Advanced Micro Devices is a leading designer of high performance computing and graphics technologies that power PCs, game consoles, data centers, and embedded systems. Over the past decade the company has transformed its portfolio and brand, moving from a challenger to a category shaper across CPUs, GPUs, and adaptive computing. That evolution makes AMD central to trends in cloud infrastructure, AI, and edge computing.
Conducting a SWOT analysis clarifies how AMD’s internal capabilities align with external market forces in a volatile semiconductor cycle. It helps business leaders, partners, and investors understand sources of durable advantage, vulnerabilities to monitor, and where the next wave of growth could emerge. The following assessment synthesizes recent developments to frame strategic priorities.
Company Overview
Founded in 1969, AMD has evolved from a supplier of PC chips into a fabless leader in high performance computing. The company architects x86 CPUs, GPUs, and adaptive SoCs, while manufacturing is outsourced primarily to TSMC on advanced process technology. A focus on energy efficiency and scalable design underpins its competitive positioning.
Client and gaming products include Ryzen processors and Radeon graphics for desktops, laptops, and consoles. The data center segment features EPYC server CPUs and Instinct accelerators that target cloud, enterprise, and HPC workloads. The Xilinx acquisition added Versal adaptive SoCs, FPGAs, and domain specific platforms that serve communications, automotive, industrial, and aerospace markets.
AMD competes most directly with Intel in CPUs and with NVIDIA in GPUs and AI accelerators. It has been gaining share in servers and showcasing rapid traction in AI systems as customers diversify suppliers and seek performance per watt leadership. The company pursues a multi generation roadmap cadence that emphasizes chiplets, high bandwidth memory, and heterogeneous computing.
Strengths
AMD’s strengths reflect bold architectural choices, disciplined execution, and a broader mix that spans client to cloud. These advantages enable the company to serve compute intensive workloads while balancing innovation with cost and power efficiency. The following strengths capture differentiators that support sustained momentum.
Chiplet Architecture and Advanced Manufacturing Leverage
AMD pioneered modern chiplet based CPUs and GPUs, separating compute cores, I/O, and cache to scale performance efficiently. Partnering closely with TSMC, it utilizes leading process nodes and advanced packaging to deliver higher density, yield benefits, and rapid iteration. This approach underlies Zen CPU generations and RDNA graphics roadmaps.
Chiplets improve performance per watt and bill of materials while enabling flexible SKUs for different markets. They also shorten time to market by reusing building blocks across client, server, and embedded lines. The result is resilient product cadence that has resonated with OEMs and hyperscalers.
Data Center Momentum with EPYC Server CPUs
EPYC platforms have scaled core counts, memory bandwidth, and I/O to address cloud native, analytics, and virtualized enterprise workloads. Recent generations like Genoa, Bergamo, and Siena target general purpose, cloud optimized, and telco edge needs. Independent evaluations consistently point to strong performance per dollar and per watt.
This translates into expanding design wins with hyperscalers, OEMs, and solution providers across geographies. As organizations prioritize density and energy efficiency, EPYC helps reduce total cost of ownership in large fleets. The franchise anchors AMD’s data center credibility and creates pull through for adjacent products.
Accelerating AI Accelerators and ROCm Ecosystem
The Instinct MI300 family integrates compute and high bandwidth memory to serve training and inference at scale. Customers and cloud providers have announced deployments that broaden the supply base for AI infrastructure and diversify beyond single vendor stacks. Architectural focus on memory capacity and bandwidth addresses large model needs.
ROCm continues to mature with support for popular frameworks, libraries, and model development workflows. Open tooling, growing community contributions, and optimizations for transformers improve time to solution. Better software reliability reduces switching costs and increases the appeal of AMD accelerators in multi vendor data centers.
Diversified Portfolio with Adaptive Computing
The Xilinx acquisition expanded AMD into adaptive SoCs, FPGAs, and embedded computing platforms with long life cycles. These products serve communications, automotive ADAS, industrial automation, medical, and aerospace applications where determinism and customization matter. They complement standard processors by accelerating specific workloads and offloading real time tasks.
Adaptive computing also strengthens cross selling opportunities in heterogeneous systems that combine CPUs, GPUs, and programmable logic. It smooths revenue across cycles because embedded demand follows different rhythms than PCs or discrete GPUs. The business tends to carry attractive margins and deep customer relationships that are hard to displace.
