Why Most AI Doom Scenarios for Devs Are Wrong

There’s a critical factor missing from conversations about AI and developer automation. The answer lies in physics, not algorithms…

AI is redefining roles and creating profound economic ripples. But what determines whether these changes lead to new opportunities—or disruptions—for software developers?

The answer lies in the race between automation and capital accumulation. Korinek and Suh explore this answer in their macroeconomic paper Scenarios for the Transition to AGI.

Korienk and Suh provide a robust framework for understanding the economic dynamics of Artificial General Intelligence (AGI). Their analysis reveals that as AGI automates more tasks, the pace of capital investment–in infrastructure, training, and new markets–will determine whether the benefits of AGI are broadly shared or concentrated among a select few.

In this article, I’m building on their work to explore what this means specifically for software developers.

Their research provides the foundation, and I’m applying it to the realities of our industry: How will the race between automation and capital affect the demand for developers? What will it do to salaries? What needs to happen for developers to be in a good spot? What are the imminent threats that can displace many development tasks?

And, most importantly, how can you position yourself to thrive in this changing landscape?

Let’s dig in.

The Automation vs. Capital Tug-of-War

In their paper, Korinek and Suh conceptualize human labor as a collection of tasks, each with varying levels of computational complexity. As AI advances, tasks with higher complexity become automatable, increasing productivity but displacing human labor.

The critical question they raise is whether capital accumulation—investments in computing infrastructure, workforce retraining, and organizational changes—can keep up with the pace of automation.

When applied to software development, this framework raises critical questions:

• Will automation replace developers or enhance their productivity?
• Can training, tools, and infrastructure investments outpace AI’s exponential growth?
• How will this affect developer salaries and job security?

The Three Scenarios Impacting Developers

The answers depend on how this race plays out, and Korinek and Suh’s scenarios provide a roadmap for understanding the possibilities.

Scenario 1: Capital Keeps Pace with Automation

In this optimistic scenario, capital investments grow alongside automation, creating new developer roles and opportunities. Automation doesn’t displace developers—it complements their work, enabling them to focus on higher-value tasks.

What This Means for Developers:

• New Roles Emerge: Developers become AI supervisors, orchestrators, and ethics specialists, leveraging AI tools to tackle more complex challenges.
• Sustained Demand: Companies need developers at all levels, from juniors who oversee automated pipelines to seniors who design hybrid human-AI workflows.
• Specialized Knowledge Pays: Expertise in AI tools, machine learning, and cloud orchestration becomes a key differentiator.

Example: Cloud Computing’s Rise

Remember when AWS revolutionized infrastructure management? Automation reduced manual server provisioning but didn’t eliminate ops jobs. Instead, it created roles like DevOps engineers and cloud architects, leading to booming demand for these skills. Companies that invested in capital—cloud infrastructure, developer training, and organizational transformation—reaped the rewards.

Salary Impacts:

Salaries grow steadily, reflecting the increasing value of developers’ skills in an automation-augmented environment.

Scenario 2: Automation Outpaces Capital

This is the more unsettling possibility. If automation progresses faster than capital accumulation, companies might automate tasks before they’ve invested in the infrastructure or training needed to fully leverage these systems.

What This Means for Developers:

• Routine Tasks Disappear: AI tools like GitHub Copilot, which already automate code writing. With AGI, even debugging, testing, and deployment could become fully automated.
• Mid-Level Roles Shrink: Developers performing repetitive or modular tasks may find themselves displaced.
• Skill Polarization: Demand concentrates on two extremes: elite developers with niche AI expertise and entry-level roles in supervised capacities.

Example: AI-Generated Code

Imagine an AI capable of writing production-ready applications end-to-end. Many roles could disappear without the necessary capital investments—such as tools for supervising these systems or retraining developers. Companies would struggle to adapt, and displaced developers might face stagnant wages or job insecurity.

Salary Impacts:

Wages for most developers stagnate or decline. Only a small group of “superstar” developers specializing in AI system design and oversight sees salary spikes. This aligns with Korinek and Suh’s finding that wages collapse as automation fully matures, especially in industries that fail to adapt quickly.

Scenario 3: Capital Outpaces Automation

This is the silver lining scenario. Here, capital investments outpace automation, creating a demand for developers to bridge the gap between AI’s capabilities and the infrastructure needed to deploy it effectively.

What This Means for Developers:

• Expanding Opportunities: Developers must design, deploy, and maintain systems integrating automation into workflows.
• Focus on Creativity: Repetitive tasks are automated, freeing developers to focus on higher-order problem-solving and innovation.
• Upskilling Becomes Key: Developers with expertise in AI orchestration, ethical AI, or interdisciplinary skills thrive.

Example: AI-Augmented Productivity

Consider an AI tool that automates unit testing. Companies that invest in retraining developers to oversee AI-driven testing systems or conduct exploratory testing will see productivity gains without losing jobs. Developers transition to more creative and strategic roles.

Salary Impacts:

Salaries are significantly driven by high demand for skilled developers who can augment automation and solve complex problems.

Why Automation Likely Won’t Outpace Capital

One critical insight from Korinek and Suh’s research is that physical realities constrain automation—an idea I strongly agree with. Automation may advance rapidly, but capital accumulation requires time and resources. Here’s why:

1. Hardware Bottlenecks:

• Manufacturing AI chips like NVIDIA’s H100 GPUs involves up to 14-20 weeks of precision fabrication, not to mention years of planning for new facilities.
• Scaling AI data centers demands billions of dollars in investment and substantial energy capacity.
  1. Power availability issues extend construction timelines by 2-4 years, sometimes up to 6 years.
  2. Backup diesel generators can take up to 90 weeks to procure.
  3. Under ideal conditions, Microsoft’s rumored 5GW data center will take four years to complete.

