Advanced Micro Devices, Inc. vs SK Hynix Inc.: Strategic Comparison

This in-depth comparison examines Advanced Micro Devices, Inc. and SK Hynix Inc. across revenue, market value, business model, competitive positioning, and long-term growth strategy. Whether you are researching Advanced Micro Devices, Inc. on its own, evaluating SK Hynix Inc., or weighing the two companies side by side, the breakdown below highlights where each company leads and where the gap between Advanced Micro Devices, Inc. and SK Hynix Inc. is widest.

On the headline numbers, Advanced Micro Devices, Inc. reports annual revenue of $34.6B against $48.9B for SK Hynix Inc., while their respective market capitalizations stand at $195.0B and $81.5B. Advanced Micro Devices, Inc. is headquartered in United States and SK Hynix Inc. operates from South Korea, and those different home markets shape how each company competes.

Advanced Micro Devices, Inc.: $1.86. That was AMD's stock price in mid-2015. What happened between those two data points is one of the most dramatic turnarounds in technology history — and it wasn't luck. She bet everything on a single CPU architecture called Zen, outsourced manufacturing to TSMC, and told Wall Street to be patient. AMD doesn't make chips. It designs them — obsessively, expensively, brilliantly — and then hands the blueprints to TSMC in Taiwan, which does the actual manufacturing on the most advanced production lines on Earth. It's also why AMD's fate is partially in someone else's hands, but we'll get to that. The money comes from four places, and the mix has shifted dramatically in just three years. This is the crown jewel now. Pensando data processing units handle networking offload. Three years ago, this segment was half its current size. Semi-custom APUs power every PlayStation 5 and Xbox Series console sold worldwide. The console contracts provide predictable multi-year revenue but carry thinner margins than enterprise products. This is the Xilinx inheritance — FPGAs, Versal adaptive SoCs, Alveo accelerators. These go into telecom base stations, fighter jet avionics, automotive ADAS systems, medical imaging equipment, and industrial automation. The margins are excellent. The downside is cyclicality: telecom spending collapsed in 2023-2024, dragging this segment down before it recovers. The unusual aspect of AMD's economics is the margin trajectory. Gross margins have climbed toward 52-54% as the revenue mix tilts from low-margin console chips toward high-value data center products. The FY2025 results benefited from an AI infrastructure spending boom. Whether that spending level is sustainable is a question AMD can't answer alone. It does not manufacture any of them. The capital that doesn't go into factories goes into design engineering. It's Amazon. Amazon is doing something different. Every chip Amazon designs internally is a chip it doesn't buy from AMD. And Amazon is AMD's single largest customer category. Meta designs custom inference silicon. AMD can't sue them into buying EPYC. It can't lock them in with proprietary software the way NVIDIA does with CUDA. Now, Intel. The oldest rivalry in semiconductors — 55 years of it. Intel still ships more total server CPUs than AMD in absolute volume. It still has deeper enterprise relationships built over decades. EPYC went from near-zero server share in 2017 to an estimated 30-35% of x86 server shipments by 2025. If they do, AMD's share gains plateau. If they don't, AMD pushes toward 40-45% and the x86 server market effectively becomes a duopoly where AMD is the premium choice. My judgment: Intel recovers partially but not fully. AMD keeps gaining, just more slowly. Then there's NVIDIA in AI accelerators. AMD's pitch here is honest but limited: "You need a second supplier, and we're the only credible one." That's not a claim of superiority. It's a claim of necessity. NVIDIA's hardware is better today. NVIDIA's software network is vastly deeper. AMD exists in AI because the market structure demands an alternative, not because AMD has earned dominance through technical superiority. Where AMD wins decisively: platform breadth. That matters for customers managing complex infrastructure who want fewer supplier relationships. The fabless model shapes the financial profile in fundamental ways. Every major AI framework was improved for CUDA first. Every university teaches CUDA. Every enterprise AI team has pipelines built on CUDA libraries. AMD cannot manufacture a single advanced