What does Anthropic stand to gain and risk with its IPO | Explained

Anthropic is reportedly planning a $2 trillion initial public offering by September, according to a Financial Times report.

Since the AI product Claude already rakes in billions of dollars, this would likely be a blockbuster IPO were it to be held today.

Anthropic’s product is incredibly popular, with software engineers handing off increasingly large amounts of advanced coding work to agents built with the AI tool.

But before Claude, there was ChatGPT and Codex, which continues to compete neck and neck with Anthropic’s tools.

For coding-focused tasks and processes where AI tool use is needed, firms like Cursor, recently acquired by SpaceX, could also pose a threat.

And then there is the rapidly growing open source model ecosystem, where AI tools that can run on sufficiently advanced PCs are able to manage increasingly complex tasks just fine locally.

Anthropic watermarking Claude AI content | Explained Anthropic faces significant risks in terms of the competition it faces both from other companies and countries.

Fueled by several years of research and capacity building pushes, Chinese firms are gaining ground on AI, with open source models like Alibaba’s Qwen shocking users with low hardware requirements and competitive performance.

So-called open source “harnesses,” which provide an interface for running the AI, have also emerged, like Bionic.

These harnesses mimic the approachable interfaces of ChatGPT and Claude, even though they are — like all LLMs — just next-word prediction tools.

If these tools get sufficiently advanced, the potential disruptions to subscription- and usage-based tools like Claude could be significant.

But this is no reason to write off a firm as large as Anthropic.

While size is no guarantee of success in entering the AI space — Microsoft, for instance, has had more luck in being an infrastructure platform for AI rather than an “inference” (AI query-response) provider on its own merit — there is a significant advantage for a company in having its AI tools being used widely.

The AI blogger Dwarkesh Patel said earlier this month that so-called recursive learning, where an AI tool’s real world use is fed back into its training, could make a major difference in how good it is.

“Right now, there are less than five prominent AI minds, and they are all quite similar to each other on account of being trained on roughly the same data,” Patel says.

“But if AIs are learning from experience, and that experience is different between different AIs, we could see actually different AIs come out the other end.” Anthropic | Security dilemma He adds, “When deployment becomes part of training, the returns to being ahead will accelerate.

If you have the best model, and more people use your AI for more complicated and useful work, and give it lots of feedback that it can integrate beyond the session window, then your model will become even smarter.” And since Claude’s models are not open source, those real world recursive learning gains would be significant, and give Anthropic pricing power.

Then again, that is assuming that it leverages its current dominance to improve future models at a breathless pace, and assuming that open models and competitors don’t collectively erode its ability to self-improve in the coming months.