Strong Leadership and Strategic Partnerships
Under CEO Lisa Su, AMD has rebuilt technical credibility and operational discipline with consistent roadmap delivery. The company navigated supply constraints through close collaboration with TSMC and packaging partners while prioritizing high value segments. Clear messaging and focus have elevated the brand with developers and enterprise buyers.
Deep partnerships with hyperscalers, OEMs, and system integrators amplify reach from silicon to solutions. Co engineering efforts drive optimized platforms, certified stacks, and faster qualification cycles. This ecosystem focus translates into repeat wins, broader channel support, and visibility into customer needs.
Weaknesses
AMD has executed a strong multi-year turnaround, yet several internal constraints still hinder its full potential. These issues span manufacturing control, software maturity, product coverage, and organizational scale. Addressing them is critical as the company competes across data center, AI, client, graphics, and embedded markets.
Heavy Reliance on Third-Party Manufacturing at Advanced Nodes
AMD is fully fabless and depends on TSMC for leading-edge process nodes such as N5 and N4, which concentrates operational risk outside its direct control. Any capacity tightness, yield variability, or prioritization shifts at TSMC can constrain AMD unit availability and delay ramps. This reliance complicates pricing, mix optimization, and delivery timelines for products like EPYC, Ryzen, and Instinct MI300.
Competing against vertically integrated rivals that secure wafer starts early adds pressure to AMD’s supply planning. HBM supply for AI accelerators is another bottleneck, requiring deep coordination with memory partners to meet demand. The result is a narrower margin of error during peak cycles when customers expect predictable, volume shipments.
While ROCm has advanced with ROCm 6, PyTorch support, and broader framework compatibility, AMD still trails NVIDIA’s CUDA in maturity and breadth. Many production AI workloads, tools, and tutorials default to CUDA first, slowing AMD adoption despite competitive hardware. This creates friction for enterprises seeking seamless migration to Instinct accelerators.
Developer enablement remains an ongoing investment area for AMD, spanning kernels, libraries, model gardens, compilers, and performance tooling. Gaps in documentation depth, turnkey containers, and ecosystem plug ins can extend time to value for customers. Until parity is perceived, AMD faces higher sales and support effort per deployment.
AMD’s RDNA portfolio competes well in certain price brackets, but NVIDIA maintains a dominant position in discrete GPUs and the halo enthusiast tier. Inconsistent flagship cadence and fewer software differentiators in creator workloads reduce AMD’s pull at the top end. This limits brand signaling and attach opportunities across peripherals, monitors, and content ecosystems.
Feature parity in ray tracing, AI upscaling, and professional creator suites remains a moving target that demands sustained investment. Smaller marketing budgets and limited exclusive partnerships can compound visibility challenges. As premium segments drive mindshare and margins, AMD’s relative gaps weigh on ASPs and share recovery.
Exposure to Cyclical PC and Semi-Custom Markets, Plus Customer Concentration
Despite data center momentum, AMD’s revenue remains sensitive to PC demand swings and console cycles. Soft consumer spending or elongated refresh timelines can compress volumes and mix, pressuring margins. Semi-custom revenue is seasonal and tied to platform lifecycles, which reduces predictability.
In AI accelerators, early wins rely on a concentrated set of hyperscalers procuring MI300 at scale. This concentration introduces volatility if qualification schedules slip or if a top customer reallocates budget. Diversifying end markets and broadening enterprise adoption are required to stabilize growth.
Scale Disadvantage Versus Larger Rivals Impacting R&D and Go-to-Market
AMD’s R&D and sales resources are smaller than those of NVIDIA and Intel, forcing sharper portfolio trade offs. Funding parallel bets across CPUs, GPUs, NPUs, FPGAs, DPUs, software, and packaging can stretch teams. As product roadmaps converge on AI centric computing, the cost of staying competitive across all fronts rises.
Go to market reach and solution engineering depth are improving but still maturing in some enterprise segments. Larger rivals often bundle software, services, and reference architectures that shorten deployment cycles. AMD must keep strengthening enablement, channel programs, and field support to match buyer expectations at scale.
Opportunities
AMD is positioned to benefit from secular growth trends in AI, high performance computing, and intelligent edge devices. With chiplet leadership and a broad portfolio spanning CPUs, GPUs, NPUs, and adaptive compute, the company can expand across multiple profit pools. Execution on ecosystems and partnerships will be the catalyst.