chip without TSMC. Not one. The CoWoS advanced packaging bottleneck in 2023-2024 already demonstrated this — AMD couldn't get enough AI accelerators built fast enough because packaging capacity was constrained. The third issue is regulatory. China represents enormous AI chip demand, and AMD is legally prohibited from serving much of it. That's a permanent addressable-market reduction that no amount of product innovation can fix. Intel can't do GPUs or FPGAs at AMD's level. NVIDIA can't do CPUs. Qualcomm can't do servers. Xilinx couldn't do any of it without AMD's distribution and platform integration. But breadth alone isn't a defense. That's not a marketing trick. Then there's the TSMC relationship. Every dollar of R&D goes into design, architecture, and software rather than keeping a factory running. Intel bears that factory burden. AMD doesn't. AMD now has this validation at every major cloud provider. Nobody currently has all six. The dominant wager is AI infrastructure. The AI play has three layers. AMD's accelerators compete on memory capacity and capacity — the MI300X offers 192GB of HBM3, which matters for large language models that need to fit in GPU memory. Second, software: ROCm needs to reach the point where enterprises can deploy AMD hardware without rewriting their CUDA-based pipelines. The supporting bets are simpler. EPYC keeps gaining server CPU share — AMD went from near-zero in 2017 to an estimated mid-30s percentage of x86 server shipments. Ryzen AI targets the emerging AI PC category where on-device inference creates upgrade demand. The Xilinx portfolio serves long-cycle embedded markets that provide margin stability when consumer segments get choppy. That's the metric that tells you whether the AI bet is working or whether AMD remains primarily a CPU success story with AI aspirations. The CPU side is nearly settled. The irony is, None of that is uncertain enough to lose sleep over. That's the irony Lisa Su has to solve. Santa Clara, 1969. The founding thesis was simple: the semiconductor industry needed a second-source supplier for Intel's chips, and someone technically capable should provide it. For its first two decades, AMD operated largely in Intel's shadow, manufacturing compatible versions of x86 processors under licensing agreements that gave Intel legal cover for market dominance claims while giving AMD revenue. The ATI Technologies acquisition in 2006 brought graphics processing capabilities that would prove essential two decades later when GPUs became the computational substrate for machine learning. At the time, it looked like an expensive bet on gaming. In retrospect, it positioned AMD to compete in AI compute before AI compute was a market category. AMD sold its Austin campus. It laid off thousands of engineers. What remained was a pure design firm with a single viable architectural bet — Zen — that Lisa Su and her engineering team had to execute flawlessly. If AMD's software stack crosses that line — call it the point where a Fortune 500 AI team can deploy Instinct accelerators without hiring dedicated porting engineers — then data center GPU revenue doubles by 2028 and AMD becomes a $50-60 billion revenue company. EPYC owns 30-35% of x86 server shipments and Intel would need three consecutive flawless generations to reverse that — something Intel hasn't managed since Haswell. This is two very different businesses wearing the same label. When those companies increase capital spending, AMD's numbers look spectacular. The company designs CPUs, GPUs, and adaptive computing products for data centers, personal computers, gaming consoles, and embedded systems. The company that should worry Lisa Su most isn't NVIDIA. But Intel has been executing poorly since roughly 2015, and AMD exploited every stumble. The question is whether Intel's new leadership can ship competitive products on a modern process node. That's a viable position — it generates billions in revenue — but it's fragile in a way that the CPU business isn't. No other company ships x86 CPUs, discrete GPUs, AI accelerators, FPGAs, and data processing units from a single vendor. The competitive position is the strongest it's been since the Athlon 64 era. Let me be direct about what keeps AMD's leadership up at night: CUDA. The embedded business recovers as telecom spending normalizes. The near-death years of 2012 through 2016 forced choices that determined the modern company. It spun off its manufacturing operations as GlobalFoundries.

