Illustrated by Brendan Lynch/Axios
There is an old saying that the surest way to make money in the gold rush is to bet on companies that provide pickaxes. And that idea is igniting tool makers and wholesalers in today’s generative AI boom.
why it matters: While it’s relatively easy to see the broader tech trends on the horizon, it’s often much harder to figure out who wins on what timeframe. fulfill.
Tom Siebel, a software veteran who now runs enterprise AI company C3.ai, sees a broad swath of the tech industry benefiting from the AI boom.
- In a statement to Axios, Siebel said, “Enterprise AI and large language models are poised for exponential growth and widespread adoption over the next few years.
- “Products and services directly tied to the adoption of pureplay AI products will benefit from similar growth, including enterprise AI software providers such as C3 AI, LLM providers, cloud computing providers, GPU or AI This includes chip providers, system integrators, and consulting service providers.”
Here are three categories of companies that act as modern-day ax vendors for AI:
The giant of cloud computing
The first tool you need to mine AI gold is computing capacity. The models that power current-generation generative AI tools like ChatGPT and Dall-E are complex, with billions of parameters.
- As a result, systems, especially those available to the public, consume enormous amounts of computational power both during the development or training phase and during normal use.
that’s a good omen Cloud computing giants: Microsoft, Google, Amazon.
- Among them, Microsoft is at the forefront, combining its Azure cloud with its OpenAI partnership, which is powered by billions of dollars of investment.
- Google has also been offering machine learning services for some time and wants to show that it can compete with Microsoft in other areas.
- Amazon, a leader in the entire cloud computing business, has long been viewed as not a player in generative AI. The company hopes to change that with Thursday’s series of announcements, including both a partnership and the release of its own foundational model.
chip manufacturer
All this computing work means you’ll need a lot of chips to power all those AI servers. They rely on several different kinds of chips, such as CPUs from Intel and AMD, and graphics processors from companies like Nvidia.
- Many cloud providers, such as Amazon and Google, are also developing their own AI chips. Meanwhile, chip makers, including Nvidia, are also developing their own base models.
- “We also see a lot of healthy competition for cloud computing providers between Nvidia, AWS, AMD and Google. (a particular kind of logical operation at the heart of AI), said Alexandre Kostin, vice president of Generative AI at Adobe. Adobe recently announced his partnership with Nvidia as well as OpenAI.
- AI performance isn’t about pure chip speed, it’s about how well a particular chip runs a particular model. Many say that Nvidia has an edge over AMD, Intel and various AI chip startups.
- Intel, meanwhile, points to results published by Hugging Face, which it says show that Intel’s AI hardware accelerators perform inference faster than any graphics chip on the market. increase.
human element
Counter-intuitively, artificial intelligence actually requires a lot of real human labor, at least in the early stages of development, to check the work of computers, add metadata, and mitigate bias. It turned out that
- One of the big players here is ScaleAI, a startup with a $7.3 billion valuation that helps AI companies annotate data.
- The AI boom should also benefit the multitude of consultants and system integrators helping large companies develop and implement new technology strategies.
Yes, but: Many of the companies directly developing AI technology position themselves as ax vendors.
- For example, OpenAI has a mega deal with Microsoft to provide a wide range of developers with text-to-image API access and access to GPT-4 text models.
- Many others, including Stable Diffusion, aim to enable other companies to build customized consumer and business tools based on their own underlying algorithms and models.
