KUALA LUMPUR (Feb 17): The advancements in large language models (LLMs) and DeepSeek technology had allowed startups to build artificial intelligence (AI) products at lower costs, said an AliBaba Cloud Intelligence officer.

This is because startups can now sidestep the heavy lifting traditionally required to develop the infrastructure behind AI systems by integrating AI models into cloud platforms and just focus entirely on innovation.

"These new startups only need to focus on their ideas — the rest, like the cloud and AI models, are already available. It's like having access to electricity or water — ready to use, with no need to reinvent the wheel,” said Leo Liu Binxing, vice president of international business general manager, North Asia Pacific and Southeast Asia at Alibaba Cloud Intelligence.

Speaking at the China Conference Southeast Asia 2025 here on Monday, Liu said that the lower barriers to entry have turned AI from an exclusive tool for large enterprises to a democratised resource, available to any innovator with a strong idea.

Meanwhile, an investment partner at OSK Ventures International Bhd, Sarah Lim, said that some startups have yet to fully explore the potential of LLMs and other cost-effective AI solutions.

However, even without fully delving into LLMs, Sarah pointed out that startups are already using them to create modules, enabling sophisticated functionality without needing to develop from scratch.

This transformation is a game-changer for smaller players in the AI space, she added. This means that more companies can now explore AI for applications that were previously out of reach, such as AI-driven avatars or virtual agents, she said.

One company, Sarah added, uses two different LLMs to create lifelike avatars, ensuring each function is optimised for realism — proof of how specialised models can combine to create powerful solutions.

While the cost benefits are clear, for these startups, the challenge lies in scaling up, as many are still in the early stages of understanding the full potential of AI, said Jackie Cheong, head of enterprise data governance at CelcomDigi Bhd (KL:CDG).

"Startups need time to digest these options and integrate them into their models. The cost reduction is massive, but it will take time for these businesses to fully leverage the possibilities," said Cheong.

He added that the foundation for success in AI lies in robust data governance.

"Data governance is really the key. Without a solid data foundation, no matter how cost-effective the technology becomes, the AI output will not meet expectations," he said.

 "Startups may invest in DeepSeek technologies and lower-cost AI tools, but without a clear understanding of their data foundations, the results will fall short,” he added.