The Missing Link
In this edition of the Smart Investor newsletter, we spotlight a foundational enabler of AI-scale data connectivity. Despite our earlier plans to trim some holdings, we are postponing those moves for now as markets navigate heightened uncertainty driven by the current geopolitical flare-up. Within that context, let’s review the latest Smart Portfolio developments.
1
Portfolio News and Updates
❖ Taiwan Semiconductor Manufacturing, aka TSMC (TSM) posted outstanding fourth-quarter and full-year 2025 results. The world’s leading chip foundry sailed past both top and bottom line expectations, with revenue up 25.5% year-over-year to $33.73 billion and net earnings up nearly 40% to $3.14 per ADR. Annual revenue topped $100 billion for the first time, rising 36% year-over-year, while margins also surpassed analyst expectations. For the full year, gross margin climbed to 59.9% and operating margin jumped to 50.8%, reflecting favorable mix toward advanced processes.
TSMC expects this growth to continue to be supported by surging demand for its advanced chips. The tech giant’s HPC division – responsible for AI and 5G applications – made up over 55% of total company revenue in Q4. Additionally, advanced chips measuring 7-nanometer or smaller contributed 77% of total wafer sales during the quarter, making up three quarters of the total in 2025.
However, it is the company’s outlook that moved the stock and lifted the broad AI trade last week. The chipmaking leader forecasted Q1 sales to rise 38% year-over-year at the midpoint. 2026 revenue growth is expected to be “close to 30%” – which indicates only a small deceleration from 2025 in the worst case, and that’s if the results don’t beat TSMC’s own projections. Investors applauded the guidance that dispelled fears that the AI-sparked expansion could not last.
The company also boosted its long-term guidance. Over the five-year period starting in 2024, TSMC expects U.S. dollar revenue to increase at a CAGR near 25%, versus the previous guidance of 20%. Meanwhile, the business that’s far more significant for investor sentiment than headline numbers – responsible for producing AI accelerators like Nvidia’s and AMD’s GPUs and other advanced silicon – is now slated to surge at a CAGR of “mid-to-high 50%” through 2029, significantly above the previous estimate of a mid-40% CAGR.
The AI chip foundry also increased its 2026 capex projections, which are expected to surge from 2025’s $41 billion to $52-56 billion, with 70-80% directed to advanced process technologies. Contrary to some hyperscalers that saw their stocks dip following the increased capex announcement, TSMC apparently enjoys far stronger investor confidence, as its outlay plans were met with enthusiasm.
Up to now, industry players have reportedly been frustrated by TSMC’s subdued production expansion, given the booming demand for AI chips and the accelerating capacity shortages. The company tends to be conservative and risk-averse, so that this notable spending ramp-up apparently signals greater confidence in strong long-term demand, particularly in megatrends like 5G, AI, and HPC.
After having spent several months in talks with its chipmaking customers and their customers, hyperscale cloud providers and other buyers of AI accelerators, TSMC’s chairman and CEO, C.C. Wei now appears convinced that AI is indeed a powerful megatrend, whether there is a bubble in AI stocks or not.
Meanwhile, news of a historic trade agreement between the U.S. and Taiwan, finalized on January 15, is also providing a long-term safety net for the chip sector through facilitating the establishment and expansion of the full semiconductor supply chain and ecosystem in the U.S. Under the new deal, Taiwanese chip and technology companies will invest at least $250 billion in production capacity in the U.S., with the country’s government providing credit guarantees. The U.S., on its part, will limit “reciprocal” tariffs on Taiwan to 15%, down from 20%, and commit to zero such tariffs on several trade items. According to Commerce Secretary Howard Lutnick, the long-term objective is to bring 40% of Taiwan’s entire supply chain and production into the U.S., enabling self-sufficiency in the capacity of building semiconductors.
TSMC has been expanding globally in recent years, with major projects underway in Japan, Europe, and Arizona, where it is accelerating capacity expansion. The chip producer’s CEO said the company has recently purchased additional land in Arizona to support new facilities, with the new U.S.-Taiwan deal providing a supporting network for these plans.
1
❖ Boston Scientific (BSX) announced its agreement to buy medical device company Penumbra in a cash-and-stock deal with an enterprise value of about $14.5 billion. The acquisition, which is expected to close in 2026, will considerably expand BSX’s cardiovascular device portfolio into fast-growing vascular segments, helping the company continue its impressive revenue growth running at ~20%. The impact to GAAP EPS is expected to be dilutive in the first year and less dilutive or increasingly accretive further on, with management confirming its commitment to long-range double-digit EPS growth targets.
The cash part of the deal (about $11 billion) will be financed through a combination of cash on hand and new debt. The size of the deal – significantly larger than the previous bolt-on acquisitions – coupled with a notable increase in leverage and the communicated near-term EPS dilution, as well as concerns over integration risks, drove down the BSX stock.
