Nvidia’s CEO Jensen Huang delivered a keystone speech ahead of Computex 2024 in Taipei, raising questions about whether new entrants can stake a claim to the artificial intelligence bonanza. Nvidia, known for making processors essential for training generative AI’s large language models, has become Big Tech’s newest member, leading to a surge in stock market valuations not only for Nvidia but also for other tech companies like Oracle, Broadcom, and HP.
However, with the dominance of existing model makers like Microsoft-backed OpenAI, Google, and Anthropic in the field of generative AI, startups seeking attention from venture capitalists are being challenged to innovate. Mike Myer, founder and CEO of tech firm Quiq, expressed skepticism about the opportunity to start a foundational AI company at this time, indicating the difficulty of competing with established players head-on.
Venture capital veteran Vinod Khosla highlighted the distinction between applications that add value to existing AI models and those that may become obsolete as AI capabilities progress, cautioning against building applications that merely mimic the powers of big models. Khosla cited chip design as a ripe field for innovation in AI, pointing out the demand for specialized processors that offer specific powers to meet the evolving demands of AI technology.
Rebecca Parsons, CTO at tech consultancy Thoughtworks, emphasized the need for more specialized processing capabilities to cater to the diverse requirements of AI. Groq, a startup specializing in building chips for AI deployment, has emerged as a strong competitor in this space, focusing on providing specialized processing for AI inference, in contrast to Nvidia’s dominance in AI training. Groq CEO Jonathan Ross compared Nvidia to Michael Jordan in basketball, acknowledging their strengths in generative AI training but highlighting the opportunities in other aspects of AI.
A key opportunity identified by Khosla is the potential for highly specialized AI solutions that offer expertise and know-how based on proprietary data, which cannot be easily replicated by tech giants. Companies like Cohere are developing specifically made models for businesses that are hesitant about adopting AI technology. Cohere CEO Aidan Gomez emphasized the importance of winning enterprises’ trust by proving the reliability, trustworthiness, and security of AI solutions, especially for industries that are risk-averse to new technologies.
Gomez, a co-author of the influential paper “Attention Is All You Need” that introduced Transformer, the architecture behind popular large language models, founded Cohere to provide tailored AI solutions for businesses. With funding from Nvidia and Salesforce Ventures, Cohere is positioned as a leader in providing highly specialized AI models and is valued in the billions of dollars. The evolving landscape of AI technology presents unique opportunities for startups to carve out a niche in specialized AI applications that offer value to businesses and remain relevant as AI capabilities continue to advance.