SK Hynix Inc.: SK Hynix swung from a $3.5 billion net loss in FY2023 to $4.66 billion in net income in FY2024. That $8.16 billion turnaround in a single fiscal year is one of the most violent recoveries in semiconductor history, and it happened because one product — High Bandwidth Memory 3E — went from niche AI accelerator component to the most constrained commodity in global technology supply chains. The Icheon, South Korea company controls an estimated 50% of global HBM3E market share. That means when Nvidia needs the memory stacks that make the H100 and H200 AI accelerators function, roughly half those stacks come from SK Hynix. The company's proprietary MR-MUF packaging technology — which reduces thermal resistance by more than 20% compared to Samsung's competing method — secured the primary Nvidia design win and established the supply relationship that drove FY2024's $48.9 billion in total revenue. Founded in 1983 as Hyundai Electronics by Hyundai Group founder Chung Ju-yung, the company went through a near-death experience in the early 2000s as the memory cycle collapsed and then another brush with insolvency during the 2008 financial crisis before SK Group acquired it in 2012. The rescue gave SK Hynix access to the capital required to compete in advanced DRAM fabrication, where new facilities routinely cost $15 billion to $20 billion and the difference between a competitive process node and a lagging one determines market share for five years. The 2021 acquisition of Intel's NAND flash business for $9 billion created Solidigm, an enterprise SSD subsidiary that gave SK Hynix a second revenue leg beyond DRAM. The NAND market is more commoditized and lower-margin than advanced DRAM, but the acquisition instantly made SK Hynix the second-largest NAND vendor globally. The strategic question now is whether the company can maintain its HBM leadership as Samsung and Micron accelerate competing HBM programs — and whether the AI infrastructure buildout sustains the demand that turned FY2024 into an extraordinary year.

