The AI race enters its expensive era

OpenAI is reportedly considering major price cuts as competition with Anthropic for enterprise customers heats up.

For the first few years of the AI race, the focus was simple: build the smartest model. But now companies are starting to ask a different question: how much does it actually cost to use?

As businesses roll out AI across thousands of employees, token usage is exploding and AI bills are becoming harder to ignore. Some companies are already looking for cheaper models or being more selective about when they use expensive frontier systems.

That creates a new challenge for OpenAI and Anthropic.

Both companies are racing toward massive valuations and potential IPOs, but their customers are also becoming more cost-conscious. If businesses can easily switch between models, pricing becomes one of the biggest competitive weapons.

The next phase of AI may not just be about who has the smartest model. It may be about who can deliver the most intelligence for the lowest cost.

Benchmarks on clean audio don’t tell you how a model handles your call center recordings, noisy interviews, or fast-paced financial conversations.

Solaria-3 was built and evaluated on real production audio — the kind that breaks other models.

Try it on your own audio for free.

Elon Musk has revealed SpaceX’s plan for AI1, a new satellite designed to run AI chips directly in orbit.

The idea sounds like science fiction, but it’s solving a very real problem. AI companies need enormous amounts of computing power, and Earth-based data centers are increasingly running into issues around energy, land, and infrastructure.

SpaceX believes orbit could offer a different path by using constant solar energy and avoiding many of the power constraints slowing data center expansion on Earth.

Musk says each AI1 satellite could eventually deliver the computing power of one of Nvidia’s top server racks, with designs focused around solar panels, cooling systems, and laser connections instead of traditional satellite hardware.

The bigger picture is that the AI race is quickly becoming an infrastructure race.

The companies building the future of AI don’t just need better models anymore. They need more chips, more energy, and maybe even a completely new place to put them.

Anthropic CEO Dario Amodei says the biggest problem with AI right now might not be the technology itself. It’s the speed mismatch between AI development and the institutions trying to manage it.

In a new essay, Amodei argues that AI has moved from a chatbot experiment into a technology with national-level consequences much faster than governments expected. He points to systems like Claude Mythos, which showed frontier models can already reshape areas like cybersecurity.

The concern is that traditional policy moves in years, while AI capabilities are now changing in months.

Amodei says governments need to move beyond simply watching AI companies and start creating stronger systems around model testing, safety, cybersecurity, and economic disruption before the next wave arrives.

The bigger picture is that even the people building frontier AI are starting to talk about it less like another software boom and more like a new industrial revolution.

The question is no longer whether AI will become powerful enough to change society. It’s whether society can adapt quickly enough when it does.