저자: TomLee · 외국인
요약 보기
AI 우려로 증시 조정이지만 기술적 지지선 근처이며 펀더멘털은 유지된다고 주장. 브로드컴(AI매출 2027년 목표 1,150억달러, 2028년 2,300억달러 전망), 아스테라랩스(2026 EPS 117% 증가 전망), 마벨(2031년 800억달러 매출 목표)을 20~40% 조정 구간의 매수 기회로 제시. 시장 전반은 공포 구간이며 위험 대비 보상비율이 양호하다는 입장.
다듬은 글
You know, the market has kind of overcome the "wall of fear." This week, concerns about AI have intensified. And I think because it's Friday, there's a concern that with oil prices rising, tensions or war could escalate over the weekend. So it doesn't surprise me that the markets are pulling back today.
But I think before the audience thinks this is the beginning of a bigger downturn. I mean, the markets are at support levels. Our technical strategist Mark Newton believes that we could actually go higher. We hit rock bottom either today or Monday. And I think, from a fundamental perspective, as much as we should respect the view that AI debt markets are not very confident, the fundamental story and the driving forces of AI remain the same.
So to me, when fundamentals hold and markets are gripped by fear, it's actually a pretty good risk-reward ratio for investors. The stock market entered a high- volatility mode as Wall Street climbed a steep wall of fear. The Nasdaq Composite Index fell 2.1% on Friday, losing about 573 points to close at 26,140, snapping a four-week winning streak and capping a significant weekly drop of 4.5%.
The S&P 500 fell 1.1% to 7,659, ending the week down nearly two percent, while Wall Street's volatility gauge, the VIX, jumped above 22 during the day, hitting a 4-month high. A potent mix of soaring oil prices, interest rate concerns and looming geopolitical tensions has accelerated a large-scale exodus from risky assets.
Market anxiety hits specialized AI infrastructure: Shares of specialized GPU-based cloud services provider CoreWeave fell 11% on Friday, extending a brutal 25% weekly decline amid credit market concerns about debt financing. Semiconductor industry leaders also suffered heavy losses: Micron Technology fell 5.6% , and the PHLX Semiconductor index extended an 8-day decline.
However, Fundstrat's Tom Lee argues that this technical pullback offers a classic buying opportunity. Historically, in 2026, investors who ignored macroeconomic fears and bought quality technology assets during downturns were handsomely rewarded when fundamentals once again prevailed. Against this backdrop, the sharp decline in stocks of about 20-40% from recent highs has created attractive entry points for three giants of the semiconductor and communications industry: Broadcom, Astera Labs, and Marvell Technology.
With demand for specialized chips, rising AI capex, and next- generation optical networking goals, these three pressured companies are uniquely positioned to lead the next phase of AI growth. First on the list is semiconductor titan Broadcom, trading under the ticker AVGO, which has retreated about 28–31% from its June record highs near $480 to close around $ 328, creating a great opportunity for forward-thinking investors.
While macroeconomic concerns focus on funding constraints and debt market frictions, Broadcom's fundamentals demonstrate that the bar for achieving ambitious multi-year growth targets is much lower than the market assumes. At the heart of Broadcom's business model are two powerful engines: specialized AI accelerators known as XPUs, and market-leading AI networking hardware.
In the third fiscal quarter, AI revenue grew 221% year-over-year to $16.7 billion, driven by mass production of Google's TPU v7 and v8i chips, as well as the first shipments of specialized "jalapeno" silicon for Open AI. Management expects AI revenue to reach approximately $ 21.7 billion in the fourth quarter, which would put the company easily on track to reach its goal of $ 115 billion in AI revenue by fiscal year 2027.
In fact, from a Q4 base, Broadcom only needs about 12% quarterly growth in AI revenue to reach the $115 billion mark, a significant slowdown from recent 30% –54% jumps, indicating a significant margin of safety built into management. Importantly, AI networks account for nearly 27% of AI sales and are projected to grow as fast as specialized chips in the coming years.
Switch architectures such as Tomahawk 6, Jericho 4, and optical interconnects are deployed in almost every major AI cluster, including those running on competitor GPUs. Because networking equipment requires much less high-bandwidth memory than specialized chips, these high-margin sales help stabilize overall gross margins above 73% while generating strong free cash flow, which exceeded $ 13.7 billion last quarter alone.