2. Energy Constraints:

• AI workloads are power-hungry. Scaling them requires significant energy investments, creating a natural buffer against runaway automation.
  1. Ten years ago, a 30-megawatt (MW) data center was considered significant. Today, 200 MW facilities are becoming standard for AI workloads.
  2. Average power densities in data centers have more than doubled in just two years, from 8 kW per rack to 17 kW, and are expected to rise to 30 kW by 2027.
  3. Training models like ChatGPT can consume more than 80 kW per rack.
  4. Nvidia’s latest chip, the GB200, and its servers may require rack densities of up to 120 kW6.
  5. Due to infrastructure limitations, some utilities can only offer power in 15-25 MW tranches for new 100 MW data center campuses6.

These physical and logistical constraints make it unlikely for automation to outpace capital accumulation consistently.

Wildcards: Where Automation Could Surge Ahead

That said, there are scenarios where automation could leap ahead:

1. Dramatic Cost Reductions in AI Tokens: If running AI models becomes significantly cheaper, automation adoption could accelerate faster than industries can adapt.
2. Emerging AI Architectures: Innovations like the Composition of Experts (CoEs) could enable AI to achieve far more with existing infrastructure, compressing the timeline for automation.

These factors could disrupt even well-prepared industries, underscoring the need for vigilance.

What This Means for Demand for Developers & Their Salaries

The race between automation and capital will shape not just the demand for developers but also their earnings:

• Scenario 1 (Capital Keeps Pace): Salaries grow steadily, particularly for AI-literate developers.
• Scenario 2 (Automation Outpaces Capital): Salaries decline for mid-level roles, while niche experts see sharp increases.
• Scenario 3 (Capital Outpaces Automation): Salaries rise significantly across the board as developers remain central to leveraging automation’s potential.

How Developers Can Stay Ahead

The rapid evolution of AI is creating new challenges and opportunities for developers. Adopting a proactive and strategic approach to skill development is essential to navigating these shifts.

One way to think about these challenges is through the Game-Changers Risk Matrix, a tool that illustrates how developers should prioritize their efforts based on the impact and timing of key AI advancements.

The Game-Changers Risk Matrix

The matrix highlights two critical dimensions for developers: the potential impact of emerging AI technologies and their expected timeframes. Each quadrant of the matrix points to specific risks and opportunities developers should prepare for:

1. High-Impact, Short-Term: Token Cost Reduction

• Token costs for AI models are dropping fast as companies like Anthropic and OpenAI optimize their infrastructure. This rapid decline could accelerate the adoption of AI coding assistants, potentially disrupting traditional development workflows sooner than expected.
• Immediate Actions for Developers:
  • Start experimenting with AI coding assistants to understand their capabilities and limitations.
  • Develop prompt engineering skills to utilize AI tools more effectively.
  • Build expertise in optimizing AI tool usage to manage token costs efficiently.

2. High-Impact, Long-Term: Composition of Experts (CoE) Models

• Emerging AI architectures like CoE models represent a paradigm shift. Instead of relying on monolithic models, these systems use specialized expert models working together. This approach could dramatically enhance AI’s ability to tackle complex programming tasks using existing computing resources more efficiently.
• Long-Term Preparation for Developers:
  • A study of distributed systems and microservices architecture to prepare for CoE-based development.
  • Develop skills in AI system integration and orchestration.
  • Gain expertise in model fine-tuning and adaptation for specific tasks.

3. Low-Impact, Short-Term: Infrastructure Bottlenecks

• The massive computational requirements for advanced AI models act as a natural buffer, slowing the pace of full automation. This creates a strategic window for developers to adapt and evolve their skills.
• Infrastructure Focus for Developers:
  • Learn cloud infrastructure management and optimization.
  • Develop skills in AI-specific infrastructure, such as GPU clusters and specialized hardware.
  • Understand machine learning (ML) deployment and scaling strategies.

4. Low-Impact, Long-Term: Regulation Changes

• While regulatory changes might seem distant, they can significantly impact AI development and deployment. Staying ahead of governance trends ensures developers are ready to navigate compliance challenges when they arise.
• Risk Mitigation for Developers:
  • Stay informed about AI governance and regulation trends.
  • Develop expertise in AI safety, responsible AI development, and auditing.
  • Build skills in compliance and ethical AI standards.

Strategic Takeaways

By understanding the game-changers outlined in the matrix, developers can make informed decisions about where to focus their time and resources. The key is balancing immediate actions—like adapting to token cost reductions—with long-term preparation for transformative shifts, such as CoE-based architectures.

A Balanced Approach

To thrive in this rapidly changing landscape, here’s how you can align your efforts with the insights from the Game-Changers Risk Matrix:

• Short-term Focus: To stay ahead, embrace AI tools now, experiment with their capabilities, and master prompt engineering.
• Long-Term Preparation: Strengthen your knowledge of distributed systems and AI orchestration to anticipate CoE models and other architectural innovations.
• Infrastructure Mastery: Use the current slowdown caused by hardware bottlenecks as an opportunity to build expertise in cloud and AI-specific infrastructure.
• Stay Informed: Keep up with regulatory changes, ethical considerations, and governance practices to remain relevant as AI evolves.

The Big Picture

The insights from Korinek and Suh’s paper highlight a critical transition for software developers. The race between automation and capital will determine our industry’s future and define our roles, opportunities, and earning potential.

What are your thoughts on this? How do you see the race between automation and capital accumulation play out in your organization?