However, Wall Street remains supportive, with all analysts that follow Boston Scientific currently rating it a “Buy.” Strategic acquisitions are viewed as critical to sustaining above-peer growth rates. Moreover, Boston Scientific believes the deal will help it enter into new, fast-growing segments in the vascular space, noting that Penumbra’s portfolio includes devices to treat conditions such as pulmonary embolism, stroke, deep vein thrombosis, acute limb ischemia, heart attack and aneurysms. The combined vascular and neurovascular markets are estimated at $6-10 billion with low penetration rates, with BSX’s recent deals expected to be accretive to organic growth once annualized, expanding addressable markets and strengthening the company’s procedure-led growth story.
1
❖ CrowdStrike (CRWD) also continued its string of acquisitions aimed at expanding capabilities and solutions portfolio. Just a week after announcing a buyout of identity management startup SGNL for $740 million, the cybersecurity leader said it’s acquiring Seraphic Security, a browser runtime security company, in a cash-and-stock deal worth about $420 million. The transaction is expected to close in early 2027. These moves also build on 2025 acquisitions, which include AI security firm Pangea for $260 million and data startup Onum for $290 million.
CRWD said it plans to integrate Seraphic’s in-session browser protection capabilities with its Falcon platform’s vast endpoint telemetry and threat intelligence, adding SGNL’s identity protection tools to the mix. This will produce a one-stop-shop enterprise AI protection solution at the browser layer, supporting CrowdStrike’s goal of delivering a flexible, unified next-gen security framework that protects every interaction from the endpoint, through the browser session, and into the cloud. CRWD sees AI as both a risk and an opportunity – posing security challenges while offering tools to counter them.
Analysts and industry participants welcomed CrowdStrike’s recent deals, describing them as logical moves amid widening cybersecurity threats, necessitating stronger and more unified tools to control how people and AI access sensitive systems. Wedbush’s Daniel Ives – one of the top Wall Street tech-stock experts – tagged CRWD as one of his top AI stocks for 2026, highlighting AI as a tailwind for the firm amid cybersecurity’s role in the AI Revolution.
Meanwhile, CrowdStrike’s stock slid along with security software peers last week following media reports that Chinese authorities had told domestic firms to steer clear of around 12 U.S. and Israeli software vendors on national security grounds. This is a part of Beijing’s broader push to replace Western-made technology with domestic alternatives amid escalating U.S.-China tech tensions. Regarding CRWD, this is largely a knee-jerk reaction, as the company does not sell to China, and has no offices, employees, or infrastructure in the country.
1
❖ Another Smart Portfolio’s cybersecurity holding, Check Point (CHKP), also saw its stock hit by sector-wide sentiment pressure. The Israeli-American firm was mentioned in the reported Chinese directive, explicitly included alongside U.S. names like Palo Alto Networks, Fortinet, and Broadcom (AVGO)-owned VMware. In contrast with CrowdStrike, CHKP has offices in Shanghai and Hong Kong, but its China operations primarily support international firms operating there. Check Point has a negligible revenue exposure to China, with the country accounting for about 1% of its total sales.
1
❖ According to a Bloomberg report, Alphabet’s (GOOGL) Google is in early discussions to create a joint venture with a U.S. broadband operator Radiate, which is owned by a private equity and infrastructure firm Stonepeak Partners. The new joint company would combine the fiber network assets and operations of both sides, i.e., Google’s residential fiber-to-the-home internet service GFiber, and Radiate’s Astound Broadband label. Astound is one of the larger regional cable and fiber providers in the U.S. Stonepeak is planning to invest about $1 billion in preferred equity to back the JV.
If closed, the deal would be highly beneficial for all involved parties. Demand for very high-speed, low-latency fiber connectivity is exploding because of massive growth in AI infrastructure, cloud computing expansion, and a surge in remote work, along with online gaming and streaming. At the same time, Google has historically found GFiber difficult to scale nationwide due to the extremely capital-intensive nature of the business. By putting GFiber into this JV, Google can deconsolidate the fiber business from its balance sheet and shift the heavy lifting of financing and scaling the physical networks to partners, while still benefitting from the growth in fiber demand.
1
❖ Vistra Corp (VST) dropped last week along with other independent power producers on news about a Trump administration-backed proposal – supported by several Mid-Atlantic governors – to overhaul how PJM Interconnection, the largest U.S. grid operator, handles surging power demand from AI data centers. Many analysts called the VST sell-off an overreaction, but the prevailing sentiment of “the government may cap profit surge” weighed on the stock.
Massive AI data center growth has caused explosive electricity demand in the PJM region, covering about 67 million people across 13 states. As that demand continues to accelerate amid ongoing AI buildout, recent PJM capacity auctions hit record highs. This has led to much higher capacity payments overall, benefiting existing generators like Vistra, but has drawn increasing criticism as consumers face rising utility bills – and power-hungry data centers are blamed for about 40-45% of the increase.
Reportedly, Trump admin’s plan urges PJM to hold one-time auction where hyperscalers bid on 15-year power purchase contracts to fund new power plants. This would shift the burden – with hyperscalers’ data centers locking in and directly paying for new dedicated supply instead of relying on the general grid. This gives the new plants upfront payments and long-term revenue certainty to get built faster. Additionally, PJM would have to temporarily cap prices paid to the existing power plants in the regular capacity market. By capping those prices for about two years, the proposal aims to stop the rapid bill increases for residential and commercial customers.
