TipRanks Smart Growth Portfolio #57: Rise of the Racks
Dear Investors,
Welcome to the 57th edition of the Smart Growth Portfolio and Newsletter, where we spotlight a company turning silicon into AI infrastructure. But first, here are some news and updates.
1
1
Under Review
❖ We are placing Aviat Networks (AVNW) on negative watch for a possible sale following a short report by GlassHouse Research (GHR) and our reassessment of recent financial trends.
GHR’s track record is mixed: while some of their previous assessments spotlighted real and pressing accounting issues, there are also several counter-examples of companies thriving or being acquired at premiums after their reports. In addition, as a rule, we do not follow short sellers blindly, since their goal – to capitalize on short positions in the companies they cover – is at odds with our long-only, long-term approach.
However, the April 1 report on Aviat highlights several issues that deserve attention. The key claims center on aggressive revenue recognition supported by elevated unbilled receivables, extended supplier payment terms reflected in high accounts payable, and the disclosed NEC arbitration1 demanding additional component purchases plus escrow and outstanding payables. They also argue that declining remaining performance obligations (RPOs) signal weakening forward demand.
Some of these concerns are supported by Aviat’s filings. RPOs have now declined for six consecutive quarters. Accounts payable remain elevated relative to the company’s cost structure, and the arbitration is real – and explicitly disclosed in Note 12 of the company’s most recent 10-Q filing. Ongoing material weaknesses in revenue-recognition controls remain unremediated.
At the same time, the most recent quarter showed partial stabilization: unbilled receivables declined sequentially, operating cash flow turned positive year-over-year, and cash increased. The balance sheet and related-party disclosures support the bear case, while the income statement and cash flow trends support the bull case. That tension remains unresolved. Importantly, year-over-year revenue turned negative in Q2, reinforcing the cautionary view, although H1 was still up nearly 6% from the year-ago period.
Moreover, Roth Capital came to the defense of AVNW, calling the ~13% GHR-driven sell-off “sorely misplaced.” Roth argued that the short report breaks no new ground and relies on dated analysis, pointing to clear signs of improvement in DSOs (days sales outstanding), working capital, and free cash flow in the December quarter. Roth remains positive on Aviat’s core business and highlighted progressing opportunities in Verizon’s MDU (multi-dwelling unit) market, along with Nokia-related optionality. The firm maintained its Buy rating and $38 price target, implying over 90% upside from current levels.
Roth’s constructive stance provides an important counterpoint, but it does not fully resolve the accounting, working-capital, and forward-demand questions clouding the story. Still, we are not executing a sale immediately due to portfolio process discipline. The Smart Growth Portfolio trades are implemented at Friday’s market open, and markets are closed this week for Good Friday. Executing a “virtual” trade would not be replicable for subscribers. Therefore, AVNW remains Under Review with a negative tilt, and we will reassess ahead of next Friday’s open unless meaningful positive developments or a credible company response materially changes the risk profile.
—
1 – In November 2023, Aviat bought NEC’s Pasolink microwave business. As part of the deal, NEC became Aviat’s largest supplier under a long-term Manufacturing Supply Agreement (MSA). The current dispute centers around NEC’s claims that Aviat owes it money for outstanding accounts payable and additional component purchases that NEC says Aviat committed to under the MSA – a claim Aviat disputes – as well as the release of the escrow from the original acquisition. This is one of the more concrete red flags in GHR’s report – it is independently verifiable from the company’s own 10-Q filing, and it signals potential strain in what was meant to be an ongoing, cooperative supplier relationship post-acquisition.
1
1
Portfolio News
❖ Micron (MU) staged a powerful recovery after a brutal sell-off driven by fears surrounding Google’s TurboQuant memory-compression algorithm and its potential impact on high-flying memory stocks. As we noted in the previous newsletter – with our sentiment echoed by several analysts and strategists – TurboQuant remains in its early research phase and does not yet address the most acute bottlenecks in the AI buildout. Memory shortages are real – driven by the AI supercycle – and are expected to remain a notable constraint in the near- to medium-term, until new storage architectures can be implemented around 2028-2029.
