From Renting Machines by the Hour to Renting Capabilities by the MSeconds • Dhaval Nagar • GOTO 2025
Dhaval Nagar chronicles the evolution of cloud economics from hourly-billed virtual machines a decade ago to the current 'capability economy.' The talk is structured in three acts, detailing the journey from the initial launch of AWS Lambda, through the maturation of the serverless ecosystem with frameworks and new platforms, to the present day where complex capabilities like AI models are consumed as millisecond-metered APIs. This shift demands a new developer mindset focused on composing services, event-driven architecture, and eliminating infrastructure management.
This talk traces the evolution of cloud computing over the last decade, shifting from a model of renting machines by the hour to renting granular capabilities by the millisecond. The journey highlights a fundamental change in how we build and pay for software, moving from managing infrastructure to composing services.
The Past: Developer Reality a Decade Ago
A decade ago, development was dominated by manual infrastructure management. Key challenges included:
- EC2 Provisioning: A complex and often error-prone process of selecting instance sizes, allocating disk space, and configuring networking, which rarely worked on the first try.
- Hourly Billing: Costs were tied to the uptime of infrastructure (EC2, databases), regardless of actual usage. As the speaker noted, billing didn’t switch to per-second until around 2017.
- Opaque Dependencies: Only the person who configured a virtual machine truly knew what dependencies and services were running on it.
An anecdote about a server with an uptime of over seven years (2,673 days) illustrates this past reality. While impressive, its load average revealed it was barely used, yet the company continued to pay for it, exemplifying the “pay for infrastructure, not for value” model.
Three Acts of Serverless Evolution
The talk divides the last ten years into three distinct acts:
Act 1: The Dawn of Serverless (2015-2017) This period saw the birth of serverless with the general availability of AWS Lambda in 2015. Initially, its purpose was very focused, exemplified by the canonical use case of triggering a function from an S3 upload to create image thumbnails. The early ecosystem was built on a few core primitives:
- AWS Lambda
- S3, SQS, SNS
- API Gateway
These components allowed developers to build event-driven workflows, asynchronous backends, and simple APIs without managing servers.
Act 2: The Great Unbundling (2018-2021) The ecosystem matured rapidly during this phase. The focus shifted from isolated functions to building complete applications. Key developments included:
- Frameworks: The Serverless Framework played a pivotal role in making serverless accessible. More advanced, developer-friendly frameworks like SST later emerged.
- New Players: Companies like Vercel and Netlify rose to prominence, building their platforms around a serverless-first philosophy.
- APIs as the Norm: The concept of “everything as an API” took hold, making it easier to adopt serverless patterns for distributed systems. The value proposition solidified around paying for value, not idle time.
Act 3: The Capability Economy (2022-2025) This current era is defined by the abstraction of not just infrastructure, but complex capabilities into API calls. It’s now possible for startups to operate without any dedicated infrastructure, relying instead on a composition of managed services.
- Wider Compute Options: Beyond traditional Lambda, developers now have Lambda@Edge, Cloudflare Workers, and Vercel Functions, running code closer to the user.
- AI/ML as APIs: The most significant shift is the rise of LLMs, vector databases, and AI agents, all consumed as API endpoints. Instead of writing complex code, developers pass natural language instructions and data to a model.
- Infrastructure-Light Businesses: This model allows companies to build and scale rapidly by consuming services on demand, from compute to AI processing, paying per token or per query.
The New Developer Mindset
This technological evolution necessitates a new developer mindset:
- APIs Over Infrastructure: Think about the capabilities you need, not the servers you have to run.
- Event-Driven by Default: Model systems around events, as everything can be treated as one.
- Composition Over Custom-Builds: Assemble solutions by combining focused, best-of-breed services rather than building a single monolithic application.
- Pay for Value: The focus is on aligning cost directly with the value generated for the user, not with idle infrastructure.
Ultimately, the developer’s role transforms from an infrastructure manager to a composer of capabilities. As the speaker concludes, “A decade ago, best engineers managed infrastructure. Now they eliminate it.”