Google just committed $40 billion to Anthropic. $10 billion now, $30 billion conditional, at a $350 billion valuation. This happened on April 24, 2026. The same week, Anthropic disclosed $30 billion in annualized revenue, passing OpenAI’s $25 billion. Sixteen months ago, Anthropic’s annualized revenue was $1 billion. That is not a typo. A company grew revenue 30x in sixteen months, and the largest technology company on earth responded by writing the largest single AI investment check in history. If you are still “evaluating AI strategy,” you are not early. You are late. And the cost of waiting another six months is no longer measured in lost efficiency. It is measured in permanent exclusion from the ecosystem being built right now.
The numbers that should keep you up tonight
Walk through the revenue trajectory. January 2025: $1 billion annualized. June 2025: $4 billion. December 2025: $9 billion. April 2026: $30 billion. Each jump accelerated faster than the last. That growth did not come from consumers paying $20 a month for a chatbot. Eighty percent of Anthropic’s revenue comes from enterprise API and developer contracts. Over 1,000 enterprise customers now spend more than $1 million per year with Anthropic. That number doubled since February.
Read that again. The customer base spending seven figures annually doubled in two months.
These are not pilot programs. These are production deployments at scale. Organizations running real workloads, processing real transactions, making real decisions with AI systems embedded in their operations. And Anthropic is doing all of this with roughly 5,000 employees. That ratio of revenue per employee is almost unheard of in enterprise software. It tells you something about where the value sits in this new stack.
What $40 billion actually buys
Google is not making a charitable donation. The $30 billion conditional tranche is tied to performance milestones. Google is buying guaranteed access to the infrastructure layer that a thousand enterprises already depend on. They are buying priority in a compute buildout that includes 5 gigawatts of capacity coming online starting in 2027. For context, 5 gigawatts powers roughly 3.5 million homes.
This is physical infrastructure. Data centers. Power plants. Cooling systems. Supply chains for chips that take years to secure. The kind of capital commitment that creates moats measured in concrete and copper, not code.
When that infrastructure comes online, who gets first access? The organizations already embedded in the ecosystem. The ones with existing contracts, existing integrations, existing relationships with the engineering teams building the next generation of models. The ones whose usage patterns and feedback are shaping product roadmaps today.
Not the ones who plan to “start exploring AI” in Q3.
The ecosystem is forming without you
Here is what most executives miss about this moment. The gap is not about capability. Claude, GPT, Gemini. The models are available to everyone. You can sign up today. The gap is about practice.
The 1,000+ organizations spending $1 million or more per year with Anthropic alone have built something you cannot buy off the shelf. They have built internal competence. They have engineers who understand how to design systems around model capabilities. They have operational workflows that incorporate AI at decision points, not as an afterthought bolted onto existing processes. They have data pipelines tuned for the specific ways these models consume and produce information.
They also have something less tangible but more important: influence. When you are spending seven figures annually with an AI company, you get direct access to the product team. Your use cases shape the roadmap. Your edge cases become training priorities. Your feedback loops are tighter, your deployment support is better, your access to new capabilities comes sooner.
The organizations planning to start are planning to enter an ecosystem that was designed without them.
Six months from now is a different world
The question I hear most from business leaders right now: “What happens if we wait six more months?”
Here is the honest answer. In six months, the 1,000+ enterprise customers currently at $1 million or more will be at 2,000 or more. Their systems will be more deeply integrated. Their competitive advantages will be more entrenched. The talent they have hired, the engineers and product managers who understand how to build with these tools, will be harder to recruit because demand will have doubled while supply has barely moved.
In six months, the 5-gigawatt compute buildout will be further along, with capacity already allocated to existing customers. New entrants will face longer wait times, less favorable pricing, and less influence over how the infrastructure is configured.
Meanwhile, consider what the research already shows. Ninety-seven percent of enterprises report running AI agents in some capacity. But only 12 percent have centralized control over those deployments. Only 29 percent report seeing real ROI from their AI agent investments. The gap between “using AI” and “getting value from AI” is enormous. And it is growing because the organizations getting value are compounding their advantages while the rest accumulate fragmented experiments with no coherent strategy.
This is a practice gap, and practice gaps compound
A capability gap can be closed with a purchase. Buy the software. Hire the consultant. Flip the switch. A practice gap compounds daily. Every day an AI-native organization runs production workloads, they learn something. They optimize a prompt chain. They discover a failure mode and build a guardrail. They find a workflow that saves 40 hours a week and redeploy that capacity into the next optimization.
You cannot skip that learning curve by signing a contract in October that you could have signed in April. The contract gives you access to the same models. It does not give you the six months of operational learning you chose not to accumulate.
Google understands this. That is why they committed $40 billion. Not because the models are magic. Because the ecosystem of organizations building on top of those models is reaching a density where the network effects become self-reinforcing. The infrastructure gets built for the organizations that are there. The products get optimized for their use cases. The talent gravitates toward the companies doing real work.
The investment is the signal. The ecosystem is the story.
Forty billion dollars is an attention-grabbing number. It should be. But the real story is the 1,000 enterprises spending $1 million or more per year, the 30x revenue growth in sixteen months, the 80 percent enterprise revenue mix. The real story is that an entire operating layer for business is being built, right now, by and for the organizations that showed up.
Every month you wait, the cost of entry goes up and the value of entry goes down. That is what compounding means. It works for the people in motion, and it works against everyone else.
The infrastructure is being poured. The ecosystem is forming. The question is whether your organization is shaping it or will be shaped by it.