美股投资young

美股科技投资者 ai狂热者分享每日快讯. all in ai

300Following

695followers

Feed

Livestream

Feed

$COHR Coherent's data center and communications business revenue reached $1.615 billion in Q4, with Nvidia also investing $2 billion in it. But what is more worth studying is that this company’s revenue for fiscal year 2026 exceeded $7.1 billion, yet operating cash flow was only $79.5 million. Why is there such a large gap between revenue growth and cash generation capability for a company positioned at the core of the AI optical communications industry? Coherent is a U.S. photonics technology and advanced materials company, with its business mainly divided into two segments: data center & communications and industrial. Many people simply understand it as an optical module company, but in fact, Coherent’s products cover indium phosphide lasers, VCSELs, photodetectors, high-speed optical transceiver modules, optical materials, industrial lasers, and other optical systems. It can produce some underlying optical components as well as provide integrated optical communication products. This puts it in competition with companies like $LITE and $AAOI, but each company’s manufacturing depth, product portfolio, and customer structure are not exactly the same. According to the Q4 fiscal 2026 financial report, Coherent’s revenue was $2.045 billion, a year-over-year increase of about 34%, with a GAAP gross margin of 38.5%, non-GAAP gross margin of 40.2%, and non-GAAP earnings per share of $1.74. Among this, data center and communications business revenue was $1.615 billion, up about 59% year-over-year, accounting for about 79% of total quarterly revenue; industrial business revenue was $431 million. Full-year revenue reached $7.118 billion, with data center and communications contributing $5.275 billion. It is clear that Coherent is increasingly relying on growth in AI data centers and communication infrastructure. Its technological advantages mainly manifest in three areas. First is indium phosphide (InP) manufacturing capability. InP is an important material for high-speed optical communication lasers. With increasing demand for 800G and 1.6T optical modules, advanced laser production capacity is gradually becoming a key constraint in the industry chain. Coherent is expanding its 6-inch InP wafer manufacturing facility in Sherman, Texas, USA. Compared to traditionally smaller wafers, larger wafers have the potential to increase production volume per batch and reduce unit costs, but whether a sustained advantage can be formed depends on yield, capacity utilization, and equipment depreciation. The second advantage is vertical integration capability. Coherent not only provides lasers but also has optical modules, photonic devices, and related material technologies. Customers can purchase complete products or select key components. For large data center customers requiring long-term stable supply, supply chain control capability is very important. The third growth direction is CPO and silicon photonics technology. As AI clusters continue to expand, traditional pluggable optical modules may not meet all connectivity needs. Co-packaged optics attempts to integrate optical engines more closely with switching chips to improve bandwidth density and power consumption. Coherent is developing related lasers and optical devices for these architectures. In March this year, Nvidia announced a $2 billion investment in Coherent and signed a cooperation agreement including multi-year purchase commitments and future capacity arrangements. Nvidia’s support for both Coherent and Lumentum also indicates it wants to expand advanced optical supply rather than bind the entire system to a single vendor. However, Coherent’s biggest investment controversy lies in cash flow. In fiscal 2026, the company’s operating cash flow was about $79.5 million, while capital expenditures reached $1.103 billion. Calculated as operating cash flow minus capital expenditures, free cash flow was about negative $1.023 billion. This contrasts sharply with the company’s net profit attributable to Coherent exceeding $800 million for the full year. Factors causing this difference may involve working capital occupation, expansion investments, and other cash flow items, so it cannot be simply assumed that the company’s business is not profitable. But it indicates Coherent is still in a capital-intensive expansion phase. In contrast, Lumentum’s operating cash flow for the same fiscal year was about $751 million, capital expenditures about $451 million, already generating positive free cash flow. Although both companies benefit from AI optical communications growth, their short-term cash flow structures differ. Coherent expects Q1 fiscal 2027 revenue to reach $2.2 billion to $2.4 billion, with non-GAAP gross margin of 39.5% to 41.5%. If future revenue growth can continue to drive margin improvement and new capacity gradually comes online, cash flow may improve. Risks come from laser capacity expansion being too rapid, customer demand falling short of expectations, intensified product competition, and rising manufacturing costs. Even if industry long-term demand is upward, it does not mean every expansion project will achieve ideal returns. In the long term, Coherent has a complete technology system of advanced optical materials, InP manufacturing, lasers, and high-speed optical modules, and its industry position is worth attention. But from an investment perspective, the most important verification in the coming quarters is not how much revenue can still grow, but whether that growth can gradually convert into stable, sustainable free cash flow. AI optical communications remain a direction worth long-term study, but $COHR’s current financial structure determines that its expansion efficiency and capital return rate require more attention than pure revenue growth.