Looking ahead, we see huge multi-year drivers that have yet to be fully tapped . The creation of multi-gigawatt specialized chips for Open AI, Anthropic, and Meta points to management's goal of doubling AI revenue to $230 billion by fiscal 2028. Trading at an attractive growth-adjusted valuation, relative to its $24 billion cash reserve and annual revenue growing to $106 billion, Broadcom remains a major AI powerhouse ready to capitalize once market fears dissipate.
Next on the list, moving from specialty silicon to high-speed server interconnects, is hyper-growth communications solutions provider Astera Labs, trading under the ticker ALAB. Although Astera has retreated about 40–43% from its 52-week peak near $500 to around $285, this sharp technical decline opens up an interesting entry point for investors focused on critical AI data center infrastructure.
Astera operates on a highly efficient fabless model , specializing exclusively in specialized connectivity hardware, including intelligent DSP retimers, memory controllers, and optical interconnect cables, enabling racks of powerful GPUs and CPUs to seamlessly interoperate without data transfer bottlenecks.
Recent performance metrics highlight the explosive fundamental momentum in building cloud infrastructure for AI. In its most recent quarterly report, revenue grew 104% year-over-year to $392.4 million, marking four consecutive quarters of roughly double-digit sales growth and bringing trailing 12-month revenue to more than $1.2 billion.
Non-GAAP earnings per share reached 80 cents, up 82% from a year earlier, driven by exceptional profitability, including a 75% gross margin and 30% net margin. The company maintains an impeccable balance sheet, with over $1.1 billion in cash and short-term investments and almost zero long-term debt, generating over $275 million in free cash flow per year.
The catalysts for future growth are accelerating along with an intense new product launch schedule. Astera recently introduced the Scorpio XM switch series, the Aries 7 PCIe 7 signal conditioning portfolio, and the Orion modular rack-scale architecture capable of supporting up to 160 AI accelerators per unit.
As hyperscalers ramp up deployment of their own chips, consensus analyst forecasts predict a 117% jump in earnings per share in 2026 to $3.99. At the same time, in 2027, profits will grow by another 62% to $6.48 per share, with revenue increasing by almost 95%. While Astera has a short history of public trading following its 2024 IPO and a high concentration of customers among key cloud partners, its indispensable position in high-speed GPU interconnect creates a powerful structural tailwind.
With rising spending on AI infrastructure and next-generation data center architectures , Astera Labs offers an attractive opportunity for investors looking to capitalize on the demand for high-speed equipment. Finally, optical networking leader Marvell Technology (ticker: MRVL) rounds out the top three, combining data center interconnects with specialty chip manufacturing capabilities.
Marvell has retreated about 21-25% from its recent local high of around $287 to $227 following a general pullback in the technology sector. Despite this temporary dip, Marvell's recent Investor Day outlined an extremely aggressive growth plan that puts the company at the heart of a multi-year expansion of AI hardware.
Essentially, Marvell provides data center interconnects, switching, and specialized integrated circuits for specific tasks, or custom XPUs. The company holds a dominant market share, approximately 60–70% , in PAM4 electro-optical digital signal processors— critical components that enable massive GPU clusters to process high-speed data over optical fibers regardless of chip architecture.
In the most recent quarter, optical interconnect revenue was approximately $1.4 billion, demonstrating how the core interconnect franchise scales directly in line with the overall capital expenditures of hyperscalers. Looking ahead, management has significantly raised its forecasts. Marvell now forecasts total revenue to reach $20 billion in fiscal year 2028 , up significantly from $8.2 billion in 2026, and foresees the overall market growing rapidly at a compound annual rate of 45% to $400 billion by 2030.
In the long term, management plans to generate $80 billion in annual revenue by fiscal 2031, driven by non-GAAP operating margins of 44-46% and projected earnings per share of over $30. Strong partnerships in specialty semiconductors provide significant growth potential. Thanks to strategic alliances with industry giants like Nvidia and Alphabet, revenues from specialized computing solutions and XPU chips are projected to exceed $12 billion in 2029 and approach $30 billion by 2031.
With significant operating leverage that drives profitability and a large amount of contracted revenue through fiscal 2028, Marvell looks highly attractive to investors looking to capitalize on the long-term growth of data center networking. While macroeconomic volatility may affect market sentiment in the short term, the underlying demand driving the development of remains unchanged.
The significant decline in shares of Broadcom, Astera Labs, and Marvell Technology opens up strategic opportunities to evaluate three proven leaders who are driving the development of dustry.
Comments 0