For hyperscalers like Microsoft (MSFT), Amazon (AMZN), Google (GOOGL) and others, this is a mixed bag. On the one hand, they will pay capped prices while waiting for the dedicated capacity to come online, which can take several years depending on a type of plant. On the other hand, they bear the upfront long-term costs while committing to pay premium down the line, with their overall long-term costs likely higher. Yet, this price may be worth the security and predictability of the dedicated power supply.
For VST and its independent peers – the owners and operators of existing gas and nuclear plants – this proposal could be negative in the short term. These companies benefit hugely from sky-high capacity prices, with caps potentially lowering the upside from scarcity pricing. However, Vistra remains a top player in PJM for reliable baseload and dispatchable power, with a huge existing gas and nuclear fleet and active expansion plans. As such, it is well-positioned to capture a large chunk of the $15 billion+ new build wave, especially given their recent Meta nuclear wins and gas expansion momentum.
Importantly, it is still a proposal under review. Meanwhile, prominent voices opposing the change are already rising. Wedbush Securities’ global head of technology research, Daniel Ives, noted that “China is spending incrementally more across new and existing power technologies into 2030, putting greater pressure on the U.S. to fuel its lofty AI ambitions.” This echoes Nvidia CEO Jensen Huang’s November warning that China will win the AI race as it has a more favorable regulatory environment and cheaper access to power. At the moment, Trump admin’s different goals – facilitating U.S. technological superiority and lowering prices for households – seem to be in conflict.
1
❖ Morgan Stanley (MS) delivered stellar fourth-quarter and full-year 2025 results, beating estimates across the board. Q4 net revenue arrived at $17.89 billion versus $17.77 expected, while EPS came in at $2.68, surging past the consensus of $2.44. Full-year net revenues rose by more than 14% and reached a record $70.6 billion, while adjusted EPS jumped over 28% to $10.21. The company delivered a strong ROTCE of 21.6%, while its expense efficiency ratio improved to 68% from 71% in 2024.
All MS’s business units performed well in 2025. Institutional Securities segment delivered record full-year net revenues of $33.1 billion, up 18%. The segment’s revenue growth in Q4 was driven by Investment banking, which surged 47% year-over-year, driven by higher fees from strong M&A activity across all served regions. For the full year, Equity division posted the fastest revenue growth of nearly 28% year-over-year. Morgan Stanley’s relatively small and agile Investment Management unit grew 11% year-over-year to a net revenue of $6.5 billion, with its assets under management (AUM) at a record $1.9 trillion.
The company’s Wealth Management – which is slightly smaller than IS, but performs as a key driver of overall results, providing durable, stable, and high-quality growth that anchors performance through market cycles – generated a record $31.8 billion in net revenue, up 12% year-over-year. WM attracted over $350 billion in net new assets in 2025, pushing total Morgan Stanley’s AUM up 14% to a record $9.3 trillion. Massive and accelerating asset inflows drive future revenue compounding, positioning WM as the scale-up engine for long-term growth.
Several analysts lifted their price targets on Morgan Stanley following the earnings release, with BofA Securities saying it is “a hard-to-replicate franchise” thanks to the combination of a best-in-class global capital markets division with the outstanding U.S. wealth management business. BofA added that MS could exceed Wall Street’s earnings estimates for 2026 and 2027, while raising its PT and reiterating its buy recommendation on the stock.
1
❖ Citigroup (C) reported excellent operational results in Q4, with record adjusted quarterly EPS of $1.81 (excluding a one-off $1.1 billion Russia divestiture charge) – up 35% year-over-year and notably above the consensus estimate of $1.65. The adjusted revenues of $21 billion surpassed the consensus of $20.72 billion, rising 8% year-over-year excluding the Russia-related notable item. Net interest income (NII) rose 14% to $15.67 billion, sailing past analyst expectations. Citi’s loan loss provision in the quarter was $2.2 billion, about $330 million below expectations. Adjusted ROTCE came in at 7.7%, with full-year adjusted returns improving to 8.8%.
In 2025, the largest U.S. consumer bank saw all of its five divisions delivering record revenues and positive operating leverage. Services delivered 28%+ ROTCE with 8% revenue growth; Markets achieved record revenues and 11.6% ROTCE; Banking posted 11.3% ROTCE with record M&A performance; Wealth grew revenues 14% with 12%+ ROTCE; and U.S. Personal Banking reached mid-teens returns with ROTCE more than doubling year-over-year to 13.2%. Excluding the effects of Russia divestiture and Q3 goodwill impairment of $726 million, total revenues came in at $86.4 billion, up 7% year-over-year, with adjusted EPS of $7.97 surging 27% year-over-year.
The bank’s CET1 ratio of 13.2% provides 160 basis points above regulatory requirements, supporting continued growth and capital returns. During 2025, Citi returned over $17.5 billion to shareholders – the most since the pandemic – including $13 billion in buybacks.