Wall Street has decisively pushed back on TurboQuant-related jitters, arguing that the underlying story remains intact and continues to warrant a Strong Buy rating. Cantor Fitzgerald has reconfirmed its Street-high $700 price target, implying potential upside of over 90% from MU’s current level, and reiterated the stock as its Top Pick based on expectations of a multi-year upcycle driven by AI demand, particularly for high-value memory such as HBM and advanced DRAM.
However, although the TurboQuant turbulence appears to be passing, Micron’s stock remains pulled in different directions on a daily basis – along with other technology names – amid volatility tied to headlines surrounding the Iran war, oil prices, and broader macro developments.
1
❖ Applied Digital (APLD) is scheduled to report fiscal Q3 2026 results on April 8. Ahead of the release, Roth MKM reiterated its Buy rating and $58 price target – implying a potential upside of nearly 160% – naming APLD a top pick. Analysts highlighted the upcoming Q3 FY26 report as a pivotal milestone because it will be the first full period to reflect lease revenue recognition, including the CoreWeave-related agreements, which should give investors a clearer read on how the company’s data-center contracts are converting into revenue.
1
1
1
This Week’s Top Growth Pick: Advanced Micro Devices (AMD)
Advanced Micro Devices – commonly referenced as AMD – operates at the center of modern computing, designing the processors and platforms that power everything from personal devices to large-scale data centers and demanding AI workloads. The company develops high-performance CPUs, GPUs, and adaptive computing solutions that deliver faster processing, smarter systems, and more efficient infrastructure. As demand accelerates across cloud computing, artificial intelligence, gaming, and edge applications, the need for scalable, energy-efficient performance continues to rise. AMD positions itself in that critical layer – where raw compute power meets real-world demand – supplying the engines that drive increasingly complex digital experiences across consumer, enterprise, and hyperscale environments.
Source: AMD Q4 and Full-Year Earnings Presentation, February 2026
1
Silicon Reprinted
Advanced Micro Devices traces its origins back to 1969, but its modern identity was shaped far more recently. After decades as a challenger in PC processors, the company reset its trajectory in the mid-2010s with the introduction of its Zen architecture, which restored competitiveness in CPUs and laid the foundation for sustained share gains against larger rivals.
Over the past five years, that foundation has been scaled into a broader compute platform strategy. AMD expanded well beyond traditional PCs, pushing aggressively into data centers, high-performance computing, and artificial intelligence. The 2022 acquisition of Xilinx marked a pivotal step in that evolution, adding adaptive computing capabilities through field-programmable gate arrays and enabling AMD to address a wider range of workloads across cloud, edge, and embedded systems. The addition of Pensando in the same year further strengthened its position in data center networking and distributed computing.
At the product level, AMD continued to iterate rapidly across both CPUs and GPUs. Successive generations of EPYC server processors gained traction with hyperscale cloud providers and enterprise customers, while its Instinct accelerator lineup positioned the company more directly in the AI training and inference market. These platforms increasingly targeted the same high-growth workloads driving industry investment – large-scale data processing, AI model training, and cloud-native applications.
Strategic partnerships have played a central role in this expansion. AMD deepened relationships with major cloud providers and AI infrastructure players, securing design wins across data center deployments and supercomputing systems. Its collaboration with Meta Platforms is one of the more visible examples, reflecting growing adoption of AMD’s AI and data center hardware within hyperscale environments.
At the same time, the company invested in software and ecosystem development to support its hardware ambitions, recognizing that performance alone is insufficient in AI-centric markets. Initiatives around open software platforms and developer tools aimed to improve compatibility and reduce barriers to adoption across enterprise and cloud customers. These moves transformed AMD from a cyclical PC-focused chipmaker into a more diversified compute company aligned with structurally growing markets – particularly data centers and AI infrastructure – where demand is driven by the rising complexity and scale of modern workloads.
1
All Systems Go
AMD sits at the center of one of the largest structural shifts in computing – the transition toward AI-driven infrastructure. What the company sells is no longer just chips. It increasingly delivers the compute backbone of modern data centers – spanning processors, accelerators, and now full systems designed to run large-scale AI workloads.
