저자: EdYardeni · 외국인
요약 보기
AI 관련 신용 우려로 인한 하락은 과잉반응이며 펀더멘털(FIMO, S&P500 EPS 2027 $425·2028 $500)은 견고하다는 주장. 채권시장의 신용 규율은 1999년식 버블 붕괴를 막는 필터로 봄. 마이크론은 FY26 4분기 매출 $54.2B(+379%)·영업이익률 80%+, 저평가(PER ~5배)로 제시. 브로드컴은 주가 30%+ 하락 후 2027 EPS 기준 17배로 매력적이라 주장. Vertiv는 수주잔고 $15B, 2026 매출 가이던스 $14B로 데이터센터 인프라 병목 수혜주로 제시.
다듬은 글
Every day we get headlines that change everyone's mind. Wasn't it just a few weeks ago that we were all afraid that it would kill 10% of the world's population, and now we're worried that AI credit will become too expensive and we'll lose our competitive advantage? But look, I think there may be an overreaction here, and it's entirely possible that a company like Coreweave is a perfectly legitimate business and there's no reason to be concerned.
So I think the market will figure it all out. I think in the last few weeks we've kind of thrown off the 1999-2000 analogy of this bubble, and in the last week or so we've really shown a path that's sustainable in terms of access to credit for AI companies. What you just heard from Ed Yardeni is not a dress rehearsal for the funeral of artificial intelligence.
This is the sound of market sense returning to Wall Street. While retail investors panic over the sharp sell-off in speculative AI stocks, the bond market is quietly doing what the rating agencies have failed to do : impose discipline. Recent volatility in high-beta technology stocks was triggered by the yield on Coreweave’s 2032 unsecured bonds jumping to 15% and 5- year credit default swaps exceeding 1,100 basis points.
For speculative borrowers who have tens of billions in total liabilities and spend significantly more on capital investments than they generate in free cash flow, cheap debt is officially dead. But , as Yardeni points out, the fact that “bond vigilantes” are turning off the tap on bad debt is a necessary filter that prevents a full-scale 1999-style dot-com crash by directing capital back into real cash flows.
Behind this temporary pressure on asset valuations lies an extraordinary fundamental underpinning that Yardeni calls FIMO— fantastic earnings momentum. Hyperscale tech giants plan to spend more than $700 billion on infrastructure, fueled by huge pent-up demand that continues to exceed Wall Street forecasts.
S&P 500 earnings are forecast to reach $425 per share next year and rise to $500 per share by 2028. While high inflation and a rise in the 10-year Treasury yield to 5.1% may lower valuation multiples in the short term, a long-term path to 10,000 for the S&P 500 remains very real. The recent credit scare caused massive mispricing, dragging down debt-free industry leaders along with speculative junk .
To capitalize on this downturn, smart money isn't buying leveraged startups or expensive consumer platforms. They buy physical "bottlenecks" that the balance sheets of large tech giants are forced to finance. Specifically, we consider three category-defining companies . The only supplier of high-speed memory chips, without which artificial intelligence processors physically cannot function.
The semiconductor titan that controls the specialized silicon and networking devices that every major hyperscaler uses. And the undisputed leader in the field of liquid cooling and thermal energy infrastructure necessary for the operation of data centers. First on the list is Micron Technology, ticker MU, the cornerstone of the entire hardware ecosystem that powers artificial intelligence.
If GPUs are the central engines of machine learning, then high- bandwidth memory serves as a vital fuel system that keeps these engines running without limiting performance. The recent general sell-off in technology stocks has caused an exceptional underpricing of Micron , driven largely by short-term sentiment around the announcements of Chinese memory architecture from Changxin Memory Technologies and macroeconomic changes in profitability, rather than a deterioration in fundamentals.
Looking inside, Micron operates a highly profitable business model based on 1-gamma DRAM and G9 NAND memory nodes. In the fourth fiscal quarter of 2026, the company generated $54.2 billion in revenue, showing a staggering 379% year-over-year growth with an operating margin exceeding 80%. This structural profitability is underpinned by 26 strategic customer agreements that provide over 35% of Micron's total trading capacity by 2030, offering unparalleled revenue transparency.
Management recently noted that customer demand for memory supplies in fiscal years 2027 and 2028 will be significantly higher than in 2026, giving the firm enormous pricing power in both cloud data centers and enterprise applications. The fundamental thesis behind Micron's growth is centered around physical supply constraints.
Producing next-generation HBM3E and HBM4 silicon requires three to four times more wafer area than standard DDR5 DRAM, which inherently creates an industry-wide supply shortage that protects gross margins. Additionally, Micron's close partnership with Nvidia to implement custom HBM4E solutions, combined with the growing demand for enterprise SSDs for AI caching workloads, positions the firm as an indispensable beneficiary of hyperscaler capital expenditures.