Surging AI Accelerator Demand for Training and Inference
Global demand for AI compute and memory bandwidth is expanding rapidly, creating room for a credible second source to NVIDIA. AMD’s Instinct MI300X and MI300A, coupled with HBM rich configurations, target training and increasingly large scale inference. Wins with hyperscalers and OEMs can translate into multi year, multi generation platform footprints.
As model sizes balloon, memory capacity per GPU becomes a key differentiator that aligns with AMD’s design choices. Broadening availability of turnkey software stacks and validated clusters can shorten time to deployment. If supply scales with HBM partners, AMD can systematically grow share in AI infrastructure.
Windows AI PCs and Enterprise Endpoint Refresh
The rise of AI PCs and Copilot class experiences elevates on device AI acceleration as a purchasing driver. AMD’s Ryzen AI 300 series with XDNA 2 NPU targets 50 plus TOPS, enabling real time local inference and power efficient workflows. This creates an opportunity to capture premium notebook share as enterprises refresh fleets.
As AI workloads move on device for privacy and cost reasons, CPU GPU NPU balance becomes a design advantage for AMD. Tight co optimization with Windows, ISVs, and OEMs can differentiate battery life and responsiveness. Success here improves brand pull across consumer and commercial segments.
EPYC has steadily grown x86 server share on performance per watt, memory capacity, and TCO. The Zen 5 based Turin family extends core density and bandwidth leadership, appealing to cloud, HPC, and analytics workloads. Strong Genoa and Bergamo momentum provides a foundation for further expansion in 2025 procurement cycles.
As enterprises modernize from legacy platforms, consolidation to fewer, denser sockets benefits AMD’s economics. Platform partnerships with major OEMs and cloud instances accelerate workload certification and migration. This flywheel can raise recurring revenue from services layered on standardized EPYC fleets.
Growth in Embedded, Automotive, and Edge AI from Xilinx Integration
The Xilinx portfolio opens expanding markets in industrial automation, communications, aerospace, and automotive. Versal adaptive SoCs and FPGA platforms enable low latency, deterministic AI at the edge where power and reliability matter. These segments value long lifecycles and customization, supporting attractive margin profiles.
Combining CPUs, GPUs, and adaptive compute lets AMD deliver heterogeneous solutions for gateways, robots, and vision systems. Toolchain investments and domain specific reference designs can speed customer deployment. As edge AI proliferates, AMD can capture design wins beyond the data center.
Custom Silicon, Advanced Packaging, and HBM Partnerships
AMD’s chiplet architecture and packaging capabilities open doors to co designed silicon for hyperscalers and OEMs. 3D V Cache, HBM stacking, and interconnect advances can deliver workload specific performance gains. Custom accelerators and semi custom designs diversify revenue and deepen strategic relationships.
Closer alignment with TSMC, memory suppliers, and OSATs can secure capacity and improve cost curves. As workloads fragment across training, inference, and data processing, tailored configurations become more valuable. This approach creates defensible moats through long term roadmaps tied to customer platforms.
Threats
AMD operates in markets that shift quickly with technology cycles and capital intensity. External factors from geopolitics to supply constraints can alter demand and access to critical components. Competitive dynamics in AI accelerators and alternative architectures add persistent pressure on share and pricing.
AI accelerator arms race and Nvidia dominance
Nvidia continues to set the pace in datacenter AI with a deep software moat in CUDA, mature tools, and an aggressive hardware roadmap that includes H100, H200, and Blackwell generation parts. Although AMD’s MI300 series has gained traction, switching costs for enterprises remain high because many production workloads are optimized for Nvidia. This entrenched position can slow AMD’s share gains and force greater pricing concessions to win footprints.
At the same time, hyperscalers are rationalizing spend and measuring total cost of ownership across silicon, networking, and power. Nvidia bundles software and systems, which can compress room for AMD to differentiate beyond price and memory capacity. If customers perceive higher integration risk with alternative stacks, procurement cycles may favor incumbents and delay AMD deployments.
