Kimi Secures $39 Billion in Funding Over Six Months, Valuation Reaches $31.5 Billion Pre-Deal

Author | Hua Hua

Six months ago, Kimi's founder, Yang Zhilin, said something.

He said that the Moon's Dark Side has sufficient cash on hand and feels no urgency to list in the short term.

What’s the result?

On June 30, The科创板Daily exclusively reported: Kimi has just completed its previous funding round, and a new round has already been launched. The post-money valuation of the previous round was $20 billion, and the pre-money valuation of the new round is $31.5 billion.

This is the fourth round in six months, with a total funding of over $3.9 billion over the half-year.

Six months ago, Kimi's post-investment valuation was only $4.3 billion.

In today’s Chinese venture capital market, it is taken for granted that Kimi should continue raising funds, just as OpenAI and Anthropic should as well.

This itself is more worth discussing than the $31.5 billion valuation.

Internet companies raise capital with a simple logic: exchange money for market share. Losing money today is fine—once you hit the inflection point, returns come back exponentially. For the past two decades, VCs have fully embraced this logic.

Large model companies are an exception; AI companies are inherently capital-intensive businesses.

For each new generation of the model, thousands of high-end GPUs, billions of dollars in compute costs, salaries for hundreds of researchers, and an entire data cleaning and post-training pipeline must be prepared.

Take OpenAI for example—it generated $13.07 billion in revenue for the entire year of 2025. That sounds like a lot, doesn't it?

However, the total cost and expenses amounted to $34 billion. Just R&D alone cost $19.18 billion—more than the company’s entire annual revenue. Of this, the compute bill paid to Microsoft was $10.59 billion. Added to this were $7.5 billion in inference costs and $5.73 billion in sales and marketing expenses, resulting in an annual operating loss of $20.92 billion.

By the first quarter of 2026, data showed that cash burn continued to expand: revenue amounted to $5.7 billion, cash consumption reached $3.7 billion—burning through more than half of revenue—and research and development spending totaled $8.6 billion.

OpenAI predicts its full-year cash burn will be $25 billion in 2026 and $57 billion in 2027.

If OpenAI demonstrated why AI companies burn cash, then Meta and Google have shown that the entire industry is entering a capital expenditure race.

Meta's full-year capital expenditure guidance for 2025 is $70 to $72 billion. For 2026, it has been directly raised to $125 to $145 billion, nearly doubling. This increase is entirely allocated to heavy assets such as AI servers, in-house developed accelerators, and global data centers.

Google is the same. Its capital expenditure guidance for 2025 was $75 to $78 billion, but in the latest update in June 2026, it was raised to $180 to $190 billion. The CFO stated verbatim that capital spending will continue to increase significantly in 2027.

This money isn’t spent and gone—it’s largely converted into computing power as fixed assets. The AI industry doesn’t consume one-time cash flow; it requires continuously expanding computational resources. The funds you invest directly correspond to hardware and racks in data centers.

Internet companies solve information distribution. Once the product is built, the marginal cost of serving one more user approaches zero. As scale increases, almost all incremental revenue becomes profit.

AI companies address the ongoing demand for computing power. Once infrastructure stops investing in additional computing capacity, model iteration and service capacity begin to decline immediately.

Kimi raised $3.9 billion in six months, seemingly breaking the conventional pace of the primary market. However, from a financial standpoint, this funding must cover team salaries, data center electricity costs, and high-end GPU purchases, sustaining operations until the next round of model development and commercialization.

These inputs ultimately become tangible assets. What you're buying are cards, data centers, and kilowatt-hours of electricity.

Having ample cash on hand doesn't mean you can stop.

In the large model industry, every link in the full chain is burning cash—training foundational models, providing online inference services, and building the Agent ecosystem. Reducing investment in any one area causes the growth curve to plummet immediately.

The harsh reality of this industry is that while your competitors invest tens of billions annually to iterate their foundational models, you're only investing hundreds of millions—the gap isn't just about the speed of iteration; it's a generational divide.

The capabilities of large models grow nonlinearly. New abilities only emerge once computational scale surpasses a certain threshold. Until you reach that point, your model is not even on the same competitive curve as leading vendors.

In this industry, technological gaps are almost immediately reflected in business outcomes.

Insufficient computing power → Model falls one generation behind → Developers migrate to competitors → End users churn → Revenue dries up → Funding channels narrow.

If any link in the chain fails, the entire business will slow down.

Kimi continues to raise funds not to pursue rapid expansion, but solely to maintain the baseline of technological iteration and avoid being left behind by industry advancements.

All top global AI companies are the same: OpenAI, Anthropic, and others all maintain frequent and substantial fundraising. Kimi has simply fully revealed this high-capital survival logic in the Chinese market.

