Springboard Series: Competition is booming up and down the AI stack—how AI is disrupting tech ecosystems
On Monday, we offered an overall landscape assessment of the AI “stack,” starting at the application layer. Today, we will dive into the choice, innovation, and competition inherent in the data, cloud, and hardware layers of the AI stack.
A proliferation of open data is fostering new competitors
Contrary to claims that proprietary data will cement the advantage of tech firms, the reality is that large amounts of data are available to almost anyone. The Georgetown Security Studies Review writes: “When machine learning was first developed, there were very few places researchers could reliably access sufficiently large enough training data to build high-performance systems. Today that is no longer the case due to the proliferation of open-source datasets.”
For example:
— This past February, Cohere for AI, the non-profit arm of the AI startup, released Aya—an open-source LLM that is fluent in 101 languages—and the corresponding open-source dataset.
— Also last month, Google released the Gemma family of open-source models. They can be run on a laptop and hosted by rival cloud platforms, lowering barriers to entry and spurring innovation.
— In August 2023, the Allen Institute for AI released the largest open-source dataset to date, Dolma.
— Bloomberg is developing a variety of AI training tools, all of which are available to the public on github.
— Over 97,000 open-source datasets are available on the AI repository Hugging Face.
Cloud computing companies are leveling the playing field for AI startups
The infrastructure that cloud providers offer stimulates innovation and competition. To hold the massive amounts of data that AI models use, developers need an extensive cloud infrastructure.
This is an extremely attractive proposition for AI developers, especially startups: it is unfeasible for many of them to build their own AI-powering cloud, but without large-scale cloud platforms, it would be impossible to develop AI applications and models. AI startups that use either Amazon Web Services or Google Cloud include Anthropic, Cohere, Character.ai, and Theia Insights. Cloud services are therefore leveling the playing field by giving AI startups the infrastructure that only well-established companies would have access to otherwise.
The competitive trends in cloud computing go both ways: cloud is supporting the many startups that are making AI a hotbed for competition, and AI is bringing new competition to the products that cloud providers offer.
AI chip companies are facing more and more competition
Companies are pouring millions of dollars into innovations to disrupt chip developers Nvidia and AMD. Chips are vital to the success of AI because they determine key features, such as how much data an AI model can store, how well it processes queries, and how fast it works. Currently, Nvidia and AMD are the main suppliers of AI processing chips, but competition is growing. Microsoft, Amazon, and Google are investing in in-house chip production, while chip startups took in roughly $8 billion in venture funding in both 2021 and 2022, The Wall Street Journal reports.
The steady startup funding is already stimulating innovation and expanding what AI chips can do. Kavout writes this about the startup Groq: “The AI chip market, traditionally dominated by NVIDIA, AMD, and Intel, is witnessing a seismic shift with Groq’s emergence. The company’s ability to deliver real-time inference, which is crucial for instant responses from generative AI products, has set a new benchmark for the industry. The Groq LPU’s role in enhancing the speed of AI-driven platforms like chatbots and consumer electronics is pivotal, as evidenced by its implementation by customer and partner aiXplain. The significance of Groq’s achievements cannot be overstated. The company’s technology is not only demonstrating superior performance in benchmarks but is also rapidly gaining traction in the market. With the opening of API access to its real-time inference capabilities, Groq is enabling a new wave of fluid end-user experiences across various applications.”