Supply chain concentration and HBM packaging constraints
AMD depends heavily on TSMC for leading edge nodes and advanced packaging, including CoWoS required for large AI accelerators. Industry reports through 2024 highlighted tight CoWoS capacity and long lead times, which can bottleneck shipments even when silicon is ready. Concentration risk is amplified by natural disasters and regional instability that have periodically disrupted Taiwanese production.
High bandwidth memory is another choke point, with supply led by SK hynix, Samsung, and Micron and allocations prioritizing the largest buyers. Tight HBM availability and cost inflation can erode AMD’s margin profile or constrain unit growth in accelerators. If rivals secure priority allocations, AMD could face unfavorable delivery schedules that weaken competitive positioning in large tenders.
Export controls and geopolitical tensions
Expanded United States export controls in 2023 and 2024 restricted shipments of advanced AI accelerators to China and certain regions, curbing a significant demand pool. Vendors have attempted to design compliance variants, but rule changes add uncertainty and engineering overhead. Further tightening or allied coordination could reduce accessible markets or delay approvals for new products.
Broader geopolitical risk around the Taiwan Strait and US China technology rivalry raises the odds of supply disruption or sanctions that affect semiconductor flows. Compliance burdens increase sales cycle friction, particularly for multinational customers with complex footprints. Currency volatility and divergent data sovereignty regimes can further complicate pricing, support, and localization strategies for global accounts.
Rise of ARM and alternative architectures
ARM based CPUs and integrated NPUs are accelerating in client and server markets, with Apple’s M series reshaping performance per watt expectations and Qualcomm’s Snapdragon X Elite bringing ARM to Windows laptops. Microsoft’s Copilot Plus PC initiative elevates on device AI capabilities as a purchase driver. These shifts may compress x86 share and challenge AMD’s premium positioning in notebooks.
In servers, alternatives such as AWS Graviton, Ampere ARM, Nvidia Grace, and in house accelerators reduce dependence on x86 and third party GPUs. As hyperscalers tune software stacks around their own silicon, the addressable market for merchant parts can narrow. RISC V investments add a long term vector of disruption, especially for edge and specialized workloads.
PC and gaming cyclicality with pricing pressure
Client PC demand remains cyclical and sensitive to macroeconomic conditions, with the post pandemic digestion and inventory corrections creating uneven recovery patterns. While AI PCs may lift premium tiers, mainstream buyers often prioritize price, which can pressure average selling prices. Competitive promotions from rivals can intensify during weaker quarters and weigh on margins.
Gaming revenue is also cyclical, with console generations aging and seasonal promotions resetting expectations on price performance. Discrete GPU demand can swing with new title launches, streamer trends, and regional economic shifts. If consumer wallets tighten or upgrade cycles elongate, unit volumes and mix can shift toward lower margin products.
Challenges and Risks
Inside the company, execution and operations determine how well AMD converts opportunity into sustained share and profits. The pace of innovation, manufacturing economics, and customer mix all influence outcomes. Managing these factors during rapid AI growth is complex and resource intensive.
Software ecosystem and ROCm maturity
Bridging the gap with Nvidia’s CUDA remains a core challenge despite visible progress in ROCm, framework compatibility, and libraries. Developers expect turnkey stability, broad operator coverage, and first class support for PyTorch, TensorFlow, and inference runtimes. Any regressions in drivers or toolchains can slow adoption and increase cost to serve strategic accounts.
Independent software vendor certification and performance tuning across hundreds of models require sustained investment and close collaboration. Enterprise customers value predictable roadmaps for features like mixed precision, memory virtualization, and cluster management. Without comparable depth and ease of use, AMD risks being limited to cost sensitive deployments rather than becoming a default choice.
Product roadmap and execution timing
Coordinating CPU, GPU, NPU, and adaptive compute roadmaps is difficult when leading edge nodes, chiplets, and packaging all move in parallel. Schedule slips or yield challenges can cascade across platforms and OEM launch windows. Missing a key back to school or holiday build can forfeit design wins that are hard to recover.
Server buyers plan multi year refreshes and require stable firmware, security features, and platform validation. If platform enablement lags, competitors can lock in sockets with long lifecycles. Balancing rapid feature delivery with reliability expectations is an ongoing risk for brand reputation.
Margin mix and cost structure
AI accelerators rely on expensive components such as HBM stacks and advanced packaging, which can compress gross margins if pricing softens. Any yield variability on large dies or interposers exacerbates cost per unit. As the mix tilts toward accelerators, profitability hinges on tight cost control and negotiated component pricing.
