From Figma to Scale AI: Lessons in playing the long game

Dylan Field and Evan Wallace probably never expected their 2012 dorm room project would become Figma, the design force that forced Adobe to pay a billion-dollar breakup fee. Four years later – living in a cramped pool house, sleeping on air mattresses – Alexandr Wang and Lucy Guo were building unglamorous data-labeling software at Scale AI. That would eventually make Wang Mark Zuckerberg's first chief AI officer.

Two companies, one playbook: the long game. The stories of Figma and Scale AI offer startups a blueprint for navigating patience, persistence, and strategic positioning in technology markets that reward endurance over flash.

Figma’s founders spent four years in development before its first public release in 2016. Field and Wallace, armed with a $100,000 Thiel Fellowship, resisted the Silicon Valley urge to ship fast and break things. They built browser-based design tools when conventional wisdom said it was impossible, betting that collaborative design would eventually matter more than feature completeness.

 Wang and Lucy Guo to a not dissimilar approach. They started with a simple data-labeling service through a summer at Y Combinator in 2016. Rather than pivoting to sexier AI applications, they doubled down on the unglamorous work of data annotation and model evaluation. The timing proved prescient as the AI boom created massive demand for high-quality training data.

 Both sets of founders, barely out of their teens – Field and Wang were 19 when they started – seemed to understand that infrastructure plays can be far deeper moats versus beat application plays over longer time horizons. Figma built the foundation for collaborative design; Scale AI built the foundation for AI model training. Neither chased immediate monetization at the expense of platform strength.

What others didn’t see

The long game requires reading market shifts before they become obvious. Figma anticipated the move from desktop to browser-based tools and from individual to collaborative design workflows. Field and Wallace saw that design was becoming more team-oriented and interdisciplinary, requiring tools that enabled real-time collaboration rather than file-passing workflows.

 Wang and Guo recognized early that data quality would become the bottleneck in AI development. While competitors focused on model architectures and computing power, Wang built human-in-the-loop systems for generating the labeled datasets that make AI models actually work. The company positioned itself as critical infrastructure rather than another AI application.

Both companies avoided the trap of solving today's problems with yesterday's assumptions. They built for where their markets were heading, not where they currently stood.

Saying no to $20 billion

Figma's rejection of Adobe's $20 billion acquisition offer in 2022 illustrates long-game thinking at its most dramatic. Regulatory concerns ultimately killed the deal, but the failed acquisition revealed Figma's true market position. The company emerged stronger, completing a successful IPO in July 2025 that valued it at $68 billion - more than three times Adobe's offer.

The Adobe episode demonstrated how playing the long game creates strategic optionality. Figma's independence allowed it to maintain relationships across the design ecosystem and continue innovating without the constraints of Adobe's corporate structure. The $1 billion breakup fee provided additional runway without dilution.

 Scale AI's Meta partnership represents a different strategic choice. Wang's decision to join Meta as chief AI officer while maintaining Scale AI's independence reflects the complexity of long-term positioning in the AI market. The $14.3 billion investment gives Scale AI resources to compete while giving Meta access to critical AI infrastructure talent.

Different metrics

 Long-game companies optimize for different metrics than their growth-stage counterparts. Figma prioritized user adoption and workflow integration over immediate revenue maximization. The company's freemium model allowed designers to experiment with collaborative workflows before committing to paid plans.

 Scale AI invested heavily in global operations and human annotation systems that competitors dismissed as non-scalable. The company built facilities across Southeast Asia and Africa, creating the human infrastructure needed to generate high-quality training data at scale. This operational complexity became a competitive moat as AI models required increasingly sophisticated data preparation.

Both companies understood that sustainable competitive advantages come from capabilities that are difficult to replicate quickly. Figma's browser-based architecture and real-time collaboration engine took years to build properly. Scale AI's global annotation network and quality control systems require operational expertise that can't be conjured up overnight.

 Timing

 The timing of liquidity events reveals long-game thinking most clearly. Figma could have sold to multiple acquirers throughout its development but waited until it had established market leadership. The company's 2025 IPO came after achieving $749 million in annual recurring revenue and dominant market share in UI/UX design.

 Scale AI's path with Meta represents a hybrid approach - taking strategic investment while maintaining operational independence. The partnership gives Scale AI access to Meta's resources and distribution while preserving the company's ability to serve other clients. Wang's move to Meta while remaining on Scale AI's board was seen to create alignment.

 More recently, there have been reports of some friction, and so the jury is out on this partnership – illustrating how the challenges never go away.

 Lessons for AI founders in India

 Indian AI startups can extract specific lessons from the Figma and Scale AI playbooks. While going global should be the long-game positioning, local context and domain expertise create sustainable advantages.

 First, focus on infrastructure. Startups like Sarvam AI are building foundational language models for Indic languages rather than competing directly with OpenAI or Google. This approach mirrors Scale AI's focus on data infrastructure rather than model development. 

In our own portfolio, Emergent is deploys agentic AI IDEs for the entire software development lifecycle, embedding infrastructure for code generation, testing, and project orchestration. Metaforms is re-architecting market research infrastructure using AI agents that automate surveys, bids, and analytics — highlighting new data processing and integration layers.

 Gibran focuses on foundational models that train on small datasets for enterprise and drug discovery, and creates lifelong learning agents, combining LLMs with nature-inspired systems for scalable inference infrastructure.

Second, build global capabilities from day one. Scale AI's international operations model provides a template for Indian startups to serve global clients while leveraging local talent advantages. India's skilled workforce and cost advantages in data annotation and model training create natural opportunities for infrastructure plays.

Third, target vertical specialization over horizontal platforms. Indian AI startups are increasingly focusing on domain-specific solutions in healthcare, finance, and manufacturing rather than general-purpose AI tools. Companies like Signzy (BFSI) and Dozee (healthcare) are building deep expertise in specific industries, creating switching costs and network effects that protect against larger competitors.

 Fourth, prioritize patient capital over rapid scaling. The Indian AI market raised $524 million in the first seven months of 2025, with enterprise software companies leading investment. Overall some 140+ AI startups have raised more than $1.5 billion since 2020.

Indian AI startups should study how Figma and Scale AI built defensible positions through operational excellence and strategic patience. The companies that will dominate India's AI future are likely those that resist the temptation of quick wins and instead build the infrastructure and capabilities needed for long-term market leadership. Even in the hyperkinetic world of AI, the biggest winners will be those willing to build for decades rather than quarters.