The First Product Manager Who Became the Third CEO

On June 3, 2024, Matt Garman officially became the third CEO of Amazon Web Services, inheriting a business generating over $100 billion in annual revenue and commanding roughly 30% of the global cloud infrastructure market. For most observers, the appointment was noteworthy but unsurprising—AWS had announced the leadership transition four weeks earlier, giving Garman time to prepare for one of the most consequential roles in enterprise technology. But beneath the surface of this orderly succession lay a remarkable story: Garman had joined AWS as a summer intern 19 years earlier, when the service existed only as an unnamed "internal Amazon tech startup" that Andy Jassy was pitching to MBA students.

Garman's journey from intern to CEO traces the entire arc of cloud computing's transformation of enterprise IT. He was AWS's first product manager, helping define the service level agreements and pricing models that made cloud infrastructure practical for businesses. He built and launched Amazon EC2's elastic block storage, creating the persistent disk capabilities that convinced enterprises they could run serious workloads in the cloud. He spent four years leading all of AWS Compute, the foundation on which every other AWS service depends. Then, in a move that signaled his path toward the CEO role, he pivoted to sales leadership, spending four years understanding how customers actually buy, deploy, and extract value from cloud services.

Now, as CEO, Garman faces challenges that are both familiar and unprecedented. The familiar part: defending AWS's market-leading position against aggressive competition from Microsoft Azure and Google Cloud. The unprecedented part: doing so during the biggest platform shift since cloud computing itself—the emergence of AI as a fundamental infrastructure layer that could remake the economics, competitive dynamics, and strategic positioning of the entire cloud industry.

Part I: The Intern Years—Joining an Unnamed Startup Inside Amazon

Summer 2005: When Cloud Computing Was Still Stealth Mode

In the summer of 2005, Matt Garman was a first-year MBA student at Northwestern University's Kellogg School of Management, looking for an internship that would let him apply the product management skills he'd developed during five years at startups after graduating from Stanford with degrees in industrial engineering. The opportunity that caught his attention wasn't from a prestigious consulting firm or tech giant—it was from Amazon, specifically from an executive named Andy Jassy who was recruiting for what he described only as an "unnamed internal Amazon tech startup."

Jassy couldn't reveal many details during the recruiting process. AWS was still in stealth mode, a secretive project that most Amazon employees knew nothing about. What Jassy could say was that Amazon was building infrastructure services that would let external developers use the same robust, scalable systems that Amazon had built for its own e-commerce operations. The vision was audacious: Amazon would expose its infrastructure as services that any developer could consume via APIs, paying only for what they used, with no upfront commitments or long-term contracts.

For Garman, the pitch resonated. His background in industrial engineering gave him appreciation for infrastructure, systems thinking, and operational efficiency. His startup experience taught him how painful it was to provision servers, manage data centers, and scale infrastructure—all tasks that distracted from building actual products. If Amazon could abstract away that complexity and let developers focus on applications rather than infrastructure, it could be transformative.

Garman accepted the internship and spent summer 2005 working on what would become AWS. The timing was perfect—he got to see cloud computing's foundational concepts take shape while they were still malleable enough to be influenced by fresh thinking. More importantly, he impressed Jassy enough that when Garman graduated in 2006—the same year AWS officially launched with S3, SQS, and EC2—he received a full-time offer as AWS's first product manager.

2006-2010: Building the Foundations of Cloud Computing

When Garman joined AWS full-time in 2006, the organization was tiny—only three people in sales, a small engineering team, and a product management function that consisted entirely of Garman. His early responsibilities spanned everything from defining service level agreements to creating pricing models to specifying new features that customers needed. It was the ultimate generalist product role, requiring both technical depth and business acumen, combined with the willingness to do whatever needed doing.

One of Garman's most significant early contributions was helping establish AWS's service-level agreements (SLAs) and pricing transparency. Traditional enterprise IT had been defined by opaque pricing, multi-year contracts, and service levels that were more aspirational than guaranteed. AWS's approach—publish prices publicly, commit to specific uptime percentages, and offer service credits when SLAs weren't met—was revolutionary. Garman helped translate Jassy's vision of customer obsession into concrete policies that customers could depend on.