On the client side, promotions to defend share against ARM and x86 rivals can dilute margins. Semi custom console contracts typically carry lower margins yet consume engineering resources over long commitments. Navigating this mix while funding heavy R and D and software investment is a persistent financial balancing act.
Customer concentration and demand volatility
Datacenter revenue is increasingly concentrated among a handful of hyperscalers and large cloud service providers. Their procurement cycles can shift abruptly as architectures or budgets change, creating quarter to quarter volatility. Price leverage from these buyers can pressure terms and service level expectations.
Semi custom revenue is tied to the strategies of a few gaming platform owners, with limited control over volumes late in the cycle. In client PCs, OEMs may rebalance vendor exposure due to platform constraints or marketing incentives. This concentration amplifies the impact of any single account’s delay or reprioritization.
Integration and talent retention
Integrating Xilinx and Pensando capabilities into coherent roadmaps and unified software is strategically valuable but organizationally complex. Cross selling adaptive compute with CPUs and GPUs requires aligned sales motions and support. Inefficiencies during integration can slow revenue synergies and distract engineering focus.
The global war for AI hardware and software talent raises compensation and retention risks. Losing key architects or developer relations leaders can set back platform credibility. Building deep expertise in compilers, graph optimization, and distributed training at scale is critical and time consuming.
Strategic Recommendations
To strengthen competitive position, AMD should align investments with the most durable growth vectors while reducing operational fragility. The emphasis should be on software leadership, resilient supply, targeted customer collaboration, and platform level value. These actions tie directly to external threats and internal execution risks.
Invest aggressively in ROCm and developer experience
Close the ecosystem gap by prioritizing end to end reliability, from drivers to cluster orchestration, with measurable service level objectives for key frameworks. Expand engineering embedded with top model builders to upstream optimizations into PyTorch, JAX, and inference stacks. Accelerate pre optimized model suites and reference pipelines that demonstrate time to accuracy and cost advantages.
Formalize a certification program with ISVs and cloud marketplaces that guarantees versioned support and performance baselines. Grow community adoption through grants, hackathons, and simplified installers that reduce friction on popular Linux distributions. Where possible, champion open standards like OpenXLA and Triton to broaden portability and reduce reliance on competitor controlled ecosystems.
Lock in advanced capacity and HBM through strategic agreements
Secure multi year supply and packaging reservations with TSMC and alternative advanced packaging partners, including capacity options for demand surges. Negotiate long term HBM agreements with multiple suppliers that include price bands and priority allocation. Co invest where feasible in substrate and CoWoS expansions to mitigate bottlenecks observed through 2024.
Develop second source contingencies for critical steps such as test and assembly to reduce single point failures. Standardize chiplet interfaces to allow flexible memory configurations across product tiers, improving yield harvesting and cost control. Maintain export compliant variants in reserve to address policy shifts without derailing build plans.
Expand custom silicon and chiplet platforms with hyperscalers
Leverage chiplets, Infinity Fabric, and Xilinx adaptive compute to co develop domain specific accelerators and network offload engines with top cloud providers. Offer modular blueprints that integrate CPUs, GPUs, NPUs, and DPUs with tailored memory footprints. This approach can secure anchor commitments while differentiating on power, latency, and total system cost.
Create joint optimization teams that align compiler work, kernel tuning, and data pipeline tooling with each customer’s stack. Share clear cost roadmaps that reward volume and early collaboration on next generation features. Custom engagements can reduce head to head price battles and build stickier, multi year partnerships.
Double down on efficiency leadership by tuning silicon, firmware, and platform power policies for real world workloads. Partner closely with OEMs on thin and light Ryzen AI designs that meet Copilot Plus readiness while sustaining battery life and thermals. Provide turnkey reference designs to compress time to market and ensure consistent user experience.
In servers, deliver hardened platform kits with validated firmware, security features, and management tools aligned to enterprise standards. Expand channel enablement, sizing tools, and migration playbooks to reduce perceived switching risk from incumbent platforms. Maintain pricing discipline while positioning total cost of ownership advantages that resonate across procurement and operations teams.
Competitor Comparison
AMD competes head-to-head with Intel in x86 CPUs and with NVIDIA in discrete GPUs, while also intersecting with Qualcomm, Apple, and specialized silicon vendors in select segments. The landscape is dynamic, with performance, power efficiency, software stacks, and supply partnerships shaping relative advantage.