Despite this huge profit-making potential, the stock is currently trading at an extremely low forward valuation of around five times earnings, a huge discount to the broader tech sector and its own multi-year growth trajectory . With a massive $150 billion expansion of manufacturing capacity in the U.S.
and international hubs, as well as long-term demand from agent-based AI workloads and autonomous vehicles requiring hundreds of gigabytes of RAM per unit, Micron offers one of the most attractive risk- reward profiles in the entire semiconductor sector. Moving from memory issues to custom computing architectures, our second company—Broadcom Inc., ticker AVGO, the semiconductor titan whose recent decline represents one of the most striking disconnects between market sentiment and financial reality.
While Wall Street has become nervous over headlines about customer concentration and fears about capital constraints at leading AI labs like Anthropic and OpenAI, Broadcom's performance has been exceptional. The recent pullback has pushed the stock down more than 30% from its peak, temporarily valuing this highly profitable company at a multiple close to the market average.
An absurd valuation for a company operating at the very peak of semiconductor innovation and enterprise is built on a powerful dual engine: dominance in co- designing AI accelerators and stable, highly profitable enterprise infrastructure software , where VMware plays a key role. In the third fiscal quarter of 2026, total revenue grew 86% year- over-year to $29.6 billion , driven by a whopping 221% growth in AI semiconductor sales to $16.7 billion .
Operating margin rose to a record 67.9%, generating $13.7 billion in quarterly free cash flow. Without relying on speculative debt, Broadcom maintains an A- credit rating with over $179 billion in liabilities, providing strong earnings predictability for years to come as management expects total revenue of $230 billion by fiscal year 2028.
The main catalyst driving Broadcom forward is a structural shift toward specialized chips for specific tasks, or XPUs. As capital becomes more expensive for hyperscalers and research labs, compute efficiency per dollar becomes paramount. Broadcom's specialized chips deliver significantly better performance per dollar than off-the-shelf GPUs, and their Tomahawk industrial Ethernet switches serve as a critical backbone for networks in massive cluster deployments.
Trading at just 17 times expected fiscal 2027 earnings and about 11 times the 2028 forecast, Broadcom offers a rare combination of high growth, strong cash flow, and a margin of safety in its valuation. As market concerns subside and multi-gigawatt capacity deployments continue in data centers , Broadcom is ideally positioned to deliver exceptional long-term returns.
Moving from chip design to the physical infrastructure that powers these massive computing clusters, our third and final company is Vertiv Holdings Co., ticker VRT, the undisputed leader in mission- critical power management and cooling systems for AI data centers. Although recent tech pullbacks have pushed Vertiv shares down nearly 20% from their peak, sparking skepticism among short-term traders, the company continues to grow earnings at an incredible pace .
Every high-performance AI accelerator in a modern server rack requires massive power consumption and advanced liquid cooling to prevent thermal shutdown. Vertiv is at the center of this physical infrastructure bottleneck, making it an indispensable player in delivering the “tools” for the entire digital ecosystem.
Vertiv's operational "moat" lies in comprehensive end-to-end integration capabilities that combine power management, precision cooling, modular infrastructure, and lifetime maintenance into a single ecosystem. According to the results of the second quarter of 2026, Vertiv received more than $ 3.27 billion in revenue, which is 24.1% more than last year, while the adjusted operating margin increased to 22.6%, and adjusted free cash flow to $ 925 million.
The company has a huge order backlog of $15 billion , providing almost a year and a half of projected sales. Management raised its full-year 2026 revenue forecast to $14 billion , reflecting 31% organic growth and a surge in demand among hyperscale customers. The primary catalyst for Vertiv's growth is the structural shift to high- density power architectures and liquid cooling.
Vertiv is working directly with industry leaders like Nvidia on next-generation 800V DC power systems for future architectures. Additionally , the strategic acquisition of Utility Innovation Group for $1.45 billion expands the company’s presence in microgrids and local energy generation, directly addressing the grid connectivity bottlenecks that currently hold up data center construction.
With approximately 85% of the company's revenue directly related to data center infrastructure, Vertiv operates in a powerful oligopoly alongside traditional industry players, but trades at a much more attractive growth-to- value multiple. As mega-cloud providers spend hundreds of billions of dollars of cash flow on physical data center infrastructure, Vertiv is uniquely positioned to transform its vast order book into sustainable revenue growth.
The bond market is not killing the artificial intelligence revolution. It separates high- risk debt traps from high-margin industry leaders. Which of these three financial strongholds will you add to your watchlist today? Let us know in the comments below. Like and subscribe to receive quality institutional-level market analysis .
Comments 0