Garman also worked on early feature definition for EC2, which launched in August 2006 as AWS's compute service. The initial version offered virtual servers that customers could start and stop via APIs, paying by the hour for capacity. But customers quickly identified a critical gap: EC2 instances had only "ephemeral" storage that disappeared when instances shut down. For any serious workload—databases, file servers, stateful applications—customers needed persistent storage that survived instance restarts.

This led to one of Garman's most important projects: leading the team that defined, launched, and operated Elastic Block Storage (EBS). EBS, which launched in August 2008, provided network-attached block storage that could be attached to EC2 instances but persisted independently. It sounds simple in retrospect, but EBS was technically complex—ensuring low latency, high durability, and seamless integration with EC2 required solving numerous distributed systems challenges. More importantly, EBS was strategically crucial: it convinced enterprises that they could run production databases and business-critical applications on AWS rather than just using cloud for development and testing.

Part II: The Product Leadership Years—Building AWS Compute

2010-2020: From EC2 Product Manager to Compute Services GM

Garman's success with EBS led to expanding responsibilities. He became the first product manager specifically focused on EC2, rather than working across all AWS services. This specialization reflected AWS's growth—the organization was large enough to need dedicated product managers for individual services, and EC2 was important enough to warrant full-time product leadership.

As EC2 product manager, Garman oversaw rapid innovation in instance types, pricing models, and capabilities. EC2 Reserved Instances (launched 2009) let customers commit to multi-year usage in exchange for significant discounts, creating predictable revenue for AWS while reducing costs for customers. EC2 Spot Instances (launched 2009) let customers bid on unused capacity, creating a market mechanism that improved AWS's infrastructure utilization while giving price-sensitive customers deep discounts.

In 2016, Garman was promoted to General Manager of AWS Compute Services, overseeing not just EC2 but the entire compute portfolio: Lambda (serverless computing), Elastic Container Service, Lightsail (simplified virtual private servers), and AWS Batch. This role required orchestrating multiple product teams, balancing investments across different compute paradigms, and maintaining coherent strategy as compute diversified from simple virtual machines to containers, serverless functions, and specialized workloads.

The Compute GM role also put Garman at the center of critical strategic decisions about AWS's future. Should AWS build its own custom silicon, or continue relying on Intel and AMD processors? Should AWS compete directly with Kubernetes, or embrace it despite it being developed by Google? How aggressively should AWS push serverless computing, which cannibalized higher-margin EC2 usage? These weren't just product questions—they were bets about how cloud computing would evolve and where AWS needed to maintain leadership.

The Custom Silicon Decision That Would Shape AI Infrastructure

One of the most consequential decisions during Garman's compute leadership tenure was AWS's move into custom silicon design. In 2015, AWS acquired Annapurna Labs, an Israeli chip design company, signaling intentions to develop proprietary processors rather than depending entirely on Intel and AMD. This led to AWS Graviton processors (ARM-based chips optimized for cloud workloads) and, crucially for the AI era, AWS Inferentia and Trainium chips designed specifically for machine learning inference and training.

At the time, the custom silicon strategy faced skepticism both internally and externally. Intel and AMD had decades of experience and massive R&D budgets. NVIDIA dominated accelerated computing with GPUs that were becoming increasingly important for machine learning. Why would AWS, a cloud infrastructure company, try to design chips—one of the most capital-intensive, technically complex, and risky endeavors in technology?

Garman and his team made several arguments that ultimately proved prescient. First, custom silicon could be optimized specifically for cloud workloads in ways that general-purpose processors couldn't match—Graviton could deliver better price-performance for the specific mix of operations that AWS customers actually ran. Second, vertical integration would reduce dependence on external suppliers who might prioritize other customers or whose roadmaps might not align with AWS's needs. Third, as AWS's scale grew, the economics of custom chip development improved—spreading NRE (non-recurring engineering costs) across millions of instances made chips economically viable that wouldn't make sense at smaller scale.