The most interesting number in this funding round isn't the $31.5 billion valuation.

It's ARR, Annual Recurring Revenue.

In March 2026, Kimi ARR surpassed $100 million. In May, it exceeded $200 million. By mid-June, it officially broke through $300 million.

Tripled in three months, with over 70% coming from API developer calls.

A $31.5 billion valuation reflects the market's pricing of future potential. A $300 million ARR represents proven, verifiable real revenue.

Over the past three years, valuations in China’s large model sector have been sustained primarily by industry narratives. From Zhipu to MiniMax to Moonshot AI, the universal story boils down to three phrases: China’s version of GPT, catching up to OpenAI in the short term, and China’s Anthropic.

This time, Kimi did something different: it directly presented measurable, consistently growing real financial revenue to the primary market—$300 million in annual recurring revenue (ARR), with APIs accounting for 70% of revenue, overseas paying users up 400% year-over-year, and the product deployed in over 200 countries.

This is the first time in China's AI industry that valuations have been supported by actual commercial revenue rather than just technological narratives. This signifies that the large model industry is entering a new phase where revenue validates technology.

The impact of this change is much greater than the $31.5 billion valuation itself.

Starting from mid-2026, domestic large model companies have completely split into two categories. One group is still competing on model benchmarks and technical specifications, while the other has established steady, recurring revenue from paying customers.

Among companies that have generated revenue, internal divisions are also emerging:

Zhipu, listed on the Hong Kong Stock Exchange in January 2026, simultaneously launched its A+H listing on the STAR Market. Full-year 2025 revenue reached RMB 724 million, with a gross margin of 41%, primarily focusing on localized MaaS deployments for government and enterprise clients.

MiniMax follows Zhipu’s listing on the Hong Kong Stock Exchange. In the first three quarters of 2025, revenue reached $53.43 million, with 73% coming from overseas markets, driven by consumer subscription products Talkie and Hailuo AI.

Jieyue Xingchen, with its Pre-IPO round led by telecom industry capital, aims to submit its Hong Kong stock listing application within 2026. Full-year revenue for 2025 is projected to be approximately RMB 500 million.

ZeroOne Technologies has voluntarily exited the general-purpose foundation model race; in June, it partnered with CP Group to establish Wanfeng Intelligence, focusing on egg-laying chicken farming.

BaiChuan Intelligence is scaling back its general-purpose large models and redirecting all resources and funding toward the healthcare vertical.

The former "Six Little Tigers of AI" are now in entirely different sectors, markets, and commercialization paths, and no longer compete in the same group.

Kimi has walked a path never fully traversed by domestic manufacturers—prioritizing APIs, driving growth through developer ecosystems, and generating revenue through paid overseas users. This strategy closely follows Anthropic’s playbook.

Anthropic's post-H-round valuation in May 2026 reached $965 billion, with its revenue growth trajectory serving as highly instructive: its annual recurring revenue (ARR) was only $1 billion by the end of 2024, rose to $14 billion by February 2026, and officially surpassed $47 billion in annualized revenue by the end of May; from the end of 2024 to May 2026, its annualized revenue expanded 47-fold.

If Kimi can successfully implement this overseas API route, a $31.5 billion pre-money valuation is just the beginning. If commercial growth falls short of expectations, this valuation will mark the peak of a private market bubble.

In traditional primary markets, it's commonly understood that the higher a company's valuation, the harder it is to raise funds, and the fewer institutions are able to make large investments.

The AI sector is completely reversed: the higher the valuation of leading companies, the easier it is for them to secure large, long-term funding.

The key change is: the person funding it has changed.

The share of traditional VC is shrinking. Industrial capital, state-owned industrial funds, and global sovereign wealth funds have become the main players.

DeepSeek recently completed its first round of external funding, raising over RMB 50 billion, with a post-money valuation of approximately RMB 400 billion (nearly $59 billion). The investors are primarily national-level AI industry funds and leading industrial conglomerates.

Over the past six months, Kimi’s investors have included Meituan Dragonfly, China Mobile, CPE Source Peak, Temasek of Singapore, and Abu Dhabi’s MGX. In Anthropic’s Series G and H rounds, investors such as GIC, MGX, and the Qatar Investment Authority also appeared.

These long-term institutional players aren't seeking short-term returns of 3 to 5 times. They're seeking entry into the next-generation digital infrastructure.

Tencent’s investment in Kimi differs from its earlier investments in JD.com, Meituan, and Pinduoduo. Previously driven by financial returns, today’s strategic positioning in leading large models aims to secure a foundational position in the underlying industry chain.