$BE This earnings report is already strong enough, but the market still chooses to sell off. On October 22 Eastern Time, the company announced third-quarter revenue increased by 138% year-over-year, adjusted EPS rose from $0.15 last year to $0.94, and the full-year guidance was raised again, yet the stock price continued to weaken after hours. In Q3 2026, $BE's revenue reached $1.236 billion, exceeding market expectations of about $1.07 billion. Adjusted EPS was $0.94, also significantly above the expected approximately $0.69. Product revenue reached $1.074 billion, up 179% year-over-year, adjusted gross margin improved to 34.5%, and adjusted operating profit reached $286 million. This growth is not simply driven by market expectations but reflects actual business expansion shown in revenue and profit. $BE's core product is solid oxide fuel cells, which can generate power directly near data centers, reducing reliance on the traditional grid expansion schedule. As the power demand of large AI data centers continues to increase, on-site power generation is becoming an important choice for some customers. The company's customers and projects are also expanding. Management mentioned progress in cooperation with companies like Oracle and Nebius, as well as AEP's plan to use 700MW of Bloom equipment supporting carbon capture to supply power to Google. The full-year revenue guidance has been raised to $4.4 billion–$4.6 billion, with adjusted EPS expected at $3.10–$3.40, significantly higher than previous forecasts. Management also stated that backlog continues to expand, growing faster than revenue. However, there is an issue in this earnings report that cannot be ignored. The company’s operating cash flow this quarter was negative $27.15 million, free cash flow was negative $181 million, and capital expenditures reached $153 million. Although there is still about $2.5 billion in cash on the books, inventory, contract assets, and accounts receivable have grown significantly, indicating that business expansion is consuming a large amount of capital. The growth on the income statement looks very good, but cash collection and capacity expansion still require time. So $BE’s current problem is no longer about demand but whether future growth can be continuously realized and how much valuation the market is willing to assign to these future profits. This is also the contradiction facing the entire AI infrastructure sector. Corporate profits are still growing rapidly, but stock prices have already begun to be repriced. Long-term optimism about AI power demand is not an issue, but in the short term, no matter how strong the earnings report is, it may not stop the market from continuing to deleverage and compress valuations.

$AVGO Broadcom's single-quarter AI chip revenue reached $16.7 billion, a year-over-year increase of 221%, with free cash flow reaching $13.7 billion. However, unlike $NVDA, Broadcom does not rely on selling standardized GPUs to all customers for profit. Its business model may be even more complex than many imagine. Broadcom mainly has two major businesses: semiconductor solutions and infrastructure software. According to the Q3 fiscal 2026 earnings report, Broadcom's revenue was $29.591 billion, up 86% year-over-year. Semiconductor business revenue was $20.839 billion, accounting for about 70% of total revenue; infrastructure software revenue was $8.752 billion, about 30%. The semiconductor business includes not only AI chips but also network switching chips, broadband, wireless communication, and other semiconductor products. Infrastructure software centers on enterprise software businesses like VMware, targeting large enterprises and data center customers. Broadcom's most important growth point in the AI era is custom ASICs. NVIDIA mainly provides a general-purpose GPU computing platform, while large cloud providers like Google and Amazon have their own AI chip development plans. For enterprises with highly fixed training and inference tasks, specially designed chips may have advantages in energy efficiency, cost, and specific computing performance. But designing chips in-house does not mean having to independently complete all chip R&D, interconnect design, and production supply chain management. Broadcom's key value lies here. It can collaborate with large customers to develop custom AI accelerators and provide high-speed interconnects, network chips, and related technologies. For example, Broadcom has already reached a long-term cooperation agreement with Google to participate in the development and supply of multiple future generations of TPU chips and provide related network components, with the agreement extending through 2031. The advantage of this business model is that once customers complete chip architecture design, software adaptation, and data center deployment, subsequent generations of products are often continuously advanced. However, concentration of major customers also brings obvious risks; losing important projects could significantly impact revenue. Besides ASICs, Broadcom also operates a globally important data center network chip business. AI clusters are growing larger, with tens of thousands or more accelerators requiring high-speed communication. Broadcom's Tomahawk, Jericho switching chips, and high-speed Ethernet technology mainly solve data transmission issues between different servers and compute nodes. Therefore, Broadcom and NVIDIA are not entirely zero-sum competitors. Even if a cloud provider continues to purchase NVIDIA GPUs, its data center network may still use Broadcom products. Beyond AI chips, VMware is also an important profit source for Broadcom. This business mainly involves virtualization, private cloud, and enterprise infrastructure software, with revenue characterized by strong software subscription and long-term customer relationships. Compared to businesses highly dependent on chip shipment cycles, enterprise software can provide an alternative cash flow source. Broadcom's operating cash flow in Q3 was about $14.2 billion, free cash flow about $13.665 billion, accounting for 46% of revenue. GAAP operating profit for the same period was about $15.955 billion. The company expects total revenue in Q4 to reach $34.8 billion, with AI semiconductor revenue around $21.7 billion. If realized, the AI business will further increase its share in the company's revenue structure. But risks cannot be ignored. As of the end of Q3, Broadcom still had about $59.4 billion in short- and long-term debt, while holding about $24 billion in cash. Although current cash generation is strong, future needs must balance debt, shareholder returns, and business investment. More importantly, demand for custom AI chips heavily depends on capital expenditures of large tech companies. If cloud providers slow down data center construction or AI commercialization returns fall short of expectations, Broadcom's order growth may also be affected. Another issue is valuation. Even if company profits continue to grow, if the market has already priced in ASIC revenue for the next few years, high growth alone may not be enough to drive the stock price higher. In the long term, Broadcom simultaneously holds three important business lines: custom AI chips, network interconnects, and enterprise software, making its business model more diversified than pure GPU competitors. $NVDA has a strong general computing platform and CUDA ecosystem, while $AVGO is better at serving large customers' custom computing needs. Both may benefit from AI infrastructure expansion, but their profit sources and competition methods differ. The key points to watch for Broadcom going forward are AI revenue growth rate, number of custom chip customers, network product orders, VMware business profit margins, and whether free cash flow can continue to grow alongside revenue.