The AI boom has vindicated this strategy spectacularly. NVIDIA GPU shortages in 2023-2024 left cloud providers scrambling for capacity and AI startups unable to access the compute they needed. AWS's Trainium chips, while not matching NVIDIA's top-end performance, provide alternatives that reduce dependence and often deliver better price-performance for specific workloads. As Andy Jassy put it in earnings calls, AWS having multiple silicon options—Intel, AMD, NVIDIA, and its own Graviton/Trainium/Inferentia chips—creates pricing leverage and supply security that purely NVIDIA-dependent competitors lack.

Part III: The Sales Pivot—Learning to See AWS Through Customers' Eyes

2020: The Surprising Move to Sales Leadership

In 2020, after 14 years building AWS's product organization, Garman made a career move that surprised many observers: he became Senior Vice President of AWS Sales, Marketing, and Global Services, leaving product management entirely. For someone who had spent his entire AWS career on the product side, the pivot seemed risky. Sales and product require different skill sets, different organizational approaches, and different mindsets. Why would AWS's most experienced compute product leader take on sales responsibilities he'd never held before?

The answer, according to people familiar with Jassy's thinking, was succession planning. Jassy knew that whoever succeeded him as AWS CEO would need to understand not just what AWS built but how customers actually bought it, deployed it, and extracted value from it. AWS's sales motion had grown enormously complex—from simple self-service signups to multi-year enterprise agreements worth hundreds of millions of dollars, from transactional relationships to strategic partnerships that shaped customers' entire technology strategies.

Garman approached sales leadership with product manager discipline. He spent time with sales teams understanding their processes, pain points, and what information they needed to close deals. He reviewed customer conversations to understand objections, feature requests, and competitive dynamics. He participated in executive business reviews with AWS's largest customers, learning how they made build-vs-buy decisions, evaluated AWS against Azure and Google Cloud, and measured ROI on cloud investments.

The sales role also exposed Garman to AWS's partner ecosystem in ways the product role hadn't. AWS's business increasingly depended on consulting partners (Accenture, Deloitte, KPMG) who helped customers architect and deploy cloud solutions, technology partners (VMware, SAP, Salesforce) who integrated their products with AWS, and managed service providers who operated AWS infrastructure on customers' behalf. Understanding how these partners made money, what motivated their recommendations, and what AWS could do to make them more successful became crucial parts of Garman's education.

The COVID-19 Acceleration and Digital Transformation

Garman's timing in taking the sales role proved fortuitous—or perhaps Jassy's timing in moving him was prescient. The COVID-19 pandemic hit months after Garman's transition, accelerating enterprise cloud adoption dramatically. Companies that had been gradually planning multi-year migrations to the cloud suddenly needed to enable remote work in weeks. Digital transformation timelines compressed from years to months as pandemic lockdowns made in-person operations impossible.

AWS's sales organization, under Garman's leadership, had to scale rapidly to meet exploding demand while also helping customers navigate unprecedented challenges. How could retailers handle 10x normal e-commerce traffic when customers couldn't visit physical stores? How could schools and universities suddenly deliver education online? How could healthcare providers handle telemedicine at massive scale? How could financial services firms maintain operations with employees working from home?

Garman's product background proved valuable in translating customer challenges into technical solutions. He could bridge conversations between sales teams who understood business needs and engineering teams who could build capabilities to address them. He could evaluate when AWS should build new services to address pandemic-driven use cases versus when customers should combine existing services creatively. And he could make judgment calls about when to provide customer concessions—like waiving data transfer fees or providing free tier extensions—that built customer loyalty even if they reduced short-term revenue.

Part IV: The CEO Transition—Inheriting AWS at a Critical Moment

May 2024: Adam Selipsky Steps Down

On May 14, 2024, AWS CEO Adam Selipsky announced he was stepping down, with Matt Garman named as his successor effective June 3, 2024. The announcement came as AWS faced mounting questions about its AI strategy and whether it was falling behind Microsoft and Google in the AI race. Selipsky had led AWS for three years after Andy Jassy became Amazon CEO in 2021, but investor and customer concerns about AWS's AI positioning had intensified throughout 2023 and early 2024.