Advanced Micro Devices, Inc. and SK Hynix Inc. pursue distinct approaches to generating revenue, and understanding how each company operates is the foundation of any fair comparison between Advanced Micro Devices, Inc. and SK Hynix Inc..

Advanced Micro Devices, Inc. business model: When they pull back, or when they design their own custom chips to reduce dependence on merchant silicon, AMD feels it immediately. TSMC in Taiwan runs the actual production lines on the most advanced nodes in the world — 4nm, 3nm — and AMD pays them to do it. But hyperscalers hate single-vendor dependence because it gives NVIDIA pricing power and supply use that no procurement team can tolerate indefinitely.

SK Hynix Inc. business model: The pricing architecture for SK Hynix's products is bifurcated between highly commoditized, spot-market pricing for legacy consumer memory, and negotiated, contract-based pricing for advanced-node enterprise and AI memory. Conversely, during a downcycle, the fixed depreciation and interest expenses rapidly consume cash reserves, forcing the company to slash capital expenditures and reduce wafer starts to stabilize pricing. The primary financial risk is the immense depreciation burden associated with its new fab construction; as the Yongin and Indiana facilities come online in 2026 and 2027, the company will incur billions of dollars in new depreciation expenses that will require sustained high memory pricing and high use rates to absorb, creating a high break-even point that could result in significant losses if another memory downcycle occurs before the fabs reach full scale. This packaging advantage is critical for AI data centers, where the thermal output of AI server racks is the primary bottleneck preventing the deployment of higher-density computing clusters; by using a liquid molding compound that fills the microscopic gaps between the stacked dies and acts as a highly efficient heat spreader, SK Hynix's MR-MUF process reduces the thermal resistance of the HBM package by over 20% compared to the traditional non-conductive film (NCF) method used by Samsung, creating a compelling economic value proposition that transcends simple per-gigabyte pricing and has secured SK Hynix the primary design win for Nvidia's H200 accelerator. The founding philosophy was simple but audacious: to design and manufacture the most advanced, highest-density memory chips in the world, competing directly with the entrenched Japanese conglomerates like Toshiba, NEC, and Hitachi who were then dominating the global memory market with superior quality and aggressive pricing, and the emerging American startups like Micron who were pioneering new process technologies.

The durability of a company's moat often decides long-term winners. Here is how the competitive advantages of Advanced Micro Devices, Inc. stack up against those of SK Hynix Inc..

Advanced Micro Devices, Inc. competitive advantage: Instinct AI accelerators — the MI300X, MI325X, and the newer MI350 — sell to hyperscalers who need alternatives to NVIDIA's $40,000 GPUs. That's a treadmill, not a moat. The x86 server CPU business generates high margins with multi-year design win cycles — once an AMD EPYC chip is designed into a hyperscaler's server rack, that customer doesn't switch architectures for three to five years. The FY2025 acceleration reflects MI300X AI accelerator shipments at scale. The switching cost isn't technical — it's organizational. Set aside the word moat for a second. The real advantage is architectural. The chiplet approach — assembling large processors from smaller, higher-yielding dies connected by Infinity Fabric — gives AMD a manufacturing economics advantage that Intel has struggled to replicate. It's a genuine engineering innovation that translates directly into cost-per-transistor advantages. What rarely gets discussed is server ecosystem validation. Once EPYC is validated in AWS's infrastructure, the switching cost to move away from it is enormous — not because the hardware is irreplaceable, but because the qualification investment is sunk.

