OpenAI completes largest funding round ever, with Anthropic even more eager to go public

Author | Lin Wanwan

On March 31, 2026, OpenAI announced the completion of a $122 billion financing, with a valuation of $852 billion, the largest private financing in human commercial history.

Amazon invested $500 billion in OpenAI. $150 billion was received immediately, while the remaining $350 billion is contingent on a certain condition being met.

This condition is for OpenAI to complete an IPO or achieve AGI.

One is to go public, the other is to create superhuman-level general intelligence. The largest e-commerce company on Earth has placed a bet with an amount of money higher than the annual military budget of most countries on an "or" scenario.

Let's break down OpenAI's entire financing structure.

NVIDIA contributed $300 billion, and OpenAI happens to be one of NVIDIA's largest GPU customers.

OpenAI's CFO, Sarah Friar, has mentioned that most of the money will flow back to NVIDIA.

Amazon's $500 billion investment was made for OpenAI to run models on AWS for inference, leading to an increase in AWS revenue and improving Amazon's financial results. Microsoft has cumulatively invested over $130 billion, and OpenAI has committed to purchasing $2.5 trillion in cloud services on Azure.

Money circulated in a closed loop and came back. Wall Street calls this circular financing.

Bernstein analyst Stacy Rasgon stated that each such transaction deepens the market's concerns about circular financing. The CFA Institute's statistics are even more unsettling, with the total amount of commitments for investments and procurement among AI companies approaching $1 trillion.

But the topic of circular financing has been discussed for a whole year, and everything that needed to be said has been said.

What truly deserves attention in this $122 billion financing is not how the funds circulate but a more direct question: What are these funds actually buying?

What is the $852 billion Buying?

The answer is, buying time. More precisely, buying time until the IPO.

OpenAI's current monthly revenue is $20 billion, translating to an annualized revenue of around $240 billion. The $852 billion valuation corresponds to approximately a 35x Price/Sales ratio. This multiple implies that the market is paying for OpenAI three to four years in advance.

Get a few reference points to feel it out. NVIDIA was trading at about 20x PS in a crazy-profitable scenario. Snowflake peaked at 100x but quickly dropped back below 30. Salesforce was trading at around 10x when it went public.

Putting 35x on a company that is still losing money is already quite aggressive.

OpenAI's own plan is to reach $100 billion in revenue and $14 billion in profit by 2029. Going from $24 billion to $100 billion requires a continuous annual growth rate of over 40% for four years. I seriously thought about software companies that have maintained this growth rate on a billion-dollar revenue base in history, and I couldn't find a single one.

An $852 billion valuation can be justified under one condition: someone is willing to take it at that price on the public market. In other words, the IPO must be successful.

Once this layer is figured out, the entire financing structure makes sense.

Out of Amazon's $500 billion, $350 billion is tied to the IPO condition, meaning the money won't be received without going public. SoftBank's $300 billion is divided into three tranches, with the first tranche being paid at the close of financing and the remaining two tranches arriving in July and October, strategically aligned with crucial phases of IPO preparation.

OpenAI first sold 30 billion shares to retail investors through a bank and will also enter ARK Invest's ETF. Retail investors buying shares and entering the ETF create natural buying pressure for the IPO opening.

The wording in the financing announcement is no longer resembling a report to private investors. "We are the fastest platform to reach 10 million users, the fastest to reach 1 billion users, and soon to be the fastest to reach 10 billion weekly active users." "Our revenue growth rate is four times that of Google and Meta at the same time." This set of rhetoric can be directly transferred to the first page of the prospectus without modification.

A PitchBook study pointed out that among the three largest AI IPO candidates, OpenAI, Anthropic, and Databricks, OpenAI has the lowest business quality fundamentals score but the highest valuation.

Every design detail of the $122 billion financing points in the same direction: take this company public and let the public market catch this valuation.

Two Companies Fighting Over the Same Faucet

OpenAI needs an IPO, but it's not the only one that needs it. This is the real show of 2026.