The timing of Selipsky's departure sparked speculation. AWS had just reported Q1 2024 growth of 17.5%—respectable but slower than Azure's 39% and Google Cloud's 32% growth in the same period. Much of that delta came from AI workloads, where Microsoft's OpenAI partnership and Google's proprietary models were winning startup customers and enterprise AI projects. While AWS offered competitive AI services through Amazon Bedrock and SageMaker, the perception was growing that AWS was losing the AI infrastructure battle despite dominating traditional cloud.

Garman's appointment signaled AWS's response to these concerns. Unlike Selipsky, who had come from outside AWS (serving as CEO of Tableau before returning to AWS), Garman had spent his entire career building AWS. He understood the technology deeply from his product years, understood customers intimately from his sales years, and represented continuity with AWS's culture and principles. Most importantly, he represented a bet that AI infrastructure wasn't fundamentally different from cloud infrastructure—it required the same customer obsession, operational excellence, and long-term thinking that had built AWS's cloud business.

June 2024: The First 100 Days

Garman's early actions as CEO focused on reassuring both customers and employees that AWS's fundamental strategy remained sound even as AI reshaped the landscape. In employee communications and customer meetings, he emphasized that generative AI was extraordinarily important but also that it built on the same foundational services—compute, storage, databases, networking—that AWS had been perfecting for 18 years. Companies needed infrastructure to train models, infrastructure to serve inference, infrastructure to store training data, and infrastructure to build AI-powered applications. AWS's scale, operational expertise, and customer obsession positioned it well to provide all of that.

Garman also moved quickly to strengthen AWS's AI partnerships. Within months of becoming CEO, AWS announced significant developments in its Anthropic relationship ($8 billion total investment) and secured a major new partnership with OpenAI (multi-year agreement worth potentially $38 billion). These announcements addressed the perception that AWS lacked preferred access to leading foundation models—now AWS could offer Claude, ChatGPT, and numerous other models through Amazon Bedrock while also providing its own Nova models for customers who wanted AWS-native options.

Internally, Garman worked to ensure that AWS's product roadmap aligned with AI requirements. This meant accelerating Trainium chip development, enhancing Bedrock capabilities, simplifying SageMaker for broader accessibility, and building new services specifically for AI workloads. But it also meant not abandoning AWS's core businesses—continuing to innovate in databases, storage, networking, and traditional compute even as AI captured headlines.

Part V: The December 2024 re:Invent—Garman's Strategic Vision

The First CEO Keynote

Garman's keynote address at AWS re:Invent in December 2024 marked his public debut as AWS CEO and provided the clearest articulation of his strategic vision. Speaking to 60,000 attendees in Las Vegas and hundreds of thousands more virtually, Garman emphasized a message of balanced innovation: "We invent, so you can reinvent."

The keynote was deliberately structured to demonstrate that AWS wasn't betting everything on AI at the expense of core infrastructure. Garman spent significant time discussing next-generation Graviton processors (45% better performance on Java workloads, 60% lower energy consumption), new storage capabilities (S3 Tables supporting Apache Iceberg, celebrating 400 trillion objects stored in S3), and database innovations (enhanced Aurora capabilities, new Neptune features). The message: AWS would continue leading in foundational infrastructure regardless of whether AI proved transformational or overhyped.

But Garman also showcased major AI announcements that addressed concerns about AWS falling behind. Andy Jassy made a surprise appearance to announce Nova, AWS's new family of foundation models, signaling that AWS would compete not just as infrastructure provider but also as model developer. Garman introduced Bedrock Model Distillation (simplifying large models for deployment efficiency), Bedrock Guardrails (ensuring accuracy and safety), and Bedrock Agents (enabling multi-agent collaboration and workflows).

Perhaps most significantly, Garman announced a $1 billion global fund for startups launching in 2025, building on AWS Activate's existing program that had helped 280,000 startups and provided over $6 billion in credits. This directly addressed the narrative that Azure and Google Cloud were winning AI startups through aggressive compute credit programs—AWS would match or exceed competitor incentives to ensure the next generation of AI companies built on AWS from the beginning.