On the earnings call, management highlighted tangible progress on Citi’s transformation, with over 80% of programs at or near target state, and a key regulatory article terminated in December, signaling improved oversight and risk management. On the technology front, AI and proprietary tools have been deployed widely, with AI adoption exceeding 70% across 84 Citi-served countries.
Looking ahead to 2026, Citi targets NII growth (ex. Markets) of 5-6%, an efficiency ratio around 60%, ROTCE of 10-11%, and continued positive operating leverage. Management expects card net credit losses to remain within 2025 guidance ranges and plans to continue prioritizing share repurchases under the ongoing $20 billion program.
Following the earnings release, multiple Wall Street firms raised their price targets on Citigroup, including Argus, Oppenheimer, Piper Sandler, Morgan Stanley, and others. Wells Fargo maintained its “Buy” rating with a Street-high PT of $150, viewing Citi as a “dominant number one pick” among banks.
1
❖ Interactive Brokers (IBKR) reported fourth-quarter and full-year 2025 financial results that beat Wall Street forecasts on the top and bottom lines.
Q4 revenue rose 15.4% year-over-year to $1.64 billion – slightly ahead of consensus – while adjusted EPS surged nearly 28% to $0.65, notably above estimates. The outperformance was driven by a double-digit surge in customer trading volumes, resulting in a 22% jump in commission revenue to a quarterly record of $582 billion. Net interest income increased 20% to $966 million on higher average customer margin loans and customer credit balances and stronger securities lending activity. Customer accounts increased 32% year-over-year; customer equity rose by 37%; customer credits grew by 34%; customer margin loans jumped by 40%. Daily average revenue trades (DARTs) increased 30% to 4.04 million. IBKR’s pre-tax income exceeded $1 billion for the fifth consecutive quarter, achieving a 79% pre-tax margin, up from 75% a year ago.
In full-year 2025, the brokerage added more than 1 million net new accounts, an annual record. IBKR’s client equity surged 37% to $780 billion, marking the first time the company ended the year with over $750 billion in client assets. Full-year commission revenue was $2.1 billion, up 27% year-over-year, driven by higher trading volumes across major product categories. Total net revenues grew 20% to $6.21 billion, exceeding $6 billion for the first time. Full-year pretax margin reached a record of 77%. Net interest income also reached a record, arriving at $3.6 billion despite lower benchmark interest rates. Adjusted EPS jumped 27% year-over-year to $2.23.
In December 2025, the firm filed an application with the U.S. Office of the Comptroller of the Currency for a national trust bank charter, under the working name Interactive National Trust Bank. The earnings call revealed that IBKR is already in dialogue with the OCC and expects the trust bank to be operational by the end of this year. The move would allow Interactive to internalize banking operations and reduce reliance on partner banks, providing its mutual fund and ETF clients with custodial and securities lending services directly. This should improve margins and operational control over time, and positions IBKR to be a more attractive, federally regulated custody option for institutional clients.
In 2025, Interactive Brokers introduced new products and platform enhancements, including the enhanced GlobalTrader 2.0 mobile platform and AI-powered tools, and expanded market access to Brazil, Taiwan, the UAE, and Slovenia, and launched new tax-advantaged accounts in Sweden, Japan, and Canada. Looking forward, management indicated ongoing global expansion, with additional countries planned for 2026 and further enhancements to platform features and tax-advantaged accounts. The company expects continued account growth in 2026 with no slowdown anticipated.
1
1
Portfolio Stocks Under Review
❖ We are placing Oracle (ORCL) in our “Under Review” bracket, as we plan to watch it closely until its next earnings report in mid-March. While we still hold a strong conviction about the company’s prospects in the medium to long term, the near-term ride may prove wild enough to shake our will to hold on to ORCL stock. On the other hand, the developments could further confirm our positive outlook.
Oracle’s fiscal Q2 2026 was a mixed bag: adjusted EPS surged well ahead of expectations, but a large chunk of that upside was a one-time item. Still, even without that addition, bottom-line growth was solid, as was the top-line expansion, even if the latter slightly missed. Importantly, in FQ2, cloud revenue jumped 33% year-over-year, while cloud-infrastructure sales soared 66%, driven by a 177% surge in GPU-related revenue. Despite that eye-popping result, cloud-infrastructure revenue still fell short of estimates due to supply constraints.
Oracle’s multiyear backlog – remaining performance obligations, or RPO – ballooned to over $523 billion at the end of last fiscal quarter, up 433% year-over-year. These outstanding numbers haven’t done much to improve sentiment at the time, as investors were already questioning Oracle’s ability to translate backlog into near-term revenue. Meanwhile, Oracle’s capex projections were higher than expected, and deeper-than-anticipated free-cash-flow pressures meant that the elevated capex needed to be financed through debt. With liabilities rising nearly as fast as doubts about ORCL’s ability to convert orders on the books into real hard cash, it is no wonder that sentiment sunk. Another issue in question is Oracle’s reliance on OpenAI, which constitutes about 60% of its RPO, coupled with worries over the AI startup’s ability to pay these staggering sums.
