. The fundamentals are already working at full capacity, as evidenced by a massive 38% year-over-year revenue growth to $10.3 billion in the first quarter, fueled by data center expansion. The upcoming catalysts are incredibly powerful, led by the upcoming shipments of next-generation Zen 6 Venice processors and the highly anticipated Helios rack platforms. By combining impressive compute performance per watt with leading total cost of ownership, AMD is strategically positioned to capture a dominant share of the monetization of the growing $1 trillion global AI infrastructure stack.
Shifting our focus from AMD, the second stock directly at the center of this architectural shift is Intel Corporation, ticker INTC, the legendary silicon pioneer that orchestrated one of the most difficult transitions in technology history. While this giant was largely written off during the initial GPU training phase, the paradigm shift toward model inference and autonomous software agents has put its core x86 competency back in the spotlight.
As agent-based workflows require intensive coordination between memory, network, and storage, the historical hardware ratio is rapidly shrinking. This structural evolution is leading to a massive resurgence in the company's data center and artificial intelligence segment, which recently posted strong revenue growth of 22% year-over-year to $5.1 billion in the first quarter. Most importantly, the market is completely underestimating the explosive margins hidden in Intel's advanced packaging business. As AI chips become exponentially more complex, the company is winning premium deals with 40% gross margins from building AI infrastructure long before its front-end business is fully mature.
This massive internal operating leverage is steadily reducing the hurdles for traditional foundry companies, setting the stage for a sharp recovery in earnings, supported by highly resilient demand for enterprise computing, multi-billion-dollar long-term supply agreements, and accelerating infrastructure development cycles. This deeply integrated manufacturing giant is systematically repositioning itself as an indispensable anchor of the next-generation AI stack.
Coming out of Intel's manufacturing reorganization, our third discovery is ARM Holdings, ticker ARM, an architectural genius that is quickly becoming the most important non-graphics engine of this entire cycle. For years, this titan operated exclusively on a high-margin intellectual property licensing model. However, a massive operational turnaround completely transformed the business, sending shares up more than 140% after it launched its first-ever proprietary data center processor, co-developed with Meta.
This dual monetization mechanism, combining pure intellectual property royalty income with high-margin direct silicon sales, caught Wall Street completely by surprise, securing over $2 billion in initial customer demand for its new line of custom silicon processors. The company's ultra-efficient architecture is uniquely positioned to address the power crisis of AI-powered data centers. Its specialized infrastructure delivers more than twice the compute performance per rack compared to legacy x86 platforms, potentially reducing capital expenditures by $10 billion per gigawatt of deployment.
Whether tech giants build their own chips using licensed blueprints from the company, as seen with custom silicon development programs from Google, Amazon, and Microsoft, or buy its discrete hardware directly, Arm is winning. By earning up to $500 in gross profit from each custom-made chip, instead of the standard 5% royalty, this capital giant is creating unprecedented profits, ideally suited to the era of high-volume logic inference.
Moving from Arm's chip architecture to heavy industrial-grade AI, the fourth breakout stock on our list is Flex Limited, ticker FLEX, a manufacturing titan that Cathie Wood lovingly calls the fundamental foundation of physical-grade AI. For decades this company was considered a traditional contract electronics manufacturer, but a massive operational reassessment catapulted it directly into the S&P 500 as an indispensable player in AI infrastructure.
While the market is obsessed with software, this company is winning precisely where AI achieves specific goals: power, liquid cooling, and data center server racks. Its core cloud and energy infrastructure segment is booming, recently posting a staggering 38% annual revenue growth as hyperscalers seek to build physical computing platforms. The growth trajectory here is simply explosive. To address huge bottlenecks in the global power grid, the infrastructure giant recently made a vital acquisition to modernize the data center power grid, allowing it to act as a single, fully integrated partner for both utilities and tech giants.
Management provided stunning guidance, predicting revenue growth for its cloud and energy divisions of up to 75% in fiscal 2027, accelerating to over 80% in fiscal 2028. By spinning off this hyper-growing infrastructure division into a separate company, Flex is unlocking a huge value trap for shareholders, making it a flawless, significantly discounted investment for the real development of the next computing era.
Leaving Flex's physical server racks behind, our latest compelling investment is Tesla Inc., ticker TSLA, the cornerstone of Cathie Wood's macro thesis: a hyper-integrated mobility giant that is aggressively using vertical integration as a weapon to rewrite the rules of autonomous transportation. For years the mainstream market mistakenly viewed this titan solely as an electric vehicle manufacturer focused on raw production numbers. But the narrative has shifted into a dominant reality driven by artificial intelligence.
The company just delivered impressive second-quarter sales of over 480,000 vehicles, beating conservative expectations and delivering a strong increase in gross margin to 21.1%, before revealing huge seasonal operating leverage. The real inflection point lies in its vision-driven network of unsupervised autonomous robotics. The giant just reached a significant milestone by officially deploying its fleet of driverless robotaxis on Miami's public roads, operating completely unattended and without internal safety drivers.
While competitors face extreme scaling challenges due to cumbersome, fragmented supply chains and infrastructure overloaded with sensors, Tesla's vertically integrated data stack is projected to reduce real-world travel costs to just 25 cents per mile. With dozens of purpose-built, specialized autonomous taxis already deployed near major transportation hubs, and fully self-driving neural networks processing real-world data at exponential speeds, this titan has the potential to unleash a multi-trillion-dollar robotic monopoly that the broader market has fundamentally misjudged.
AI 산업은 거대한 구조 전환을 겪고 있고, 가장 큰 수익은 남들보다 먼저 이 전환을 포착하는 투자자들에게 돌아갈 것입니다. 오늘 이 인프라 핵심 기업들을 중심으로 포트폴리오를 다시 점검해 보시기 바랍니다. 좋아요와 구독, 그리고 아래 댓글로 여러분의 생각도 나눠 주세요. 다음 영상에서 뵙겠습니다.
Comments 0