Mexican AI Startups Take the Lead Through Specialized Models
Mexican startups are accelerating the development of proprietary AI models and specialized applications despite investing far less than the United States in AI infrastructure, highlighting a shift toward domain-specific innovation while key industries such as manufacturing continue to lag in enterprise AI adoption.
Mexico’s startup ecosystem is increasingly proving that competing in AI does not require building the world’s largest models. While the United States continues to dominate global AI infrastructure investment, Mexican startups are carving out a different path by developing specialized models tailored to industry-specific use cases, allowing them to compete through expertise rather than scale.
The strategy reflects the country’s position within the global AI value chain. In 2025, according to Visual Capitalist, the five largest US technology companies invested a combined US$448 billion in AI-related capital expenditures. By comparison, Mexico allocated MX$33 billion (US$1.89 billion) to AI-specific investments across infrastructure, software and IT services, according to IDC.
Despite the funding gap, Mexico’s startup ecosystem is showing signs of maturity. Venture capital activity continues to support AI-focused companies, while startups increasingly prioritize proprietary capabilities over dependence on generic AI models.
Specialization Becomes Mexico’s Competitive Strategy
Rather than attempting to build foundational models comparable to GPT, Gemini or Llama, Mexican startups are focusing on smaller, domain-specific AI systems trained with proprietary data and designed for particular industries.
Gabriel Charles, General Partner at OBS and Regional Director of Entrepreneurship at Tecnológico de Monterrey, says the country’s dependence on global AI infrastructure does not prevent local innovation.
“There is an important dependence on global infrastructure, but alternatives exist through hybrid architectures, where sensitive or sovereign workloads remain local while less sensitive processes can rely on cloud infrastructure and general-purpose models,” says Charles.
This approach allows startups to leverage hyperscale cloud infrastructure for general AI workloads while retaining sensitive data and industry-specific knowledge within their own environments. The model reduces infrastructure costs while enabling companies to develop differentiated AI products.
Charles says building proprietary foundational models only makes sense for initiatives backed by significant public and private investment, such as Mexico’s Coatlicue supercomputing project.
Training large language models requires hundreds of millions of dollars in computing infrastructure, extensive datasets and years of research, placing such projects beyond the reach of most startups.
Instead, investment is increasingly flowing toward applications where specialized knowledge creates a competitive advantage. According to the Latin America Venture Capital Association (LAVCA), fintech startups attracted US$865 million in funding during 2024, followed by ecommerce with US$87 million, mobility with US$86 million and IT, data and cloud services with US$68 million.
Small Language Models (SLMs) are emerging as one of the main opportunities for Mexican startups. Unlike large language models trained for general-purpose tasks, SLMs are optimized for specific business domains such as financial services, collections, hospitality or automotive applications.
Franco Palacios, Founder and CEO of Creai, says enterprises increasingly value precision over broad creativity.
“Small Language Models are hyper-specialized. Medium-sized and large companies in Latin America need certainty, not creativity. A generic large language model introduces the risk of hallucinations, while a domain-specific SLM reduces that risk dramatically. It is a market need that still is not fully articulated, but it already exists,” says Palacios.
Native AI Startups Outpace Traditional Businesses
The emphasis on proprietary AI development is already translating into stronger operational performance among Mexican AI startups.
According to the global Growth Drivers report prepared by Strand Partners for AWS Startups, 63% of Mexico’s native AI startups have already developed proprietary AI capabilities, compared with 33% of traditional startups. Nearly half generate more than US$400,000 in revenue per employee, almost double the proportion recorded among conventional startups.
Globally, AI-native startups report average annual revenue growth of 156%, compared with 65% for startups overall. The report also indicates these companies are reaching US$1 billion valuations in approximately 3.5 years, roughly half the time required before the emergence of generative AI.
The findings suggest that startups built around AI are moving beyond adopting third-party tools and are increasingly creating differentiated intellectual property as a foundation for growth.
The trend aligns with the broader evolution of Mexico’s AI ecosystem. Venture capital funding reached US$1.1 billion in 2025, while enterprise AI adoption continues to expand across industries. However, analysts note that much of the country’s AI usage remains concentrated in basic productivity applications rather than proprietary business capabilities.
Manufacturing Illustrates Mexico’s AI Adoption Gap
While startups accelerate AI development, one of Mexico’s most important economic sectors continues to adopt the technology at a much slower pace.
An analysis by Centro México Digital, based on the 2024 Economic Census conducted by INEGI, found that only 4.8% of manufacturing companies with more than 10 employees currently use AI systems capable of prediction, recommendation or automated decision-making.
The sector employs more than 7.19 million people and accounts for approximately 90% of Mexico’s exports, yet it ranks only 12th among 19 economic sectors in AI adoption.
The study also found meaningful differences by company size. Medium-sized manufacturers reported AI adoption of 7.8%, while small manufacturers reached only 1.2%. Large manufacturers use AI approximately 14 times more frequently than smaller companies.
Although the analysis identifies statistical correlations rather than direct causation, it found that every 10-percentage-point increase in AI adoption among manufacturers is associated with 18.8% higher gross production and 5.4% higher wages. The study also links broader enterprise AI adoption with a 3.3% increase in employment across the economy.