First, look at the waiting list. CoreWeave went public last March at $40, now at $130, with a market cap of over $46 billion, setting the benchmark for other companies. Databricks, valued at $134 billion in its roadshow, with nearly $5 billion in annualized revenue. Cerebras resolved CFIUS scrutiny and resubmitted its IPO application.

The real heavyweights are Anthropic and OpenAI. Anthropic, valued at $380 billion, has engaged Wilson Sonsini for IPO legal prep. Kalshi predicts a 72% probability in the market that Anthropic will IPO before OpenAI.

This odds are tough on OpenAI. The pool of funds looking to invest in AI targets is limited, so if Anthropic consumes this batch of funds and attention first, OpenAI's IPO pricing will be compressed.

And Anthropic is indeed encroaching on OpenAI's territory. In the enterprise API market share, OpenAI dropped from 50% in 2023 to 25% in mid-2025, while Anthropic rose from 12% to 32% during the same period. Anthropic's revenue growth rate is approximately three times that of OpenAI. Some analysts extrapolate from the current trend that Anthropic will surpass OpenAI's annualized revenue by mid-2026.

Two years ago, OpenAI dominated the enterprise market, but now Anthropic is the leader in the enterprise API market. A single product of Claude Code has an annualized revenue of $25 billion, contributing to 4% of global GitHub public commits. This reversal speed is also rare in the tech industry.

Of course, OpenAI has its aces. 9 billion weekly active users, 50 million paid subscriptions, over $1 billion in annualized revenue from its advertising business, which has been in operation for six years. The brand awareness and user habits of ChatGPT are still the biggest moat in the AI industry. But the slowdown on the enterprise side is real.

Both companies are also spending money at an astonishing rate.

OpenAI is expected to lose $14 billion in 2026, with the annualized cash burn rate possibly reaching $57 billion by 2027. A $122 billion funding round sounds astronomical, supporting approximately 18 to 24 months. Anthropic is projected to spend $19 billion in 2026, $12 billion on training models, and $7 billion on running inferences.

The early listing gets the longevity. The money in the private market is almost running out for these companies, and the public market is the last untapped source. Renaissance Capital predicts there may be 200 to 230 IPOs in 2026. Just combining the IPOs of OpenAI, Anthropic, Databricks, and Cerebras could raise over $200 billion.

This is the largest technology IPO window since 2000. The last time such a wave of IPOs at this level occurred was also in 2000.

Can the Money-Making Speed Outpace the Spending Speed

All valuations, all funding structures, all IPO plans ultimately hinge on one judgment. Whether AI's money-making speed can outpace the spending speed.

If it does, $122 billion in funding is foresight, and an $852 billion valuation is a discount.

Some are already modeling scenarios where it doesn't. Analysts call it the CapEx Cliff, the capital expenditure cliff. If multi-billion-dollar data centers are built but the software running on them doesn't generate enough revenue to cover the costs, the efficiency revolution will replace the scale race. Companies that bet everything on "bigger is better" will find themselves sitting on a pile of expensive but underutilized hardware.

The progress in efficiency is faster than most people realize. To train a GPT-4 level model in 2023 cost around $79 million, but by 2026, with new hardware and techniques like distillation and quantization, the cost has dropped to $5 million to $10 million.

Last year, DeepSeek R1 trained a cutting-edge inference model for less than $300,000. In January of this year, they published a new training architecture paper, continuing to focus on efficiency. Google's latest Gemini 3.1 Flash-Lite has brought the inference cost down to $0.25 per million tokens. IBM researchers publicly stated that 2026 would be a year of divergence between cutting-edge large models and efficient small models.

If the efficiency path continues to outpace the scale path, the computing power empire built by OpenAI with an $852 billion valuation may face devaluation before it's even completed.

After the 2000 bubble burst, the Internet did not disappear; Google emerged from the ruins. What died were those companies that raised the most money at the peak of the bubble, built the most infrastructure, and never found a sustainable business model.

AI is not going away either. But whether the $122 billion and $852 billion valuations can hold up until the day of profitability is far from a done deal.

The drum is still beating, and the tempo is picking up.