Who Benefits More from AI: Startups or Incumbents?

Introduction

With forecasts predicting the market for AI products and services to grow to over one trillion USD by 2030, there are going to be many opportunities to harness this economic shift. Not only does this projection underscore the technological change currently underway but also shows the potential of AI to redefine how businesses operate. 

There is no doubt that AI is here to stay, the big question now is who will be the winners of the AI race. Will it be the startups who are notorious for their ability to stay agile and innovate? Or will it be the incumbents who have the resources to build more powerful models? 

Background on AI in Business

While the Large Language Models (LLM) product, ChatGPT, initiated the current hype cycle, the entire AI field is going through a golden period of funding, innovation, and utility. Here is a quick breakdown to put the different areas of AI into perspective: 

• Machine Learning (ML):  From predictive analytics to fraud prevention, ML is powering a wide range of applications.

• Natural Language Processing (NLP): Applications include chatbots, sentiment analysis, and language translation, fundamentally transforming customer service, content analysis, and communication.
• Large Language Models (LLMs): The largest form of AI models that are capable of generating original content with a text prompt, use in applications like Chatbots, Automated workflows, and Content Generation.

• Computer Vision: From facial recognition to object detection, computer vision algorithms are driving innovations in healthcare, retail, and surveillance.

• Deep Learning: With its ability to handle large volumes of unstructured data, deep learning fuels breakthroughs in image and speech recognition, autonomous vehicles, and drug discovery.

• Reinforcement Learning: This technology underpins advancements in robotics, gaming, and recommendation systems, offering new avenues for adaptive decision-making.

Startups: The Agile Innovators

Startups are well-positioned to benefit from AI’s potential. With small teams comes quick decision making, granting the ability to challenge the large corporations slow, methodical, and calculated advancements.

Consider the cases of Midjounrey and Stability AI, whose generative image models compete in quality and price with OpenAI’s DALL-E. Both have shown us firsthand that they can be technological leaders in an industry crowded with large companies. 

Now take Google and their search engine as another example. With LLMs becoming more powerful, Google could soon launch an answer-based search product for all of its users. However, this would risk cannibalizing large parts of their blue links business that brings in billions of dollars in revenue for them each year. Thus counterintuitively, startups hold the upper hand when it comes to launching new and innovative products.

Another way startups have set themselves apart from incumbents is their flexibility to harness open-source solutions. Using frameworks like TensorFlow and PyTorch, startups can pivot between different cutting-edge models. They can take risks and quickly iterate while benefiting from collaborating with the global AI community. This democratization has proven that collaboration can be the antithesis of the red tape that makes incumbents notoriously slow to innovate.

Startup Advantages: 

• Agility
• Ability to take risk
• Open-source solutions

Image Source: Spiceworks

Incumbents: The Resourceful Titans

When it comes to AI development, the biggest hurdle is the huge costs associated with training new models. This is where large companies have the greatest advantage in comparison to startups. The cost required to hire a team of top talent engineers, and the computational costs of training a new model can add up quickly.

Most startups just don’t have the 100 million dollars that it cost OpenAI to train GPT-4. Another aspect of development that drives up the cost is access to hardware, specifically GPUs. Recently, Meta announced that it plans on spending billions to buy new Nvidia H100s GPUs. While this will undoubtedly place Meta as one of the leading AI companies in the near future, it makes it even harder for startups to keep up.

Beyond their access to capital and market power, incumbents can attract top-tier talent globally. With opportunities ranging from entry-level positions for recent college graduates to senior roles suitable for individuals with 20+ years of industry experience, incumbents provide unparalleled career advancement that is hard to say no to. 

Incumbents also have a huge edge when it comes to AI due to their strong market position. In the past, incumbents have slowed down the development of startups and competitors by using their patent portfolios to create legal barriers to entry. One advantage of this is that it ensures that they can maintain their market share for longer. With their ever-growing number of AI patents, Microsoft and IBM are poised to do just this. 

As with any new technology, regulations are needed to help protect consumers and prevent anti-competitive behavior by incumbents. This time, big tech companies are the ones lobbying Congress to be the ones that help write those new regulations. If they are successful and change the rules of the game in their favor, who knows what that could mean for competition.  

Incumbents Advantages:

• Access to capital
• Market power
• Talent

Image Source: IP CLoseUp

The Verdict

As AI continues to evolve, startups will continue to pop up with innovations. Those who are the most successful will have carved out a niche by building targeted products for small to medium-sized audiences. Others will be acquired by the top AI companies who continue to pour tens of billions into the research and development of new models. Others will struggle and die.

At the end of the day, it will be the incumbents that benefit the most from this new wave of technological change as they always have. We are already seeing this with the massive growth in the market cap of large tech companies in the past few months. 

Image Source: TheKobeissiLetter

Federal regulators may still overcome their current gridlock but given the speed of change in the industry and lobbying efforts, they will likely always be two steps behind. There are a small group of states that have begun taking it into their own hands to begin passing AI legislation. We can hope that it will provide a framework for nationwide intervention in the future, and an even playing field for startups and incumbents alike.