However, some of the existing worries surrounding Oracle have begun to lessen. First, the sentiment regarding AI supply-chain components – including ORCL – got a boost from TSMC’s (TSM) blockbuster outlook. Additionally, according to Gartner’s recent report, global AI spending will reach $2.53 trillion in 2026 and climb to $3.33 trillion in 2027, with the bulk of that going to AI infrastructure, including Oracle Cloud Infrastructure (OCI) – a major player in providing the compute, storage, networking, and data center capacity needed to power AI workloads.
The TikTok deal, signed in late 2025, is also a major win for the company – providing long-term revenue visibility through a high-margin, sticky, big-ticket contract, and confirming OCI as a key player in AI/cloud infrastructure. With TikTok’s U.S. operations required to store all U.S. user data on OCI, analysts estimate ORCL could generate $10-20 billion, or even more, in cumulative revenue over the next 5-10 years from this business alone. This deal also validates Oracle as the fourth hyperscaler, as well as a sovereign and secure cloud play.
The question of the backlog’s heavy reliance on OpenAI is also being addressed through large partnerships, particularly with Meta Platforms. The Facebook owner has recently announced a major new strategic initiative, Meta Compute, focused on building and scaling massive infrastructure to create a decisive advantage in the AI race. Meta cannot achieve this on its own – it is aggressively supplementing its capacity with third-party cloud providers to meet immediate training and inference needs while its own gigawatt-scale campuses ramp up. Oracle plays a major role in this push through a large multi-year cloud computing agreement with Meta, signed in late 2025 and potentially worth around $20 billion. That’s on top of the already existing collaboration, where OCI provides high-performance compute capacity for training and running Meta’s models. As a reminder, last year, Oracle struck multi-billion-dollar contracts with other key industry players, which analysts believe are xAI, Nvidia, and AMD.
All in all, 2026 is expected to be the “show me the money” year for Oracle. In its fiscal Q3 2026, we expect to gain a meaningful signal about the company’s ability to convert RPO into cash, as its near-term investment case strongly hinges on execution.
1
1
Portfolio Earnings and Dividend Calendar
❖ The Q4 2025 earnings season is gaining altitude, and several Smart Portfolio holdings are scheduled to report over the next several days. GE Aerospace (GE) and Capital One Financial (COF) will reveal their results on January 22, and RTX (RTX) will reveal its results on January 27.
❖ The ex-dividend date for Bank of New York Mellon (BK) and Pfizer (PFE) is January 23.
w
1
New Buy: Credo Technology (CRDO)
Credo Technology Group Holding Ltd. is a fabless semiconductor company focused on high-speed connectivity solutions that support the rising bandwidth and power-efficiency demands of modern data centers. The company designs advanced physical-layer semiconductor IP and devices that enable data to move reliably and efficiently across servers, accelerators, and networking equipment. Its products are built for environments where signal integrity and energy efficiency are becoming as critical as raw performance. By operating at the interconnect layer rather than the system level, Credo plays an infrastructure-enabling role within hyperscale and AI-centric architectures. As cloud and AI systems grow faster and more complex, the company’s technology helps make next-generation connectivity scalable, reliable, and economically viable at scale.
1
Scaling the Signal
Credo Technology was founded in 2008 with a focused mandate: to push the physical limits of high-speed data transmission inside networking hardware. For much of its early life, the company operated behind the scenes, supplying specialized serializer-deserializer (SerDes) technology as cloud infrastructure scaled steadily but predictably. That context shifted materially in the late 2010s, when bandwidth growth began colliding with power, thermal, and signal-integrity constraints.
The company’s modern trajectory began to take shape around 2019-2020, as hyperscale data centers moved into a new design phase. Bandwidth alone was no longer the binding constraint – power efficiency became equally decisive. Higher Ethernet speeds, denser server racks, and early AI workloads exposed the limitations of traditional passive interconnects. Credo responded by expanding beyond discrete SerDes into power-efficient active electrical cables and optical digital signal processing, moving closer to architectural decision points within the data center.
From 2020 through 2022, Credo broadened its physical-layer portfolio to support next-generation Ethernet standards and the early scaling of AI-centric infrastructure. Its technology moved upstream in customer design cycles, shaping how data flows across servers, accelerators, and switches rather than entering as a late-stage component. This architectural role directly influenced the company’s roadmap, tightening alignment with emerging AI workloads and higher-speed interconnect requirements. CRDO’s position in the value chain is further reinforced by deep, iterative engagement with leading hyperscalers and tier-one networking vendors.
The company’s public listing in early 2022 provided the capital base to support its focused expansion in high-speed, power-efficient connectivity and deeper platform-level integration. Credo used the transition to accelerate investment in advanced process nodes, expand engineering depth, and run multiple product generations in parallel. Management maintained a narrow strategic focus, resisting broad diversification in favor of reinforcing its position in high-speed connectivity.
Notably, CRDO’s growth has been almost entirely organic – an outlier in a semiconductor landscape where many peers rely on acquisitions to accelerate relevance. The only acquisition it made in recent years was Hyperlume in 2024, a small, targeted technology purchase focused on capability expansion rather than revenue scale. The deal added microLED-based optical interconnect technology, now underpinning Credo’s Active LED Cable (ALC) roadmap.