Brief comparison with direct competitors
Against Intel, AMD’s EPYC and Ryzen lines have gained share by leveraging high core counts, energy efficiency, and advanced process nodes delivered through foundry partners. Intel retains scale, incumbency with OEMs, and deep platform validation, but AMD has narrowed gaps in enterprise credibility and platform breadth.
Versus NVIDIA, AMD’s Instinct and Radeon portfolios challenge a dominant AI and graphics ecosystem centered on CUDA and strong developer tooling. NVIDIA’s software moat and data center mindshare remain formidable, yet AMD competes on performance per dollar, open frameworks, and broader memory configurations in select workloads.
Key differences in strategy, marketing, pricing, innovation
Strategically, AMD focuses on high-impact segments where architectural differentiation and chiplet design deliver outsized value. Intel emphasizes vertically integrated platforms and manufacturing reinvention, while NVIDIA prioritizes end-to-end AI stacks, networking, and accelerated computing platforms to extend its ecosystem leverage.
Marketing and pricing diverge as AMD often uses aggressive performance-per-dollar positioning to convert accounts and win TCO narratives. NVIDIA typically prices to its software and platform premium, and Intel leans on incumbency and fleet-standardization economics, leaving AMD room to craft targeted bundles and partnerships.
How AMD’s strengths shape its position
AMD’s strengths include advanced chiplet architectures, strong foundry execution, and versatile product roadmaps spanning client, data center, and adaptive computing. These assets help it deliver rapid generational gains, competitive total cost of ownership, and flexible configurations for cloud, enterprise, and OEM customers.
Its openness to industry standards, combined with expanding software support, reduces switching friction and broadens partner engagement. As performance parity tightens in key battlegrounds, AMD’s credibility, supply diversity, and cross-portfolio synergies reinforce its position as the primary alternative to entrenched leaders.
Future Outlook for AMD
AMD’s near-term outlook is anchored by secular growth in AI, data center compute, and power-efficient client platforms. Execution in software ecosystems and deep partnerships will determine how fully it converts technology advantages into durable share gains and margins.
Data center AI and accelerated computing
AI training and inference represent the largest expansion vector, where AMD can compete through high-bandwidth memory designs, strong perf-per-watt, and maturing software stacks. Continued investment in ROCm, frameworks interoperability, and reference solutions will be critical to unlocking developer adoption and repeatable deployments.
In CPUs, EPYC is positioned to benefit from core density, memory channels, and TCO advantages as cloud and enterprise refresh cycles advance. Blending CPUs, GPUs, and adaptive accelerators into validated platforms can drive larger deal sizes and entrenchment with hyperscalers and OEMs.
Client, gaming, and edge momentum
On the client side, efficiency-focused CPUs with integrated AI acceleration can capture AI PC demand while preserving battery life and performance headroom. In gaming, leadership in perf-per-dollar and technologies that balance ray tracing with upscaling will influence enthusiast and OEM designs.
At the edge, industrial, telecom, and embedded use cases benefit from AMD’s adaptive computing and heterogeneous architectures. Strategic SKUs tuned for thermal envelopes and long-lifecycle support can expand TAM and create sticky design wins across verticals.
Risks, constraints, and execution priorities
Primary risks include supply-demand imbalances in advanced nodes, competitive software moats, and rapid product cycles that compress pricing power. AMD must sustain cadence in roadmaps while strengthening developer relations, ISV certification, and turnkey solutions.
Capital discipline, ecosystem incentives, and co-selling with partners can accelerate adoption and mitigate incumbency barriers. Clear messaging on workload outcomes, not just benchmarks, will help translate technical strengths into sustained revenue mix improvement and margin expansion.
Conclusion
AMD stands at a favorable juncture where architectural innovation, chiplet leadership, and power efficiency align with market demand for AI and high-performance compute. Its ability to pair competitive silicon with maturing software and strong partnerships will define the depth of its gains.
While NVIDIA’s software ecosystem and Intel’s incumbency remain significant hurdles, AMD’s cross-portfolio strengths create credible alternatives in data center, client, and edge markets. Focused execution on platforms, developers, and supply resilience can turn momentum into durable share and profitability.