SK Hynix Inc. competitive advantage: Because HBM requires significantly more wafer area per gigabyte than standard planar DRAM, and involves complex advanced packaging processes that yield lower output per wafer, the effective supply of HBM is structurally constrained, allowing SK Hynix to negotiate multi-year, fixed-price allocation agreements with hyperscalers that guarantee gross margins exceeding 50% for the HBM segment, regardless of broader memory market fluctuations. Under CEO Kwak Noh-jeong and backed by the immense resources of the SK Group conglomerate, the business has successfully pivoted its product mix toward High Bandwidth Memory (HBM3E) and advanced-node data center solutions, securing multi-year supply agreements with Nvidia and the world's largest hyperscalers to power the next generation of artificial intelligence accelerators. The company's competitive moat is anchored by its proprietary MR-MUF advanced packaging technology, its aggressive adoption of 1-beta and 1-gamma DRAM nodes, and the immense financial barriers to entry that protect the triopoly from new competition. The competitive dynamic between SK Hynix and Samsung is defined by a bitter, decades-long rivalry for absolute scale and technological supremacy in the South Korean semiconductor ecosystem; Samsung possesses a massive revenue base and vertical integration advantage, producing its own logic chips, displays, and mobile devices, which allows it to consume a significant portion of its own memory production and absorb market downturns better than pure-play memory vendors. SK Hynix's competitive advantage lies in its ability to prove superior thermal performance in HBM packaging, higher bit density in DRAM, and a comprehensive enterprise SSD portfolio via Solidigm, a value proposition that resonates powerfully with Western hyperscalers seeking to maximize the compute density of their AI clusters. The competitive moat is also defended through the sheer scale of the capital investment required to compete; with a single leading-edge fab costing over $15 billion, and the R&D required to master MR-MUF packaging and 321-layer NAND stacking running into the billions annually, the financial barrier to entry ensures that the triopoly will remain intact for the foreseeable future, protecting SK Hynix's long-term pricing power and market share. The second pillar of the competitive advantage is SK Hynix's aggressive adoption of leading-edge DRAM nodes, specifically its 1-beta and 1-gamma technologies, which use advanced multi-patterning and selective EUV integration to achieve the highest bit density per wafer in the industry. The fifth pillar is the immense financial and strategic backing of the SK Group, South Korea's second-largest conglomerate, which provides SK Hynix with access to virtually unlimited capital, deep government backing through the K-Chips Act, and a diversified ecosystem of affiliated companies that supply everything from advanced chemicals to industrial gases, insulating the company from the supply chain vulnerabilities that plague standalone semiconductor manufacturers. SK Hynix is also pioneering the concept of 'customer-defined HBM', where hyperscalers like Google and Amazon can customize the base die and memory architecture to optimize for their proprietary AI silicon, a strategic move that deepens the switching costs and locks SK Hynix into the long-term roadmaps of the world's largest cloud providers.

Future prospects matter as much as current results. The growth strategies below explain how Advanced Micro Devices, Inc. and SK Hynix Inc. each plan to expand from here.

Advanced Micro Devices, Inc. growth strategy: The growth rate here is what makes Wall Street pay attention. Ryzen processors for laptops and desktops, sold to Lenovo, HP, Dell, ASUS, and directly to enthusiasts who build their own PCs. The design-in cycles are long, meaning once a customer builds around your chip, they're locked in for 7-10 years. This fabless model means AMD carries no depreciation on semiconductor fabs, which typically cost $15-20 billion each to build. CEO Lisa Su, who took the role in 2014 when AMD's survival was not guaranteed, has built a product roadmap that covers every major segment of the computing market from gaming consoles to AI training clusters. Honestly, that's a fight AMD understands — build better chips, price them aggressively, win on total cost of ownership. It's building Graviton CPUs that replace EPYC in its own cloud. It's building Trainium accelerators that replace Instinct for its own AI workloads. The pattern is unmistakable: the four companies spending the most on compute infrastructure are all investing billions to reduce their dependence on merchant chip suppliers. It can only make its products so good, so cost-effective, and so easy to deploy that the build-vs-buy math keeps favoring buying. Goodwill impairment risk is now a real financial consideration — if Xilinx-derived products don't meet growth expectations, the accounting adjustment could materially impact reported earnings. Not NVIDIA's hardware — AMD can build competitive silicon. NVIDIA spent over a decade building CUDA into the default programming model for AI, scientific computing, and high-performance workloads. TSMC dependence is the second vulnerability, and it's existential in a way most investors don't fully appreciate. If Taiwan faces a geopolitical crisis, a major earthquake, or simply allocates more capacity to Apple and NVIDIA during a shortage, AMD's product launches slip and revenue evaporates. There is no Plan B. Building an alternative would cost $50+ billion and take a decade. Zen is now in its fifth generation, and each iteration builds on validated customer deployments rather than starting from scratch. AMD can build a 128-core server chip from eight identical compute dies plus I/O dies, achieving yields that would be impossible with a single monolithic slab of silicon. The result is higher returns on invested capital when products are competitive. AMD's growth strategy centers on a single dominant wager surrounded by complementary plays. First, hardware: MI300X shipped in volume through 2024-2025, MI350 is ramping now, and the roadmap extends through MI400. That growth should continue as long as the architecture stays competitive. The single data point that determines everything for AMD is data center GPU revenue growth rate quarter over quarter. Ryzen AI in PCs is a steady grower, not a moonshot.

