As Taiwan powers the world's AI ambitions, its software startups face an uphill battle for survival.
In the bustling tech landscape of Taiwan, a dichotomy is shaping the future of artificial intelligence. On one hand, Taiwan stands as an indispensable cog in the global AI machinery, providing the critical computing power needed for innovations across the globe. On the other hand, its own AI software industry grapples with systemic challenges, struggling to find its footing amidst a landscape skewed heavily in favor of hardware giants.
Sega Cheng, CEO of iKala, a prominent AI software company in Taiwan, provides a candid overview of this paradox. "Taiwan has always been integral to the tech supply chain, especially in hardware manufacturing," he says.
While Taiwan has cemented its place as a critical player in AI hardware manufacturing, Cheng is less optimistic about the island's prospects in AI software development. "Taiwan doesn't really have an AI industry," he states bluntly. "We are part of the AI industry chain, primarily offering the computing power for AI applications worldwide.
Credit: DIGITIMES
Taiwan's AI Dilemma: Hardware Strength, Software Struggle
This hardware-centric approach has led to a "sectoral imbalance" in Taiwan, with resources heavily concentrated in semiconductor manufacturing. This has resulted in a stock market boom for chip manufacturing and design firms and foundries. On July 4th, TSMC's stock price surpassed NT$1,000 per share for the first time, pushing its market value to NT$26 trillion(US$799 billion). For context, Taiwan's GDP in 2023 is NT$23.5 trillion(US$755 billion). TSMC's market value surpassing Taiwan's annual GDP is an unprecedented situation globally that carries systemic risks.
For AI software companies in Taiwan, the landscape is challenging. "The scenario is quite different," says Cheng. "Resources are not concentrated here unless you're in electronic manufacturing."
The government's robust support structure for hardware companies is evident, with substantial resources dedicated to infrastructure, human resources, and subsidies.
This disparity is not just financial. The intense competition for talent is another hurdle. As NVIDIA establish R&D centers in Taiwan, local companies like TSMC and MediaTek face the risk of talent poaching. "NVIDIA coming to Taiwan to recruit a thousand engineers makes TSMC and MediaTek very nervous," Cheng explains.
Credit: DIGITIMES
The Future of Language Models and AI Dominance
When it comes to large language models (LLMs), Cheng is skeptical of efforts to create a Taiwan-specific model. "Honestly, we don't need one," he states. "It's not about looking down on Taiwan, but about practical usage." The dominance of English-language models accelerates existing linguistic hierarchies. "Language hegemony will definitely occur. Everyone should prioritize improving their English to gain access to the most powerful AI tools."
As AI continues to reshape industries and redefine technological boundaries, companies like iKala are positioning themselves at the intersection of emerging trends. With a focus on vertical search and enterprise knowledge management powered by AI, Cheng sees opportunities even as the tech landscape becomes increasingly dominated by a handful of global giants.
The Global Competition
He cautions that many Taiwanese "AI software companies" are essentially system integrators, creating solutions for specific manufacturing processes that don't easily scale globally. Taiwanese software startups should aim for international markets from day one. Internationalization has been key to iKala's strategy. By expanding to markets like Japan, Korea, and Southeast Asia, iKala ensures its growth is not confined by Taiwan's limited support for AI software companies.
For global AI software companies, competing with tech giants like Google, Meta, and OpenAI remains a daunting challenge. These major players continue to dominate the generative AI market due to their economies of scale and vast computing resources.
Despite these challenges, iKala has managed to carve out a niche by focusing on B2B solutions, providing AI-driven tools that enhance business operations. Their approach is pragmatic—offering solutions that improve efficiency and reduce costs rather than just generating new revenue streams. "For example, we use generative AI for knowledge management and marketing technology, offering a comprehensive solution that meets the current demands of businesses," Cheng elaborates.
Credit: DIGITIMES
In a recent meeting with Taiwan's Vice President Hsiao Bi-khim, Cheng emphasized the critical need for enhanced computing power infrastructure. "We need long-term and substantial investment in computing power to support AI development. Currently, the existing centers are not sufficient to meet the growing demand."
Taiwan's dual reality presents both an opportunity and a challenge. While it continues to lead the AI hardware revolution, the path for its AI software industry remains uncertain with obstacles. The question remains: can Taiwan balance this dichotomy to emerge as a holistic AI powerhouse? Only time will tell.