$WDC's revenue grew 55% year-over-year, with the price per TB rising 28%, and adjusted gross margin reaching 58.1%. Amid the global semiconductor downturn, this earnings report at least proves one thing: data center storage demand remains strong, and some storage products even have higher pricing power than before. However, analyzing the impact of this earnings report on other storage companies requires distinguishing between HDD, NAND, and DRAM. In February 2025, Western Digital spun off its flash memory business into the independently listed $SNDK. Today, $WDC mainly operates in mechanical hard drives (HDD), while $SNDK primarily handles NAND flash memory. Although both companies belong to the storage industry, their products, customer applications, and profit drivers are not entirely the same. The company that directly benefits from the validation of this earnings report is $STX, Seagate. It competes with $WDC in the enterprise large-capacity HDD market. $WDC's simultaneous growth in shipment volume, pricing, and profit margin reflects a still tight supply-demand environment for cloud large-capacity hard drives. As long as the industry maintains supply discipline, $STX also has the opportunity to sustain strong profitability. For $SNDK, this earnings report provides an indirect positive signal. Data growth driven by AI does not only require HDD. High-performance storage, enterprise SSDs, data caching, and higher-frequency data access all require NAND flash memory. HDDs mainly leverage large capacity and low unit storage cost advantages, while SSDs rely primarily on performance and access speed. There is a division of labor within data centers between the two, as well as some substitution. Therefore, $WDC's strong growth indicates customers are still expanding storage scale, but it cannot be directly inferred that NAND prices will also rise by 28%. The situations for $MU and SK Hynix are even more different. They benefit from DRAM, HBM, and AI server memory demand, with SK Hynix being a key HBM supplier. $WDC itself does not produce HBM, so this earnings report cannot directly verify future HBM prices or order growth. However, the earnings call revealed a noteworthy detail: $WDC management stated that rising prices of DRAM and NAND components have begun to push up hard drive manufacturing costs. This means that at least from the procurement side of this major hard drive manufacturer, supply of some memory and flash products remains tight. This is an industry signal worth noting for $MU, $SNDK, and Korean storage companies. Longer-term growth comes from AI application expansion. The market has mainly focused on GPUs and model training in the past, but future inference, intelligent agents, video generation, autonomous driving, and robotics will all generate massive amounts of data that need to be stored, read, and processed. As computing power scales continue to expand, the underlying storage infrastructure must grow in tandem. But cyclical risks cannot be ignored. The storage industry has repeatedly experienced supply-demand tightness, price hikes, capacity expansions, and price declines. What needs to be observed now is whether cloud providers' capital expenditures can continue and how long the currently high storage profit margins can be maintained. Therefore, the long-term outlook for the storage industry remains unchanged, and $WDC's earnings report further supports the logic that demand remains strong. However, industry growth and stock price increases are not the same thing. Currently, $SOXX remains in a short-term downtrend, and $MU, $SNDK, and $STX cannot completely escape the overall sector valuation compression. The better the industry fundamentals, the more worth watching the opportunities after future valuation adjustments are fully realized. It is still reasonable to be optimistic about storage now, but there is no need to rush to go all-in to prove a long-term view.

