blinque.news
Breaking news, simply explained
Tech

OpenAI and Anthropic Show Different Paths to AI Profitability Despite High Computing Costs

OpenAI and Anthropic have shown investors different paths to profitability despite massive computing costs that eat up more than half their revenue. OpenAI projects $74 billion in operating losses by 2025, while Anthropic expects better profit margins.

April 6, 20264 sources2 min read
OpenAI and Anthropic Show Different Paths to AI Profitability Despite High Computing Costs

Two of the biggest AI companies are spending enormous amounts on computing power, but they're taking very different approaches to making money.

OpenAI expects to lose about $74 billion in 2025 - roughly three-fourths of its revenue - due to massive computing costs. The company plans to spend $9.4 billion just on training its AI models. Meanwhile, Anthropic projects much better profit margins, expecting gross margins of 50% in 2025 and 77% by 2028.

Both companies are dealing with "inference costs" - the expense of running AI models when people use them - that exceed half their revenue. Anthropic expects to spend $4.1 billion on training costs in 2025, up 5% from earlier projections.

The different approaches reflect how uncertain the AI business model remains. Companies are racing to build the best AI while figuring out how to make money from it. Computing power remains the biggest expense, requiring specialized chips and massive data centers.

Why this matters

These costs affect how fast AI tools improve and how much they'll cost consumers. The companies burning through billions today will need to make money eventually, which could mean higher prices for ChatGPT and other AI services.

What to watch

Watch for investor reactions to these different strategies and whether either company changes course on spending.

Sources
artificial-intelligenceopenaianthropicventure-capital
This story was written with AI based on reporting from the sources above. For the complete story, visit the original sources.

Was this article helpful?

0 people found this helpful