At its core, AMD’s business is built on two tightly linked engines: EPYC server CPUs and Instinct GPUs. EPYC has become a critical layer in cloud and enterprise environments, handling orchestration, data movement, and inference tasks that sit alongside accelerators. Instinct GPUs, meanwhile, target both AI training and inference, positioning AMD directly in the fastest-growing segment of the semiconductor market. Together, they form a complementary compute stack – one that aligns with how AI workloads are actually deployed today, where CPUs and GPUs operate as a coordinated system, not as separate components.
That system-level thinking is where AMD is pushing hardest. The company is moving up the stack into rack-scale infrastructure through its Helios platform, integrating compute, networking, and software into a unified architecture. This shift – from selling chips to delivering complete AI systems – mirrors the playbook that made NVIDIA dominant, but with a more open, modular approach built around chiplets, ROCm software, and flexible integration. It allows hyperscalers to optimize performance without being locked into a single proprietary ecosystem.
The opportunity set is massive. Management frames the AI infrastructure market as approaching $1 trillion over the next five years, while broader data center infrastructure spending is expanding at a similarly aggressive pace. Against that backdrop, AMD’s current share remains modest, particularly in GPUs, leaving significant room for expansion without requiring it to displace NVIDIA’s leadership. The market does not need a winner-take-all outcome – it needs a credible second supplier, and AMD is increasingly filling that role.
The Meta Platforms partnership crystallizes this shift. It establishes Meta as AMD’s second major hyperscale AI customer after Microsoft, significantly strengthening AMD’s credibility in large-scale hyperscale deployments and confirming that AMD is one of only two serious AI GPU players in the market. The deal features a multi-gigawatt deployment tied to custom silicon, CPUs, and rack-scale systems on the Helios architecture. This positions AMD not just as a vendor, but as a strategic partner and co-developer of AI infrastructure.
The foundation for this model was laid earlier with the landmark OpenAI agreement – AMD’s first 6-gigawatt, multi-year deal with a leading frontier AI lab. The recent Meta deal, along with growing traction across the broader ecosystem, reinforces a commercial model built on long-term contracts, deep co-design, and improving revenue visibility. This represents a structural upgrade in how AMD makes money – moving from largely transactional chip sales toward embedded, strategic infrastructure relationships.
Beyond hyperscalers, AMD is expanding across the ecosystem. Enterprise platforms with Nutanix, emerging “neocloud” providers like Crusoe, and sovereign AI initiatives in Asia all point to diversified of demand. At the same time, partnerships with Samsung on next-generation memory, Flex on U.S. manufacturing, and Celestica on system design strengthen the supply chain needed to support that scale. Meanwhile, agreements with TCS, Cisco, and BlackBerry broaden AMD’s reach across enterprise, infrastructure, and industrial markets. At the edge, AMD is expanding into edge AI and embedded systems through its adaptive computing portfolio – combining Ryzen Embedded processors, Versal adaptive SoCs, FPGAs, and AI accelerators – to power real-time intelligence in industrial automation, robotics, vehicles, and smart devices.
Risks remain. NVIDIA’s software ecosystem still leads, high-bandwidth memory supply is constrained, and execution on complex system deployments is non-trivial. But the direction is clear. AI is shifting toward integrated, large-scale infrastructure, and AMD is positioning itself not just to participate in this buildout, but to help define it.
1
God Mode Unlocked
AMD’s latest results show a business accelerating into a higher-quality growth phase, where scale, mix, and long-term contracts are starting to reshape the earnings profile.
AMD exited 2025 with record annual and quarterly financials. The company delivered $10.3 billion in revenue in Q4 2025, up 34% year-over-year and ahead of expectations, extending a streak of seven consecutive revenue beats. Fourth-quarter adjusted EPS also came in above consensus for the fifth time in recent quarters – surging by over 40% year-over-year – reinforcing a pattern of consistent execution. Full-year EPS rose 26% year-over-year, arriving at $4.17, also well above expectations. Full-year revenue reached $34.6 billion, up 34%, with growth increasingly concentrated in higher-margin data center workloads.