$WDC delivered an almost comprehensively better-than-expected earnings report, with the stock price initially rising about 4% after hours, then quickly reversing to fall nearly 8%. Against the backdrop of a sell-off in the entire semiconductor and storage sector, this earnings report deserves close examination. In Q1 of fiscal year 2027, $WDC reported revenue of $4.381 billion, up 55% year-over-year and 17% quarter-over-quarter, exceeding market expectations of about $4.14 billion. Adjusted EPS reached $4.64, up 161% year-over-year, also surpassing the market expectation of about $4.08. The improvement in profitability was especially notable. Adjusted gross margin reached 58.1%, compared to 43.9% in the same period last year, an increase of 14.2 percentage points. Adjusted operating margin was 49.1%, with operating income around $2.153 billion. Operating cash flow was $1.604 billion, and free cash flow was $1.377 billion. These figures indicate that $WDC not only sold more products but also achieved higher pricing power. This quarter, hard drive shipment capacity reached 248EB, up 22% year-over-year, with nearline drives accounting for 223EB. Meanwhile, the average price per TB rose 28% year-over-year. The 22% growth in shipment capacity combined with the 28% increase in unit price per capacity jointly drove the 55% revenue growth. Cloud business revenue was about $3.9 billion, accounting for 88% of total revenue. This means $WDC is now highly dependent on large cloud providers and data centers rather than traditional consumers buying external hard drives. AI training data, inference logs, video content, and large-scale data retention are all driving enterprises to continue expanding storage infrastructure. The guidance for the next quarter is also strong. The company expects revenue between $5.0 billion and $5.2 billion, with a midpoint of $5.1 billion, significantly exceeding Wall Street’s previous expectation of about $4.57 billion. Adjusted gross margin is expected to further increase to 60%–61%, with EPS projected at $5.60–6.00. However, the earnings call also revealed risks the market is unwilling to ignore. Management acknowledged that costs for components like DRAM and NAND are rising, and capital expenditures as a percentage of revenue are expected to be slightly above the long-term target of 4%–6% over the next two years. The market is also concerned about competitors like Toshiba increasing capacity and whether the currently very high hard drive prices and profit margins can be sustained. This explains why the stock price did not hold its gains after the earnings release. $WDC’s operating data does not show demand collapsing, but the market is no longer willing to pay the previously high valuations based solely on earnings growth. Currently, the stock price reflects concerns about the sustainability of future profits, not just the recently reported quarterly earnings. This earnings report supports the view that the storage industry remains in a prosperous phase but is insufficient to prove that $WDC or the entire semiconductor sector has bottomed out. Until the short-term downtrend is clearly reversed, better-than-expected earnings may still become a window for capital reduction.

