During Trump’s 1st term, Trump and Masayoshi Son announced Softbank will invest $50 B and create 50K jobs. He mostly invested in existing companies, not new enterprises. His biggest investment was “We Work”.
Like clockwork, he showed up in Mar-a-lago and announced $500 B investment and 100 K jobs, along with Open AI Sam Altman ( Of course old lion Larry Ellison was also there).
The “world domination through unlimited investment” parallel is too glaring to miss. Open AI’s current spending trajectory vs earnings, requires sustained fund raising for few (many) years. Will OpenAI continue to get investor capital? or will Open AI dramatically alter their course to get to break even or minimal cash burn? Will Sam Altman curb his ambition to raise $7 Trillion yeap $7000 Billion?
Interesting questions. When looking at $NVDA how many such great ambitions are baked in their numbers?
xAI and Anthropic better hope not. MSFT and Alphabet throw off close to $200 billion in annual profits each, so they can fund a game of that magnitude - to say nothing of the compute they can contribute as “sweat equity” in other firms. Meta’s not quite in that league, but still $50-$60B in profits. So those mega companies can fund a “game” that requires annual ten-digit capex, whether in-house or in partnership (like MSFT and OpenAI).
If AI really requires those kinds of massive resources, that’s going to be harder for smaller companies like xAI and Anthropic to stay in the hunt. Deepseek suggests that it might not actually be a $ hundreds of billions game, though…
Anthropic is leveraging Amazon. Amazon plans on $100B investment.
xAI is a new entrant and already valued at $50B. They stood up the Cortex supercomputer in record time.
First of all not all goes to AWS. Even within AWS, application development, data center, trainium chip development all will take bulk of the AWS investment. For now, AWS has invested some money in Anthropic. Your statement mis-represents that Amazon is going to spend $100 B on AI LLM. Absolutely not.
AWS offers multiple models on Bedrock, including Claude/Anthropic and hundreds of thousands of customers use them. They all use shared pools of GPUs and other AWS resources.
Dario Amodei, is the CEO of Anthropic, in this he simultaneously argues, DeepSeek is not a big deal, and we need to tighten export control of chips. He is worried.