For national capital, AI has been elevated to the same level as electricity, railways, and 5G—as the next-generation infrastructure of the national economy. Possessing a globally competitive foundational large model means controlling the话语权 over computing power distribution and data flow in the future digital economy.

So today’s AI funding is no longer merely financial investment—it’s a long-term strategic allocation of resources by major nations and industrial conglomerates.

Enterprise valuations now simultaneously reflect both industry value and strategic value, with industry weights and geopolitical strategic weights incorporated into pricing.

$31.5 billion, not the result of market-driven bidding, but a allocation jointly determined by multiple industry and sovereign capital participants.

This also explains why Kimi isn't rushing to go public.

Zhipu and MiniMax went public on the Hong Kong Stock Exchange in January this year, Jieyue Xingchen is preparing to submit its listing application, and DeepSeek has secured substantial funding from state-backed investors. Only Moonshot AI remains in the private market.

What is Kimi waiting for? It’s likely waiting for ARR to grow from $300 million to $3 billion, or even reach the tens of billions.

Money in the primary market can tolerate long-term lack of profitability, waiting for revenue to grow through compounding. The secondary market cannot—it focuses on short-term metrics such as quarterly growth rates, price-to-sales ratios, and hash rate liabilities. Once listed, the future compounding growth of revenue is preemptively discounted by the market.

Staying in the primary market, Kimi can fully complete its commercialization journey and maintain control over its narrative pacing.

The core logic of making money on internet platforms is that marginal costs approach zero.

The same product costs $10 to serve the first user, but serving ten million users barely increases the cost per user. The larger the scale, the wider the profit margin. Over the past two decades, consumer internet platforms have relied entirely on this model to generate returns.

The cost curve for AI companies is completely opposite.

As the number of users and API call volume increase, operational costs rise accordingly. Higher call volumes lead to greater losses.

If you break down the costs of an AI company, it might look something like this:

Total computing cost ≈ Number of active users × Number of inference requests per user × Unit computing price

If any one of the three increases, the total cost rises accordingly. If all three increase together, the cost expands exponentially.

OpenAI previously disclosed that when ChatGPT first launched, its daily inference cost was approximately $7 million. That was during the early stage with a small user base. Today, with user numbers and API call volumes having increased manyfold, the actual daily cost is surely even higher.

AI is not traditional software; it is more like a capital-intensive utility that continuously consumes energy.

After the computing power data center is built, scaling up user demand will not dilute fixed costs such as hardware depreciation, electricity, and bandwidth. The more users served, the more computing resources are consumed.

High asset intensity, rapid growth, and persistent funding gaps are typical characteristics of leading foundational large model companies.

Before achieving stable profitability, OpenAI, Anthropic, and Kimi will maintain a high frequency of fundraising.

Looking back at Kimi’s key metrics: a pre-money valuation of $31.5 billion, an ARR of $300 million, API revenue accounting for 70%, a 400% increase in overseas paying users, and the initiation of a new large funding round.

These figures do not describe a newly profitable company, but rather a capital-intensive business whose operations are functional yet simultaneously creating a larger gap in computing power funding.

The higher the API call volume, the greater the demand for inference hardware and power consumption. The faster the expansion into overseas markets, the heavier the capital investment required for cross-border computing nodes and multi-regional compliance. For every additional dollar of ARR, there is often a multiple-fold increase in fixed capital expenditure.

The revenue curve is rising, while the capital expenditure curve is steeper.

Broadly speaking, this isn't just a problem with Kimi—OpenAI and Anthropic face the same issue.

They are all in the same operational state, with revenue growing and capital expenditures growing even faster.

Funding for leading AI companies is not a phased tool for expansion—it is a survival imperative to sustain core computing power iteration.

The real core issue isn't whether you can generate revenue or become profitable—it's whether you can consistently secure funding and survive this computational arms race.

[Outside the layout]:

$3.9 billion has come in, and a $31.5 billion valuation hangs in the balance. These two figures don’t signify Kimi’s safe landing—they’re more like a countdown.

The higher the valuation, the higher the barrier for the next round. A $31.5 billion pre-money valuation means that, for the next round, at least one investor must be willing to take over at a $40 billion or higher valuation. Institutions capable of offering this price are few and far between globally.

With each funding round, Kimi gets one step closer to the ceiling of unfundable. But your compute bill won’t pause just because you can’t secure funding.

What Yang Zhilin is truly racing against is not technology, but time—before the funds from the primary market run out, he must push ARR to a level sufficient to support a $100 billion valuation. If he succeeds, going public is a coronation. If he fails, going public is a defeat.

The time left for him may be shorter than the market assumes.

This article is from the WeChat public account "Beyond the Layout," authored by Hua Hua.