$KLAC KLA's fiscal year 2026 revenue reached $13.579 billion, with a net profit of $4.831 billion and free cash flow of $3.767 billion. Even more impressive, the company expects a non-GAAP gross margin of about 62.5% in the next quarter. Why can a company that does not produce GPUs, manufacture memory chips, or provide EUV lithography equipment have such high profitability? KLA is one of the global leaders in semiconductor inspection and metrology equipment, primarily helping wafer fabs detect manufacturing defects, measure critical parameters, and improve chip production yield through data analysis. Its biggest difference from $ASML, $AMAT, and $LRCX lies in the manufacturing stage its products serve. ASML is responsible for lithography, Applied Materials provides a large amount of materials engineering equipment, Lam Research specializes in etching and deposition, while KLA's core value is to determine whether these manufacturing processes meet requirements and where defects occur. To understand this company, one must understand a crucial metric in semiconductor manufacturing: yield. Suppose a wafer fab invests heavily in manufacturing advanced chips, but only 70% of the products meet specifications. If yield can be improved to 80% by enhancing manufacturing processes and defect detection, the number of sellable chips would increase by about 14.3%, assuming other conditions remain unchanged. For wafer fabs producing high-value AI chips, such improvements can bring substantial economic benefits. However, yield improvement is not achieved simply by purchasing an inspection device. Wafer manufacturing requires continuous defect monitoring, identifying problem sources, adjusting equipment parameters, and maintaining stability over long production cycles. KLA's equipment and software serve this process. The company's products mainly include wafer defect inspection, photomask inspection, thin film and pattern metrology, electron beam inspection, process control software, and advanced packaging inspection equipment. Optical inspection can quickly scan wafer surface defects, electron beam inspection provides higher resolution localized analysis, and metrology equipment confirms whether critical dimensions, film thickness, and other parameters meet design requirements. These devices are not simple cameras but require complex optics, electron beam technology, algorithms, and data analysis. According to the fiscal 2026 annual report, KLA's full-year revenue was $13.579 billion, a year-over-year increase of about 12%. Semiconductor process control business revenue was $12.245 billion, accounting for about 90% of total revenue. Special semiconductor process equipment revenue was about $584 million, and PCB and component inspection business revenue was about $750 million. It is clear that KLA does not rely on broad participation in all equipment markets to make money but is highly focused on process control and yield management. This business model has a clear advantage: the more complex the advanced process, the higher the customers' requirements for inspection and metrology. As chip manufacturing enters 2nm and more advanced nodes, transistor structures, materials, and interconnect layers become increasingly complex. Tiny defects can affect chip performance or even cause entire chips to be scrapped, so investment in process control becomes increasingly important. HBM and advanced packaging also create growth opportunities. HBM requires stacking multiple layers of DRAM chips, and advanced packaging involves precise connections between different chips. As structures become more complex, material defects, connection accuracy, and packaging yield affect final product reliability. KLA can provide inspection and metrology solutions for these manufacturing processes, but HBM shipment growth cannot be directly equated to proportional growth in KLA's equipment revenue. Equipment demand also depends on customers' technology upgrades and capital expenditure plans. Another important advantage comes from customer stickiness. Wafer fabs usually require long-term validation to determine whether a set of inspection equipment meets manufacturing requirements. Once equipment is deeply integrated with production processes, process parameters, and software systems, switching suppliers may incur revalidation costs. Therefore, KLA's competitive advantage lies not only in the equipment itself but also in the long-term accumulated process data, algorithm capabilities, and customer relationships. Profitability also proves the value of this model. In fiscal 2026, KLA's GAAP net profit was $4.831 billion, operating cash flow was $4.143 billion, capital expenditure was about $376 million, and free cash flow was $3.767 billion. The company also returned about $3.35 billion to shareholders through stock buybacks and dividends during the year. Compared to chip manufacturers who need to continuously invest tens of billions of dollars to build wafer fabs, KLA does not bear wafer manufacturing capital expenditures of the same scale, allowing it to convert a significant portion of operating cash flow into free cash flow. However, high profitability for equipment suppliers does not mean there is no cyclical risk. KLA's important customers include global large wafer manufacturers such as TSMC. According to the fiscal 2026 annual report, the Chinese market contributed about 30% of revenue. Customer concentration, export controls, and global wafer fab investment slowdowns may directly affect equipment demand. Additionally, process control equipment faces technological competition. Optical and electron beam inspection require continuous improvements in precision, speed, and data processing capabilities. If new products cannot meet the requirements of more advanced nodes, existing advantages may be challenged. The company expects fiscal 2027 first-quarter revenue of about $4 billion, with a fluctuation of $200 million, and a non-GAAP gross margin of about 62.5%. The earnings report released after market close on October 28 Eastern Time will further verify demand for advanced process and memory manufacturing. In the long term, KLA may be one of the semiconductor industry’s equipment companies with a relatively excellent business model. It does not need to bet on whether Nvidia or AMD ultimately gains more market share. As long as chip manufacturing processes continue to upgrade and yield management becomes more complex, advanced inspection and metrology equipment will have opportunities for increased demand. However, a good company still requires a reasonable valuation. KLA's past technological advantages are already quite evident, and investment returns depend on whether future growth can exceed market expectations. When researching $KLAC, focus should be on process control business revenue, advanced process equipment demand, global wafer fab capital expenditures, gross margin, and free cash flow. Chip manufacturing is becoming increasingly complex, and the importance of inspection is indeed rising, but this does not mean equipment orders can completely detach from the semiconductor investment cycle.