Instead of pursuing large buyouts, the company expanded from a roughly $1.7 billion market capitalization at IPO to over $27 billion now, driven by internal R&D, process-node advances, and deeper design wins with hyperscale and networking customers. In effect, Credo built its position by solving a specific physical-layer problem earlier and more effectively, then scaling alongside the rise of AI-driven data-center architectures. IPO capital flowed into engineering depth and product breadth, not balance-sheet expansion.
Over the past several years, this disciplined execution has reshaped Credo from a niche IP supplier into a connectivity specialist aligned with the structural demands of AI-scale data centers. Today’s company reflects that evolution – grounded in early technical depth, refined through long-cycle platform integration, and positioned where data-movement efficiency increasingly defines system performance.
1
Bandwidth Is Power
Credo Technology is a high-speed connectivity specialist built around a simple but increasingly critical premise: as AI data centers scale, moving data reliably and efficiently becomes just as important as compute itself. The company designs the physical-layer technologies that sit inside cables, transceivers, and chips, enabling GPUs, CPUs, switches, and memory to communicate at extreme speeds without breaking down under power, thermal, and signal-integrity constraints.
At the core of Credo’s business are Active Electrical Cables (AECs), its flagship product and primary revenue driver. AECs look like copper cables, but embed active electronics – including proprietary SerDes and signal conditioning – inside the connectors. This allows copper links to operate at far higher speeds and distances than passive cables while consuming less power and delivering materially higher reliability than optics at short reach. In large AI clusters, where a single link failure can idle thousands of GPUs, that reliability advantage is decisive. As a result, AECs are widely viewed as the de facto standard for inter-rack connectivity up to roughly seven meters, with industry estimates placing Credo’s AEC market share at approximately 90%. AECs account for the majority of the company’s revenue today and are the foundation of its hyperscaler adoption.
Surrounding AECs is a growing IC and IP business, which includes retimers, optical digital signal processors (DSPs), and SerDes intellectual property. In practice, Credo monetizes this layer in two parallel ways: by selling finished integrated circuits used in optical modules, switches, and line cards, and by licensing the underlying SerDes IP and chiplets to customers designing their own custom silicon. These products support higher Ethernet speeds and longer reach as data centers transition from 100G to 200G per lane and toward 800G and 1.6T ports. While Credo does not disclose segment revenue percentages, management consistently frames the IC business as the second growth engine, with optical DSPs expected to scale meaningfully as AI clusters expand beyond rack-level designs.
In the past year, Credo has broadened its scope beyond cables and components into system-level connectivity. ZeroFlap optical transceivers extend its reliability-first approach into optics by actively monitoring and stabilizing optical links, addressing one of the main failure modes in large clusters. Through the acquisition of Hyperlume, CRDO entered Active LED Cables (ALCs), which use microLED-based optical signaling to bring copper-like efficiency and determinism to longer distances. Management views ALCs as a next-phase adjacency as AI clusters grow from rack-scale to row-scale and cluster-scale architectures, where AECs alone cannot physically reach.
Most recently, Credo expanded into system-level memory connectivity with OmniConnect, a platform built around its Weaver gearbox chip, which allows AI accelerators to connect to much larger pools of external memory at high bandwidth, reducing inference bottlenecks and limiting dependence on costly on-package HBM. While still early, this move extends Credo’s role beyond networking and into how AI systems are architected, as memory access increasingly constrains performance in large-scale inference deployments.
At the silicon ecosystem level, CRDO’s participation in the Arm Total Design ecosystem – led by Arm Holdings – integrates its SerDes IP and chiplets into Arm-based custom silicon platforms, embedding Credo’s connectivity technology deeper into customer ASIC and chiplet design flows. Rather than selling only finished components, Credo increasingly participates earlier in the design process, tightening alignment with hyperscaler roadmaps and raising long-term switching costs.
Together, these offerings form five connectivity pillars – AECs, IC solutions, ZeroFlap optics, ALCs, and OmniConnect – expanding Credo’s total addressable market to more than $10 billion, more than triple where it stood roughly 18 months ago. Adoption is driven almost entirely by hyperscale customers. While Credo does not formally name them, industry consensus points to Amazon, Microsoft, Meta, Google, and a newly ramping fifth hyperscaler widely believed to be Oracle, alongside OEM partners that integrate Credo technology into servers and switches.
While customer concentration remains high, it is improving, with the company no longer relying on a single dominant customer as it did earlier in its growth cycle. Competitive risks are real: large incumbents such as Marvell and Broadcom are investing in adjacent connectivity solutions, and hyperscalers are incentivized to develop secondary sources over time. However, Credo currently benefits from technical lead, reliability differentiation, and deep system integration that raise switching costs.
