There’s a story that doesn’t get told enough about today’s AI: it’s not a story of algorithms, models, or benchmarks. It’s a story of capital.
In the last weeks of May and early June 2026, three financial moves drew a clear picture of who’s betting what in the industry:
- Alphabet raised $80 billion to extend its compute infrastructure
- Anthropic filed its confidential S-1 for an IPO, according to press reports
- OpenAI and AWS formalized their strategic alliance (covered separately)
Three signals of the same phenomenon: AI is entering a phase where capital — not model intelligence — is the limiting factor.
Alphabet: $80 billion to avoid falling behind
That Alphabet raised $80 billion is not an isolated news item. It’s part of a wave of capital spending that’s redefining what “investing in technology” means.
To put it in perspective: $80 billion is more than Alphabet spent on capex in the last three years combined. It’s more than the GDP of entire countries. And Alphabet is not the exception — it’s the rule. Microsoft, Amazon, and Meta are on similar trajectories, though none has announced such a large figure in a single go.
Where is the money going? Primarily:
- Data centers: not traditional data centers, but hyperscale facilities designed specifically for AI workloads, with power densities that double or triple conventional data centers.
- GPUs and accelerators: the NVIDIA H200/B200 shortage remains a bottleneck, and hyperscalers are signing multi-year contracts to secure capacity.
- Interconnect networks: communication between GPUs in a training cluster requires specialized network infrastructure almost as expensive as the GPUs themselves.
Anthropic: from startup to potential IPO
Anthropic’s confidential S-1 filing marks a significant moment. In less than four years, the company went from being a research lab founded by ex-OpenAI employees to a corporation valued at nearly a trillion dollars ($965 billion according to Series H) preparing for the public market.
Anthropic’s IPO, if realized, would be the largest AI company IPO in history, surpassing any precedent. The timing is revealing: Anthropic is seeking public capital just as AI infrastructure spending reaches levels that even private capital can’t sustain alone.
The risk, of course, is public scrutiny. AI companies haven’t had to answer to public shareholders until now — and markets aren’t known for their patience with businesses burning cash at Anthropic’s scale.
The compute competition is the real competition
What these moves reveal is an uncomfortable truth for the industry: the battle for artificial intelligence is no longer fought primarily on benchmarks or model quality. It’s fought on the ability to secure and finance compute infrastructure.
Having the world’s best model is useless if you can’t:
- Train it (requires clusters of tens of thousands of GPUs)
- Serve it (requires globally distributed inference capacity)
- Improve it (requires continuous training and experimentation cycles)
Each of these steps costs billions. And the required scale only grows.
A new industrial logic
AI is adopting the logic of capital-intensive industries like energy or advanced manufacturing: whoever invests most in infrastructure wins. Not because money buys intelligence — but because without infrastructure, intelligence never comes into existence.
This has profound implications for industry competition. AI startups without access to infrastructure capital — or hyperscaler partnerships — simply won’t be able to compete at the frontier of models. Consolidation is inevitable.
And for hyperscalers, the question is whether this massive investment will pay off. Alphabet, Microsoft, and Amazon are spending as if AI will transform the global economy in the next five years. If they’re right, the return will be historic. If they’re wrong, the correction will be equally historic.
The capital is placed. Now we wait to see if the intelligence justifies it.