AI Is Hot. The Math Isn’t.
Everyone uses ChatGPT. Everyone talks about Claude. But using a product and owning the stock are two very different things.
OpenAI and Anthropic may be incredible companies. But if they come public near trillion-dollar valuations, the question is not whether AI is useful. The question is whether the financials can support the price.
OpenAI has reportedly filed confidentially for an IPO and could target a valuation of up to $1 trillion.
Anthropic was recently valued at $965 billion after raising $65 billion.
That is insane.
These are almost trillion-dollar companies.
But the profit proof is not there yet.
OpenAI
Reuters reported OpenAI has more than 900 million weekly active users and about $2 billion in monthly revenue.
That proves demand.
But demand is not profit.
OpenAI reportedly generated $5.7 billion in revenue in Q1 2026, but burned through $3.7 billion in the same quarter.
The Financial Times also reported OpenAI generated $13 billion in revenue in 2025, but spent $34 billion.
That included $19 billion in R&D.
R&D means research and development. For OpenAI, that means building better models, training them, testing them, and hiring elite researchers.
OpenAI also reportedly had a $39 billion net loss.
That number may be inflated by accounting charges, but even after cleaning that up, the business reportedly still lost around $8 billion from operations.
Simple version:
OpenAI has massive demand but they’re spending way too much to keep it.
OpenAI: Big Revenue. Bigger Spending.
Reported 2025 figures
Demand is real. But spending is running way ahead of revenue.
Anthropic
Anthropic is also exploding.
Reuters reported Anthropic raised $65 billion at a $965 billion post-money valuation, up from $380 billion a few months earlier.
Post-money valuation means the company’s value after the new investment.
Anthropic also reported crossing $47 billion in run-rate revenue. Run-rate revenue does not mean the company already made $47 billion. It means the company’s current revenue pace, annualized.
The revenue pace is impressive. The profit proof is still missing.
Anthropic’s Valuation Jump
Reported private valuation increase in 2026
February 2026
$380B
May 2026
$965B
The valuation more than doubled in a few months. The market is already pricing in a massive future.
Where Is the Money Going?
This is the real story.
OpenAI and Anthropic are not losing money because nobody wants the product.
They are losing money because AI is extremely expensive to build and run.
The biggest expense is compute.
Compute means the processing power needed to train and run AI models.
Training is building the brain.
Inference is running the brain every time a user asks a question.
Training is expensive once.
Inference is expensive every time.
That is the problem.
More users are good. But if every user is expensive to serve, growth can burn cash instead of creating profit.
Reuters reported OpenAI expects around $600 billion of compute spending through 2030.
OpenAI’s compute capacity reportedly grew from 0.6 GW in 2024 to 1.9 GW in 2025.
That sounds too technical, so here is the simple version:
1.9 GW is almost the power capacity of the Hoover Dam.
AI Runs on Power
OpenAI compute capacity vs Hoover Dam
1.9 GW running all year ≈ 1.5 million U.S. homes powered for a year
If that capacity ran all year, it would use about 16.6 TWh of electricity.
That is roughly enough electricity to power about 1.5 million U.S. homes for a year.
So OpenAI is not just building software.
It is building an AI factory.
And factories are expensive.
Anthropic’s AI Factory
Anthropic’s biggest expense is cloud and chips.
Amazon has invested billions in Anthropic. AWS became Anthropic’s primary cloud partner.
Google is also heavily involved.
Reuters reported Anthropic expanded its Google Cloud partnership to use up to 1 million Google TPUs, adding more than 1 GW of computing capacity starting in 2026. TPUs are Google’s AI chips.
Reuters also reported, citing The Information, that Anthropic committed to spending $200 billion over five years on Google Cloud services and chips.
The money is going into chips, cloud contracts, data centers, electricity, and engineers. AI looks like software on the screen, but financially it behaves more like infrastructure.
Final Verdict:
AI is real.
The products are amazing.
But the stock market does not only care about good products.
It only cares about profits.
Right now, OpenAI and Anthropic are growing like software companies, but spending like infrastructure companies.
That is the issue.
The question is simple:
Can they grow revenue faster than compute costs?
Until we see the full S-1 numbers, I would not trust the valuation.
AI is hot.
The math isn’t.
