Why thoroughbreds (not unicorns) are winning the AI race (startupsmagazine.co.uk)

The piece argues that in today’s AI boom, founders and investors should stop idolizing unicorn valuations and instead build “thoroughbreds” — companies that demonstrate repeatable revenue and durable product-market fit. With a rising number of high-value but weakly grounded “zombiecorns” and signs of an “AI chill,” valuation alone is a poor signal of long-term viability. The healthier milestone is becoming a “colt” (~$25M revenue) on the way to a thoroughbred ($100M+ ARR), because revenue validates that customers will pay and that the business can scale sustainably. For builders and the AI/ML community the implication is clear: prioritize technical defensibility over flashy wrappers. That means investing in robust data pipelines, vertical-specific or domain-tuned models, automation frameworks, and mission-critical integrations that are hard to replicate — not just surfacing a foundation model through a new UI. Examples cited include UiPath, DeepL, Hugging Face, Covariant and Synthesia, which scaled by solving entrenched industry problems (enterprise workflows, translation, model infrastructure, robotics in warehouses, and sticky content workflows). McKinsey’s estimate of a $4.4T annual uplift highlights the opportunity in manufacturing, healthcare and logistics — sectors where deep technical work and reliable economics win over hype.