Anthropic closed a $30 billion Series G yesterday at a $380 billion valuation. The money won’t be enough.
The raise looks massive until you see what it has to fund: $14 billion in annualized revenue demands infrastructure that doesn’t exist yet, and the $50 billion data center commitment announced last year requires power grid upgrades Anthropic must pay for itself. This isn’t a victory lap. It’s a down payment on an energy bill that could reshape how much AI costs for everyone.
Here’s what everyone missed: while the tech press obsessed over ChatGPT’s consumer dominance, Anthropic quietly won the enterprise war.
Anthropic owns 40% of enterprise AI spending while OpenAI bleeds share
Anthropic now controls 40% of enterprise LLM spending, up from 24% months ago, according to Menlo Ventures. OpenAI dropped to 27% from 50% over the same period. That’s not a market shift—it’s a rout.
Claude Code, which launched as Claude Cowork before rebranding, hit $2.5 billion in annualized revenue—more than doubling since January 2026, per SiliconANGLE. Over half comes from enterprises. More than 500 customers now pay at least $1 million annually. The company’s overall revenue run-rate grew over 10x yearly for three consecutive years.
And the Super Bowl ad war with OpenAI wasn’t just marketing theater. It signaled a fundamental brand divide that enterprise buyers noticed: Claude means no ads, no data scraping, no compromises. That message resonated—25.5% positive sentiment for Anthropic’s ads versus OpenAI’s 16.3%, per Meltwater analysis.
The narrative was wrong. ChatGPT won consumers. Claude won the companies writing million-dollar checks.
The $50 billion infrastructure bet has a power problem nobody’s solving
Anthropic committed $50 billion to data centers in Texas and New York last year, with sites coming online throughout 2026. But here’s the math that doesn’t work: those facilities need power infrastructure that takes years to build, and utilities aren’t paying for it.
Grid connections for 100+ megawatt loads take up to seven years in Northern Virginia, the densest AI data center market in the US. Gartner predicts 40% of AI data centers will face power constraints by 2027. As Nvidia’s CEO warned in January about AI infrastructure limits, the power problem isn’t theoretical—it’s already delaying projects across the industry.
Anthropic’s $30 billion raise covers model development and operational scale. It doesn’t cover transformers, substations, and transmission lines. Those costs land on whoever wants the power first—and that’s Anthropic.
The company’s partnership with infrastructure provider Fluidstack helps, but physics doesn’t negotiate. You can’t run a gigawatt-scale data center on promises.
What this means for everyone using Claude
Energy costs get passed through. Always.
Enterprise customers in healthcare and other regulated industries paying $1 million or more annually might absorb infrastructure cost increases through contract renegotiations. Smaller users won’t have that buffer. Anthropic hasn’t disclosed pricing changes between January and February 2026, but the direction is clear: $50 billion in infrastructure plus grid upgrades that take years equals higher costs somewhere in the stack.
The “no ads, no compromises” promise looks great until you realize the alternative is pricing out half your potential market. And if power constraints force capacity limits, enterprises with eight-figure contracts get priority. Everyone else waits.
This isn’t unique to Anthropic—every AI company scaling past $10 billion in revenue hits the same wall. But Anthropic got there first, which means it’s solving the power problem alone while competitors learn from its mistakes.
The question isn’t whether Anthropic can afford to build the infrastructure. It’s whether the infrastructure can be built fast enough—and what happens to the “helpful, harmless, honest” brand when the power bill comes due and someone has to pay it.