That mix shift is the key story. The data center segment generated $5.4 billion in Q4 revenue, up 39% year-over-year, driven by strong EPYC CPU adoption and the ramp of Instinct GPUs. Operating income in the segment rose to $1.8 billion, highlighting how AI-driven demand is not just lifting revenue but expanding profitability. At the same time, client and gaming delivered solid growth, though these remain more cyclical and are expected to soften in 2026 as the console cycle matures.
Profitability is scaling alongside revenue, though it is still in transition. Gross margin reached 57% in the fourth quarter, supported by a richer mix of data center and AI products, while operating income nearly doubled year-over-year, rising 94%. That said, part of the margin strength was supported by non-recurring items, including China-related MI308 sales and inventory adjustments, with management guiding margins toward a more normalized ~55% level in the near term. Importantly, operating income grew significantly faster than revenue, signaling early operating leverage as the business shifts toward higher-value workloads.
Cash generation is strong and becoming more strategic. AMD produced approximately $2.1 billion in free cash flow in Q4 and ended the year with about $10.6 billion in cash, giving it the flexibility to invest aggressively in AI while maintaining balance sheet strength. Operating expenses rose 42% year-over-year, driven by R&D, ecosystem investment, and go-to-market expansion, but this is a deliberate trade-off to capture a much larger opportunity.
Guidance reflects both near-term normalization and long-term acceleration. First-quarter revenue is expected at around $9.8 billion, implying roughly 32% year-over-year growth but a modest sequential decline due to seasonality. More importantly, management reiterated expectations for data center revenue to grow above 60% annually over the next three to five years, with AI revenue scaling into the tens of billions by 2027.
Large-scale AI agreements are starting to reshape the financial model. The Meta deal alone implies potential revenue of roughly $90-100 billion over several years, with peak annual contributions estimated at $23-25 billion and potentially adding around $5 to annual EPS at peak. These contracts improve visibility and shift AMD toward longer-duration, infrastructure-like revenue streams, though they come with trade-offs. The warrant structure tied to these deals could weigh on gross margins by 200-400 basis points and introduces dilution risk, while also aligning revenue with execution milestones.
There are execution risks – from supply constraints to software competitiveness, the complexity of multi-gigawatt deployments, and rising customer concentration as hyperscaler deals scale. But the trajectory is clear: AMD is moving toward a model with stronger revenue visibility, an improving mix, and expanding earnings power over time, supported by one of the fastest-growing segments in the global technology market.
Source: AMD Q4 and Full-Year Earnings Presentation, February 2026
1
Power Curve
AMD operates within a narrow group of high-performance semiconductor players shaping the AI infrastructure buildout. Its closest and most direct comparator is NVIDIA, while Intel remains the primary benchmark on the CPU side. Broadcom offers a relevant parallel in custom AI silicon and hyperscaler integration, and Marvell reflects the emerging wave of data center–focused infrastructure providers scaling alongside AI demand. Together, these companies frame AMD’s positioning – bridging traditional compute leadership and the push into full-system AI infrastructure.
All of these names rode the same wave over the past year, as surging AI demand and hyperscaler spending lifted the entire semiconductor sector. However, the gap in returns comes down to where each company started and how much its story improved. Intel rallied the most in the group, driven by a turnaround trade from a severely depressed starting point – a year ago, it was priced for failure. Broadcom’s gains were strong but slightly more measured, led by its position as a premium AI infrastructure play with custom silicon, networking, and scale advantages. Marvell’s advance was helped by data center and custom silicon demand, but it lagged somewhat due to lingering skepticism around its competitive positioning. NVIDIA also lagged with “only” ~60% gain, despite having the strongest fundamentals, as it was already priced for dominance and faced greater headline sensitivity tied to its market leadership.
At the same time, AMD’s roughly 110% rally over the past year reflects a real shift in perception, from a secondary player to a credible AI infrastructure partner – highlighted by its position as the only credible second source for AI GPUs besides NVIDIA and backed by major hyperscaler deals and growing data center traction. That shift helps explain Wall Street’s consensus, which points to additional upside of over 30%, with some analysts seeing gains of more than 60%.