SK Hynix Inc. growth strategy: This land-and-expand strategy within the data center is critical; as AI models grow from hundreds of billions to trillions of parameters, the memory bandwidth required to prevent the GPU from idling increases exponentially, ensuring that SK Hynix's content-per-server metrics continue to scale regardless of broader macroeconomic headwinds in the consumer electronics sector. The capital allocation strategy under the SK Group umbrella has deliberately shifted away from pursuing maximum market share in low-margin consumer electronics, focusing instead on capturing the highest-value segments of the data center and AI markets. The land-and-expand strategy within the data center is driven by the exponential growth of AI model parameters; as large language models scale from hundreds of billions to trillions of parameters, the memory bandwidth required to prevent the GPU from idling increases proportionally, ensuring that SK Hynix's content-per-server metrics continue to scale even if the total number of servers shipped remains flat. The overall business model is a masterclass in extreme industrial engineering and advanced packaging: acquire the technological capability to print the smallest possible transistor and stack the highest possible number of 3D layers, expand revenue by capturing the most demanding AI and data center workloads, retain the customer through deep architectural integration and multi-year allocation agreements, and defend the margin through relentless yield optimization and government-subsidized capacity expansion. SK Hynix counters this by completely exiting the commodity, low-margin segments and focusing exclusively on the high-performance, advanced-node segments where Chinese manufacturers lack the lithography tools and advanced packaging expertise to compete, effectively ceding the bottom 20% of the market to protect the margins of the top 80%. This consolidation has fundamentally altered the competitive dynamics, replacing the destructive, market-share-at-all-costs price wars of the 1990s and 2000s with a more rational, profit-focused oligopoly where capacity discipline is prioritized over volume growth. The financial trajectory is characterized by a deliberate shift in product mix; the percentage of revenue derived from HBM and data center-centric products has grown from less than 10% in FY2022 to over 30% in FY2024, structurally elevating the company's long-term gross margin profile and reducing its exposure to the volatile consumer electronics cycle. A secondary, acute challenge is the brutal, inherent cyclicality of the global memory semiconductor market, a phenomenon driven by the massive lead times required to build fabrication capacity and the commodity-like nature of standard DRAM and NAND products. The third pillar is the deep, architectural integration with Nvidia and other AI chip designers; SK Hynix's engineering teams work directly with Nvidia's architecture groups years in advance of product launches to co-design the custom PHY interfaces, thermal spreaders, and interposer routing required for HBM integration. SK Hynix's growth strategy is explicitly defined by the 'Advanced Node and AI Content' framework, a systematic initiative to capture specific market segments by deploying targeted technologies that expand the company's share of the AI server bill of materials (BOM) without relying on unit volume growth. The strategy is executed through the aggressive ramp of HBM3E and the development of HBM4, which will increase the memory content per AI accelerator from 80GB in the H100 to over 192GB in next-generation accelerators, ensuring that SK Hynix's revenue grows in direct proportion to the performance capabilities of next-generation AI silicon. This growth strategy is executed through a land-and-expand motion that relies on deep architectural integration with Nvidia, AMD, and custom AI chip designers; rather than competing on price in the commodity market, the engineering team focuses on co-developing the custom PHY interfaces, thermal solutions, and customer-defined base dies required for next-generation HBM stacks, creating a level of technical lock-in that guarantees multi-year supply agreements and premium pricing. The channel partner strategy is also evolving to support this framework; SK Hynix is training its network of global module makers and distribution partners to sell the advanced-node server DRAM and Solidigm enterprise SSDs as comprehensive 'AI Infrastructure' packages, offering customers validated compatibility lists and performance benchmarks that justify the premium pricing of SK Hynix's leading-edge products. The company is also pursuing strategic, tuck-in acquisitions to fill gaps in its advanced packaging and controller capabilities; recent investments in packaging startups and controller design firms are specifically targeted to enhance the HBM production yield and the performance of data center SSDs, providing customers with higher-reliability products without requiring the development of new foundational silicon technologies from scratch. The international growth strategy involves establishing a balanced, geographically diversified manufacturing footprint, using the South Korean K-Chips Act to build leading-edge DRAM capacity in the Yongin cluster, while simultaneously expanding its advanced NAND and HBM packaging facilities in the United States and Asia to maintain proximity to the global supply chain ecosystem and customer base, mitigating the geopolitical risks associated with its Chinese operations. The growth strategy also includes the development of industry-specific memory solutions for automotive, industrial, and edge AI applications, which incorporate specialized software features and ruggedized hardware designs tailored to the specific operational requirements and longevity demands of each vertical, expanding the TAM beyond the traditional data center and mobile markets. The financial target of this growth strategy is to increase the average selling price (ASP) per gigabyte across the entire product portfolio by 20% annually, a figure that will be driven entirely by the advanced-node product mix shift and the successful penetration of the AI server market, without requiring a proportional increase in the sales and marketing headcount. The transition to EUV lithography for 1-gamma and 1-delta DRAM is also a critical component of the growth strategy, allowing SK Hynix to achieve the necessary bit density reductions to maintain its cost leadership and gross margin expansion in the face of intense competitive pressure from Samsung and Micron. The company is aggressively expanding its total addressable market (TAM) by capitalizing on the exponential growth of AI training and inference workloads, which require exponentially more memory bandwidth and capacity than traditional cloud computing tasks. The introduction of HBM4, scheduled for volume production in 2026, is the cornerstone of this strategy; HBM4 will use a custom base die designed in partnership with logic foundries to integrate advanced compute capabilities directly into the memory stack, delivering unprecedented bandwidth and reducing the latency between the GPU and the memory, a critical requirement for training trillion-parameter models. The company's long-term financial model targets $80 billion in annual revenue by fiscal year 2028, a goal that requires maintaining a 15% compound annual growth rate (CAGR) while expanding gross margins to the mid-40% range through the operating leverage of the advanced-node product mix and the full absorption of the K-Chips Act and US CHIPS Act subsidies. However, the structural shift toward AI-driven computing is irreversible, and SK Hynix's technological leadership in HBM packaging and advanced-node DRAM positions it to capture the majority of the memory content growth in the AI server market over the next decade. Chung Ju-yung, recognizing that memory semiconductors were the 'rice' of the digital age, established Hyundai Electronics as a dedicated semiconductor division, tasking a small team of engineers with the seemingly impossible mission of building a world-class DRAM fabrication facility from scratch in Icheon, a rural area southeast of Seoul. The team operated out of a modest facility in Icheon, focusing entirely on building the core architecture of the company's first product: a 64K SRAM and a 256K DRAM chip that would use the most advanced n-channel MOS technology available. To bridge the technological gap, Hyundai Electronics engaged in a controversial and aggressive strategy of reverse-engineering and acquiring foreign technology, including a pivotal and highly disputed licensing agreement with Micron Technology for 64K DRAM design rights, a move that would later trigger a massive intellectual property lawsuit in the 1990s when the US ITC ruled that Hyundai had infringed on Micron's patents. The initial customer base consisted of domestic electronics manufacturers like Samsung and GoldStar (now LG), who were eager to secure a local supply of memory chips to feed their rapidly expanding consumer electronics export businesses, as well as a handful of forward-thinking US computer manufacturers who were looking to diversify their supply chains away from Japan.