Industry observers increasingly frame Credo as a “picks-and-shovels” provider for AI, supplying the connectivity layer that enables GPU and accelerator scale regardless of which compute architectures or vendors dominate at the chip level. As AI systems grow denser, more distributed, and more uptime-sensitive, Credo’s ability to deliver deterministic performance, power efficiency, and link stability across multiple layers of the data-center stack is what differentiates it – and why its connectivity solutions are increasingly treated as infrastructure rather than interchangeable hardware.
1
Insulated by Design
Credo’s exposure to the current geopolitical flare-up remains limited, both by revenue mix and by how its business is structured. While the company operates globally, its demand is anchored overwhelmingly in U.S. hyperscalers and Asia-based manufacturing ecosystems, not Europe. Recent filings and disclosures consistently show North America and Asia as the only material geographies, with Europe embedded within a broad “rest of world” category that accounts for roughly a tenth of revenue at most – and Europe itself is only a small fraction of that.
That matters in the context of rising U.S.-Europe friction. The newly announced U.S. tariffs on imports from select European countries are inbound measures, meaning they primarily affect European-sourced goods entering the United States. Credo’s core supply chain is not European. Its chips are fabricated and assembled predominantly through Asian partners, with Taiwan-based foundries and Asia-centric manufacturing hubs forming the backbone of production. There is little evidence of reliance on European suppliers for critical components, limiting direct cost exposure.
The more relevant second-order risk would be European retaliation against U.S. technology exports. Here again, Credo’s positioning provides insulation. Europe is not a primary end market, and no single European country represents a disclosed or material revenue contributor. Any countermeasures would therefore have a marginal impact relative to Credo’s revenue base, which continues to scale alongside U.S. and Asia-driven AI infrastructure investment.
Importantly, Credo’s growth engine is not cyclical regional demand but hyperscale architecture decisions made years in advance. Its connectivity solutions are designed into platforms rather than sourced opportunistically, reducing sensitivity to near-term trade noise. While broader market volatility tied to geopolitics can affect sentiment, Credo’s fundamentals remain tethered to AI cluster scale-out, not transatlantic trade dynamics. In a market increasingly sensitive to political shocks, CRDO’s geographic and supply-chain profile functions less as a risk amplifier and more as a stabilizer.
1
Unlocking the Bottleneck
Credo’s recent financial performance reflects a company that has moved decisively from early-stage scaling into a phase of profitable hypergrowth, with execution that has consistently outpaced expectations. Over the past two years, results have not merely improved – they have accelerated, quarter after quarter, as hyperscale AI deployments shifted from pilot programs to large-scale production.
That execution shows up clearly in Credo’s track record against the Street. CRDO has beaten analyst consensus on revenue for at least eight consecutive quarters, and on adjusted EPS in every quarter since fiscal Q3 2023, except for two instances of in-line results. This pattern matters because it signals not just demand strength, but improving forecasting accuracy and operating discipline in a business often assumed to be volatile.
Fiscal Q2 2026 was Credo’s strongest quarter to date and reinforced that trend. Revenue reached $268 million, up 20% sequentially and 272% year-over-year, materially above both company guidance and analyst forecasts, which had clustered closer to the mid-$230 million range. Growth was driven primarily by Active Electrical Cables, with continued momentum in IC products such as optical DSPs and retimers. While the company does not disclose segment-level revenue, management consistently emphasized that AECs remain the dominant contributor, with ICs providing incremental upside as architectures move to higher speeds and longer reach.
Profitability scaled alongside revenue. Non-GAAP gross margin reached 67.7%, exceeding guidance and underscoring CRDO’s pricing power in reliability-critical connectivity. Operating leverage was pronounced, with non-GAAP net income of $128 million, rising materially faster than revenue. Adjusted EPS came in well ahead of consensus – $0.67 versus $0.49 expected – continuing a multi-quarter pattern of upside delivery. The company has demonstrated that it can grow aggressively while sustaining elite margins and a clean balance sheet.
Cash generation further validates the quality of those earnings. Credo generated $61.7 million in operating cash flow and $38.5 million in free cash flow in Q2 alone, building on multiple prior quarters of positive cash generation. This is not a one-off inflection, but a scaling trend as margins widen and revenue grows. The company exited the quarter with a strong net-cash position and no debt, an uncommon profile for a semiconductor company expanding at this pace. That balance sheet strength gives CRDO flexibility to fund elevated R&D investment, support inventory ramps tied to hyperscaler deployments, absorb customer timing variability, and scale production without external financing pressure.
On a GAAP basis, profitability has also become increasingly consistent, though management continues to emphasize non-GAAP results as the clearest indicator of underlying operating performance in a stock-based-compensation-heavy growth phase. Importantly, Credo reached non-GAAP profitability before its IPO, underscoring that this margin structure was achieved through product economics and scale – not public-market leverage.
Guidance points to further acceleration. For fiscal Q3 2026, Credo projected revenue of $335–345 million, implying roughly 27% sequential growth at the midpoint. Non-GAAP gross margin is guided to 64–66%, reflecting some mix normalization but still exceptional by industry standards. Operating expenses are expected to rise as the company invests in OmniConnect, ALCs, and advanced SerDes platforms. For the full fiscal year, management reiterated expectations for more than 170% year-over-year revenue growth and non-GAAP net margins around 45%, implying that earnings and cash flow should continue compounding faster than revenue.