The Infrastructure Bet: Trainium 3 and Beyond

One of the most strategically important re:Invent announcements was Trainium 3, the next generation of AWS's custom AI training chip scheduled for 2025. Garman positioned Trainium 3 as offering dramatic performance improvements while maintaining price advantages over NVIDIA's competing offerings. But more importantly, he framed custom silicon as essential to AWS's long-term AI strategy for reasons that extended beyond pure performance or cost.

First, Trainium gives AWS control over its supply chain at a time when GPU shortages constrain competitors. While Azure and Google Cloud compete with gaming companies, cryptocurrency miners, and research labs for limited NVIDIA production, AWS can direct chip fabrication capacity toward its specific needs. As demand for AI compute continues growing faster than chip production can scale, this supply security becomes increasingly valuable.

Second, Trainium allows optimization for the specific workloads AWS customers actually run. General-purpose GPUs must support everything from gaming to scientific computing to AI. Trainium can be optimized exclusively for training and inference patterns that AWS observes across millions of customer workloads, potentially delivering better real-world performance than benchmark-optimized chips.

Third, Trainium creates competitive differentiation that's difficult to replicate. Microsoft can license GPT models from OpenAI, and Google has its own models from DeepMind. But neither has invested as heavily in custom ML silicon as AWS, and catching up would require years and billions of dollars. If Trainium delivers on its promise, it becomes a durable competitive advantage that compounds over time as AWS's chip design expertise deepens and its silicon roadmap extends further.

Part VI: The Competitive Challenge—Defending 30% Market Share

The Azure Threat: Microsoft's Integration Advantage

Matt Garman's most immediate competitive challenge comes from Microsoft Azure, which has narrowed AWS's market share lead substantially. While AWS still holds roughly 30% of cloud infrastructure spending versus Azure's 20%, the growth rate gap is concerning—Azure growing at 39% versus AWS's 17.5% in recent quarters suggests that the gap could close within a few years if trends continue.

Azure's competitive advantage is integration with Microsoft's existing enterprise software footprint. Companies already using Office 365, Windows Server, Active Directory, Dynamics, and other Microsoft products face significantly lower friction in adopting Azure than AWS. Microsoft can offer unified billing, single support relationships, common identity management, and enterprise agreements that span all Microsoft products. For CIOs managing complex IT estates, this integration reduces complexity even if Azure's individual services aren't superior to AWS's.

The AI era has amplified Azure's integration advantage. Microsoft Copilot—AI capabilities embedded throughout Office, Windows, Dynamics, and other Microsoft products—runs on Azure infrastructure using models from OpenAI (in which Microsoft invested $13 billion). Customers adopting Copilot naturally end up using more Azure services. And Microsoft can bundle AI capabilities with existing enterprise agreements, making it financially attractive to expand Azure usage rather than adding separate AI spending through AWS.

Garman's response emphasizes AWS's neutrality and customer choice. Unlike Microsoft, which pushes customers toward its own software stack, AWS supports whatever technologies customers want to use—Linux or Windows, Oracle or PostgreSQL, SAP or Salesforce. AWS's partner ecosystem is more open precisely because AWS doesn't compete with partners the way Microsoft does. And AWS's focus on infrastructure rather than applications means AWS can prioritize customer needs rather than promoting its own adjacent businesses.

The Google Cloud Challenge: Model Leadership and Price Pressure

While Google Cloud Platform holds only about 13% market share—roughly half Azure's—it poses distinct competitive challenges. Google's deep AI research heritage (Google Brain and DeepMind made foundational contributions to modern AI) gives GCP credibility with AI-focused startups and data science teams. Google's Gemini models compete directly with GPT-4 and Claude, and being able to access Google's proprietary models only through GCP creates exclusivity that AWS can't match with Bedrock's multi-model approach.

Google is also willing to use aggressive pricing and compute credits to win AI startups. Multiple venture-backed AI companies have reported receiving millions of dollars in GCP credits, sometimes as part of Google's investment in their funding rounds. This "pay-to-play" approach works because AI startups have enormous compute needs but limited revenue—free credits can make the difference between building on GCP versus AWS or Azure.