A closer look at the financial trajectory of Advanced Micro Devices, Inc. and SK Hynix Inc. rounds out the comparison.

Advanced Micro Devices, Inc.: Today it's worth north of $170 billion. FY2025 revenue landed at $34.6 billion. That's a 5x increase from 2019's $6.7 billion. Data Center alone — EPYC servers and Instinct AI accelerators — pulled in $16.6 billion, making it the company's largest business for the first time. Under CEO Lisa Su, the company executed a turnaround through Zen architecture, chiplet design, and TSMC manufacturing partnerships, growing revenue from $4B to $34.6B between 2014 and 2025. This fabless model is why AMD can spend $6 billion a year on R&D without also burning $15-20 billion on factory upgrades the way Intel does. Data Center: $16.6 billion in FY2025. Client: $7.6 billion. Gaming: roughly $7 billion. Embedded: approximately $3.5 billion. AMD grew from $6.7 billion in revenue in 2020 to $34.6 billion in fiscal year 2025. Data Center revenue reached $16.6 billion in FY2025, nearly half of total company revenue. The Xilinx acquisition in 2022 for $35 billion added field-programmable gate arrays to AMD's product range, and the 2024 ZT Systems acquisition brought server integration capabilities. FY2025 Data Center revenue of $16.6 billion, nearly half of AMD's $34.6 billion total, is the number that explains why the market values the company at approximately $195 billion. Revenue trajectory: $22.7 billion in 2022, $22.7 billion in 2023 (essentially flat during an AI infrastructure investment pause), then $25.8 billion in 2024 and $34.6 billion in FY2025. Net income reached $4.3 billion in FY2025 against a market cap of approximately $195 billion — a valuation that prices in substantial future growth from AI infrastructure. AMD has no capital expenditure for manufacturing facilities, so free cash flow conversion from operating income is high. The Xilinx acquisition for $35 billion in 2022 added the Adaptive and Embedded segment, which contributed revenue but also created $26 billion in goodwill on the balance sheet. AMD gets access to the world's best manufacturing without spending $20 billion a year maintaining fabs. The Silo AI acquisition ($665 million) and investments in PyTorch compatibility, vLLM inference improvement, and Hugging Face integrations are all aimed at this. Third, systems: the ZT Systems acquisition ($4.9 billion) gives AMD rack-level design expertise so it can sell complete AI clusters, not just individual chips. The entire valuation debate — whether AMD is worth $170 billion or $300 billion — reduces to a software question masquerading as a hardware company. The relationship was adversarial from the start — AMD filed antitrust complaints against Intel in 2005, alleging that Intel paid PC manufacturers to exclude AMD chips, a case that settled for $1.25 billion in 2009.

SK Hynix Inc.: Revenue of $48.91 billion in FY2024 compared to $15.09 billion in FY2023 — a 224% increase in a single year — is the most dramatic illustration available of how violently memory semiconductor financials can move when the product cycle and the demand cycle align. The $36.63 billion revenue figure in FY2022, the collapse to $15.09 billion in FY2023, and the recovery to $48.91 billion in FY2024 represent three consecutive years of extraordinary volatility in both directions. The driver of the FY2024 recovery was unambiguous: High Bandwidth Memory pricing and volume, fueled by hyperscaler capital expenditure on AI infrastructure. HBM3E commands prices an order of magnitude above commodity DRAM on a per-bit basis because the packaging complexity — stacking multiple DRAM dies and connecting them with thousands of through-silicon vias — limits production yield in ways that standard DRAM fabrication does not. SK Hynix's proprietary MR-MUF packaging process achieved better thermal performance and yield than competing approaches, securing the primary allocation in Nvidia's most advanced accelerator designs. Net income of $4.66 billion in FY2024 compared to a $3.5 billion net loss in FY2023 produced the $8.16 billion swing that made SK Hynix's annual results one of the most widely discussed financial turnarounds in global semiconductors. Market capitalization stood at approximately $81.5 billion — reflecting both the FY2024 results and the market's assessment of how long the HBM premium pricing cycle will last before Samsung and Micron close the technical gap. The 2021 acquisition of Intel's NAND business for $9 billion represents the largest acquisition in SK Hynix's history and created a revenue stream that, while lower-margin than advanced DRAM, provides some counter-cyclicality to the DRAM-heavy core business. The FY2021 revenue of $36.6 billion and FY2022 revenue of $36.63 billion represented a stable period that the DRAM downcycle then destroyed in FY2023 — a reminder that the path from the current position back to the trough, if the AI buildout slows, is steep.