Skip to content

Anthropic

TSMC buys $500M of Anthropic at the IPO price

2 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: reuters.com

TSMC, the world’s largest contract chipmaker, now owns a small piece of Anthropic. On November 19, the day Anthropic’s shares started trading on the Nasdaq, TSMC bought 6,024,096 Anthropic shares at $83 each, the IPO price, for $499,999,968. That comes to 0.03% of the company, according to TSMC’s filing with the Taiwan stock exchange.

The filing is plain. It calls the deal an equity investment, says it was paid in one lump sum under Rule 144 of the US Securities Act of 1933, and lists no relationship between the two companies. TSMC’s board approved it on November 10, and no director objected. The shares were bought through a TSMC subsidiary, TSMC Partners.

Part of a bigger block

TSMC didn’t buy in the IPO itself. Its shares are Class N common stock, a separate class from the Class A shares sold to the public. In an amended filing ahead of the listing, Anthropic said seven companies, AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC, had agreed to buy $26.0 billion of Class N stock at the IPO price, Leverageshares reports. Next to that, TSMC’s stake is small. Nvidia alone agreed to buy roughly $10 billion worth, according to CNBC.

The price was the same one public investors paid. Anthropic set it at $83, above the $70-78 range it had marketed, and raised about $93.5 billion in the IPO. By the end of its first trading day, Anthropic was worth roughly $1.8 trillion on a fully diluted basis, CNBC reports.

OpenAI too

The same board meeting on November 10 also approved up to $500 million for OpenAI preferred shares. Both amounts were set as ceilings. With the Anthropic half now bought almost exactly at the limit, TSMC has approved bets on both the maker of Claude and the maker of ChatGPT.

OpenAIOpenAIanalysis

OpenAI wants a shared limit on AI progress

GPT-3.5 scored 0.99, GPT-4 0.91, newer models under 0.8 on OpenAI’s internal next-word test, where lower is better. OpenAI made them public to push labs toward shared limits. Would you want AI labs to agree on a limit they can’t cross yet?

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.