Garman has responded by doubling down on AWS's startup program. The $1 billion fund announced at re:Invent 2024 is explicitly designed to match or exceed competitor credits. But Garman is also emphasizing AWS's scale advantages—AWS has more regions, more availability zones, more services, and more production-proven infrastructure than GCP. For startups that grow beyond the credit-supported phase, AWS's maturity becomes more valuable than free compute.

Part VII: The OpenAI and Anthropic Partnerships—Strategic Alignment

November 2024: The $38 Billion OpenAI Deal

In November 2024, AWS announced a multi-year strategic partnership with OpenAI valued at potentially $38 billion over multiple years. The deal makes AWS a primary infrastructure provider for OpenAI's workloads and brings ChatGPT Enterprise and OpenAI's models to AWS Bedrock. For Garman, barely five months into his CEO tenure, securing this partnership represented a major strategic victory that directly addressed AWS's perceived weakness in AI partnerships.

The OpenAI partnership is strategically important for several reasons. First, it legitimizes AWS as a serious AI infrastructure player—OpenAI, the company that launched the gen-AI revolution with ChatGPT, choosing AWS validates AWS's technical capabilities. Second, it gives AWS customers access to GPT models through Bedrock, closing a gap where previously only Azure had preferred OpenAI integration. Third, it creates a hedge against Microsoft's OpenAI exclusivity—now both Microsoft and AWS have significant OpenAI relationships, reducing Microsoft's differentiation.

But the deal also reflects shifting power dynamics in AI infrastructure. When Microsoft invested $13 billion in OpenAI in early 2023, Microsoft had leverage—OpenAI desperately needed capital and compute, and few companies could provide both at OpenAI's scale. By late 2024, OpenAI's position had strengthened (approaching $4 billion in revenue, multiple funding offers) while Microsoft's AI exclusivity was under regulatory scrutiny. This created space for AWS to negotiate a partnership without Microsoft-style exclusivity constraints.

The Anthropic Relationship: $8 Billion and Deep Technical Collaboration

AWS's relationship with Anthropic predates Garman's CEO tenure—the partnership began in September 2023 with a $4 billion investment and deepened with an additional $4 billion in November 2024. But under Garman's leadership, the Anthropic partnership has become central to AWS's AI strategy in ways that extend beyond just offering Claude through Bedrock.

The technical collaboration with Anthropic is particularly deep around custom silicon. Anthropic has committed to using AWS Trainium chips for model training, with plans to use one million Trainium chips by end of 2025. AWS announced Project Rainier, a massive cluster of 500,000 Trainium 2 chips dedicated to Anthropic's model training. This represents validation that AWS's custom silicon can compete with NVIDIA GPUs for frontier model development—arguably the most demanding AI workload that exists.

For Garman, the Anthropic partnership serves multiple strategic purposes. It demonstrates Trainium's capabilities at the highest performance tier. It creates a reference customer for enterprises considering Claude—if Anthropic itself trusts AWS infrastructure, why shouldn't others? And it provides feedback for AWS's AI service development—Anthropic's needs in training massive models, deploying inference at scale, and managing AI workloads inform AWS's product roadmap in ways that benefit all customers.

Part VIII: The Long-Term Vision—Platform Thinking in the AI Era

Learning from AWS's Cloud Computing Playbook

Garman's strategic approach to AI infrastructure mirrors the playbook that built AWS's cloud business. When AWS launched in 2006, few people believed that enterprises would trust critical workloads to infrastructure run by an online retailer. The conventional wisdom was that companies needed to own their data centers, control their own servers, and maintain physical security of their systems. AWS succeeded not by fighting this mindset directly but by proving value through adoption—startups used AWS because it was cheaper and faster than building data centers, then grew into substantial businesses running entirely on AWS, demonstrating that cloud infrastructure could work at scale.

Garman is applying similar thinking to AI infrastructure. Rather than claiming AWS will build the best foundation models or win the AI race through superior technology, he's positioning AWS as the essential infrastructure layer that every AI company depends on—whether they're training their own models, fine-tuning existing ones, or building applications on top of foundation models from others. This platform strategy has several advantages:

Model agnosticism—AWS wins whether Claude, GPT, Llama, Gemini, or some future model becomes dominant, as long as they run on AWS infrastructure.