At this stage, Credo’s financials describe a business that has moved well beyond proof-of-concept profitability into a phase defined by durable margins, repeatable cash generation, and zero leverage, with hardware economics increasingly resembling software-like operating leverage. That combination positions CRDO to scale through the next leg of AI infrastructure buildouts from a position of financial strength.
1
Follow the Flow
Credo belongs to a narrow subset of U.S.-listed semiconductor companies whose value is driven by high-speed connectivity inside hyperscale AI data centers. Its most relevant public peers, therefore, skew toward infrastructure-level bottleneck solvers, not diversified chip vendors. Astera Labs stands out as the closest pure-play comparison, given its similar growth stage and focus on removing data-movement constraints inside AI clusters. MACOM Technology provides a useful benchmark for physical-layer connectivity at scale, pairing optical and analog expertise with established profitability. Marvell serves as the scaled incumbent reference, competing across optical DSPs and SerDes at a much broader scope. Rambus complements the group as an IP-centric comparator, reflecting the licensing and chiplet optionality embedded in CRDO’s model.
Stocks in this cohort delivered strong performance over the past 12 months, with one clear exception, Marvell, and the dispersion largely reflects where each company sits in the AI build-out cycle and how much it surprised on earnings. Credo, a critical connectivity bottleneck owner with revenue scaling far faster than the market initially priced in and margins expanding despite rapid growth, is up roughly 85% over the past year. The stock re-rated not only on earnings acceleration, but on growing indispensability. Rambus followed with gains of around 70%, benefiting from AI-driven memory demand, DDR5 adoption, and its high-margin IP licensing model – though with steadier, less explosive growth. MACOM Technology advanced approximately 48%, as rising data-center exposure more than offset lingering softness in telecom and industrial end markets. Astera Labs gained roughly 46%, delivering solid operational execution but lagging CRDO as its premium valuation left less room for upside surprises. Marvell, by contrast, declined about 35%, weighed down by exposure to weaker non-AI end markets and China, alongside investor concerns about competitive pressure in custom AI silicon and reported customer-share losses.
Despite the outperformance over the past year, analysts still see potential upside of nearly 50% from CRDO’s current levels, with consensus remaining a “Strong Buy.” One reason is that valuation metrics have not yet fully adjusted to the company’s growth profile, which – supported by strong balance-sheet health and a critical-juncture position in AI infrastructure – rests on more durable fundamentals than most peers. Headline multiples screen as demanding, but look far more reasonable when viewed through a growth-adjusted lens. On forward non-GAAP earnings, CRDO trades at ~40x FY2 and ~33x FY3, below Astera Labs and broadly in line with MACOM despite materially faster growth. Its forward non-GAAP PEG of ~0.79x remains low both in absolute terms and versus peers, signaling that a meaningful portion of its growth trajectory is still unpriced.
That growth differential is decisive. Credo has grown – and is expected to continue growing – by an order of magnitude faster than all peers except Astera Labs, whose growth rates are still just over half of CRDO’s pace. Viewed in this context, Credo’s higher Price/Sales and EV/EBITDA multiples reflect not excess optimism, but a business compounding fundamentals faster than the market can easily reprice.
Credo is effectively growing into its valuation rather than outrunning it. While a near-doubling of the stock would typically raise concerns about overheating, CRDO’s revenue, earnings, and cash flow have compounded even faster, leaving the business more attractive on a growth-adjusted basis today than it was a year ago. As AI infrastructure spending broadens and connectivity becomes an increasingly binding constraint, Credo’s role as a bottleneck owner positions it to keep compounding fundamentals ahead of sentiment. If execution holds, valuation compression is far more likely to come from earnings catching up than from the stock coming down.
1
Investing Takeaway
Credo is a growth story built for an era where data movement, not compute, defines system performance. As AI clusters scale, reliability, power efficiency, and signal integrity have become binding constraints – and that’s exactly where Credo operates. Its technology sits at the physical layer of AI infrastructure, quietly enabling GPUs, switches, and memory to function at full utilization and at scale without interruption. What differentiates Credo is not just speed, but reliability: predictable performance in environments where failure is costly. Adoption has followed that logic, with hyperscalers embedding its solutions earlier and deeper into next-generation architectures. As connectivity becomes the gating factor for AI scale, Credo is increasingly positioned not to follow the cycle, but to shape its limits.
1
1
Smart Investor’s Winners Club
The 30% Winners Club represents stocks from the Smart Investor Portfolio that have risen at least 30% since their purchase dates.
Stocks were hit by risk-off sentiment, but our Winners list remained unchanged, still holding 18 stocks:
GE, AVGO, TSM, APH, ANET, HWM, EME, IBKR, PH, ORCL, GOOGL, VRT, MTZ, RTX, BK, MS, IBM, and CRWD.
The first contender for the Club’s entry is still CSCO with a 25.34% gain since purchase. Will it join the ranks, or will another stock outrun it to the finish line?
1
1
|