Another factor supporting further share-price opportunity is that despite the strong past rally, AMD remains moderately valued versus its peer group (excluding Intel, which still reports negative net profits and has limited analyst visibility). That said, AMD outperforms peers on trailing EPS growth, while its forward earnings growth is expected to be in line with the group. Additionally, although AMD’s trailing EBITDA growth was slightly below the peer average, it is slated to accelerate sharply in the next year, even surpassing NVIDIA by a small margin.
Against this backdrop, AMD trades at a clear discount to peers on both trailing and forward EV/Sales, while sitting broadly in line on forward earnings multiples – which are expected to trend below the peer average in the coming years as growth accelerates. Moreover, AMD’s forward PEG of ~0.75 suggests that the market is not yet fully accounting for its strengthening growth profile, improving product mix, and rising AI exposure.
On top of the prospective share-price gains, AMD actively returns capital to shareholders through buybacks, which remain a priority along with reinvesting in the business. In May 2025, AMD’s board authorized a new $6 billion repurchase program, extending the total available authorization at that time to roughly $10 billion. In 2025, AMD repurchased $1.3 billion worth of shares, with about $9.4 billion remaining available under the program as of year-end 2025.
Notably, while the Meta deal – along with the similar structure with OpenAI – includes a performance-based warrant for up to 160 million shares (roughly 10% of the current share count), these warrants are highly conditional. Vesting occurs in tranches tied to shipment milestones, with the final tranches also requiring AMD’s stock price to reach $600 – roughly 3x the current level. The full dilution will simply not materialize if the stock does not hit those targets. The warrants run through 2031 and are at Meta’s (and OpenAI’s) option to exercise.
Apart from the fact that full dilution is far from unavoidable – although some dilution risk is real if the performance targets are met – analysts such as BofA estimate that the Meta deal is strongly accretive in net terms. Each gigawatt of compute is estimated to generate $15-20 billion in revenue. The full 6 GW deal is valued at over $100 billion over five years. By granting Meta a stake, AMD financially aligns one of the world’s largest AI spenders with its long-term success. BofA estimates 13-23% EPS accretion by 2030 even after accounting for up to 10% dilution – translating to up to $25 in additional EPS power – which should give AMD ample capacity for increased buybacks to help offset any dilution.
AMD now sits at an inflection point where improving fundamentals, rising AI exposure, and still-reasonable valuation converge – positioning the stock to benefit not just from continued execution, but from further upside as market conviction continues to build.
1
To Sum It All Up
AMD is moving into a new phase of its growth story, where it is no longer just competing on chips, but positioning itself as a full-scale AI infrastructure provider. Its expanding role across CPUs, GPUs, and system-level deployments aligns directly with how modern data centers are being built, while hyperscaler partnerships add both validation and long-term revenue visibility. The shift toward multi-year, large-scale contracts is already improving product mix, margins, and earnings power, yet the valuation still lags behind that transformation. With AI demand continuing to scale globally and AMD increasingly embedded in this buildout, the company is transitioning from an alternative supplier into a core infrastructure partner. If execution continues and adoption broadens, the current setup suggests the stock is still early in the rerating cycle.
1
1
Smart Growth Portfolio
|
1
Disclaimer
The information contained in this article represents the views and opinions of the writer only, and not the views or opinions of TipRanks or its affiliates and should be considered for informational purposes only. TipRanks makes no warranties about the completeness, accuracy, or reliability of such information. Nothing in this article should be taken as a recommendation or solicitation to purchase or sell securities. Nothing in the article constitutes legal, professional, investment and/or financial advice and/or takes into account the specific needs and/or requirements of an individual, nor does any information in the article constitute a comprehensive or complete statement of the matters or subject discussed therein. TipRanks and its affiliates disclaim all liability or responsibility with respect to the content of the article, and any action taken upon the information in the article is at your own and sole risk. The link to this article does not constitute an endorsement or recommendation by TipRanks or its affiliates. Past performance is not indicative of future results, prices, or performance.