Use case agnosticism—AWS captures value from AI being used for customer service, code generation, medical diagnosis, financial analysis, or thousands of other applications.

Maturity spectrum coverage—AWS serves customers at every stage from experimentation (Bedrock's simple APIs) to production deployment (SageMaker's full ML lifecycle management) to building custom infrastructure (raw access to Trainium clusters).

The Startup Strategy: Building Tomorrow's AWS Customers Today

One of Garman's most important strategic priorities is ensuring that AI-native startups build on AWS from the beginning. This matters because today's AI startups could become tomorrow's massive AWS customers—just as today's largest AWS customers (Netflix, Airbnb, Pinterest) started as scrappy startups that grew up on AWS and now spend hundreds of millions annually.

The AWS Activate program, which Garman expanded with the $1 billion commitment announced at re:Invent 2024, is central to this strategy. By providing generous credits to early-stage companies, AWS reduces the financial barrier to choosing AWS over cheaper alternatives or open-source self-hosting. Once startups build on AWS—adopting Bedrock for model access, SageMaker for training, EC2 for compute, S3 for storage—switching costs grow as they integrate more deeply with AWS services.

Garman has also emphasized AWS's unique value proposition for AI startups beyond just credits. AWS's global infrastructure lets startups deploy AI applications close to users worldwide, reducing latency. AWS's security and compliance capabilities help startups sell to enterprises with stringent requirements. AWS's partner ecosystem provides startups with access to consulting expertise, technology integrations, and go-to-market support that they couldn't afford to build internally.

Conclusion: The Platform Leader for AI's Next Decade

Matt Garman's journey from AWS intern to CEO embodies the company's entire history—starting with an audacious bet that infrastructure could be provided as a service, proving that cloud computing could work at global scale, and now evolving to meet the demands of AI's explosive growth. His unique combination of product depth (14 years building core infrastructure), sales understanding (4 years learning how customers actually buy and deploy cloud), and institutional knowledge (19 years watching AWS grow from startup to $100 billion business) positions him distinctively to lead AWS through its next phase.

The challenges are immense. Microsoft Azure is growing faster and leveraging integration advantages that AWS can't easily counter. Google Cloud is using aggressive pricing and proprietary models to win AI-focused startups. AI workload economics remain uncertain—will customers pay enough for inference to make AI infrastructure as profitable as traditional cloud? Regulatory scrutiny is intensifying around data privacy, model training, and the concentration of AI capabilities among a few large platforms.

But Garman is playing a strategic game that extends beyond quarterly growth rates or near-term market share battles. His vision is that AI infrastructure, like cloud infrastructure before it, will become a massive, enduring market where the winners are determined not by who has the best technology today but by who builds the most comprehensive platform, earns the deepest customer trust, and creates the strongest ecosystem effects.

AWS under Garman is betting that platform advantages compound over time. Every startup that builds on AWS becomes a potential enterprise customer. Every workload that starts on AWS tends to expand into adjacent services. Every AWS region that opens creates network effects as customers want to deploy applications globally. Every AWS innovation—whether in custom silicon, AI services, or operational capabilities—gets leveraged across millions of customer workloads, creating scale advantages that competitors can't match.

Whether this platform strategy will succeed in the AI era remains to be seen. But Garman's approach—combining AWS's cloud computing DNA with rapid innovation in AI infrastructure, balancing aggressive competition with customer-centric principles, and thinking in decades rather than quarters—represents the most comprehensive answer yet to the question of how a cloud leader defends its position when a new platform emerges.

The intern who joined an unnamed Amazon startup in 2005 is now leading that startup—grown into one of the world's most important technology businesses—through its biggest test yet. The next few years will determine whether AWS maintains its infrastructure leadership through yet another platform shift, or whether the AI revolution creates new winners that displace cloud's old guard. Matt Garman's tenure will be judged on which of those outcomes prevails.