Frontier AI Labs Funding Tracker (154 deals)

Our funding tracker covering the Frontier AI Labs Market

Below is the table listing all the deals. You may find our methodology at the end of this page.

And, if you want to gain a deeper understanding of the market and its current dynamics, get our report covering the Frontier AI Labs Market.

Is Frontier AI Labs really a $320 billion funding market?

Yes. Frontier AI Labs have raised at least $320.65 billion across the financing window we studied, but that number makes the market look much broader than it really is.

We found 154 qualifying financings across 92 companies. Seven deals have no defensible exact amount, so the actual total is somewhat higher. The missing capital does not change the basic picture because most of the money sits in a handful of enormous transactions.

The median disclosed round is $100 million. The average is $2.18 billion. When the average is almost 22 times the median, the headline total tells us much more about the biggest labs than about the typical company.

That distinction has become even more important lately because the largest labs are raising amounts that used to be almost unimaginable in venture capital. OpenAI closed a $122 billion financing in 2026. Anthropic raised $30 billion and then $65 billion within a few months. These deals now sit in a different financial universe from the rest of the startup market.

Is Frontier AI funding basically OpenAI and Anthropic now?

At the top, yes: OpenAI and Anthropic together account for roughly 78% of the known capital we tracked.

Anthropic accumulated $127.5 billion across seven qualifying financings, while OpenAI accounts for $122 billion. Add xAI, Moonshot AI and Safe Superintelligence, and the top five companies reach 86.5% of the market's known capital.

The market gets even narrower when we look at the top ten companies. They hold 93.1% of all disclosed funding despite representing only around 11% of the funded-company population.

Counting deals alone can give the wrong impression. We found 154 transactions, which sounds like a fairly broad market. But 60 of the 92 companies appear only once. Only 32 raised more than once during the period. A much smaller group keeps returning for larger checks.

Company	Qualifying financings	Known capital
Anthropic	7	$127.50B
OpenAI	1	$122.00B
xAI	2	$12.00B
Moonshot AI	6	$8.00B
Safe Superintelligence	3	$8.00B
DeepSeek	1	$7.10B
Mistral AI	3	$5.99B
StepFun	3	$3.22B+
Kling AI	1	$2.80B
Reflection AI	1	$2.00B

What does a normal Frontier AI funding round look like?

A normal Frontier AI round today is closer to $100 million than to the multibillion-dollar financings dominating the news.

The median disclosed financing is $100 million, while the median company has accumulated $105 million of known funding during the period. Those two figures give a much better sense of the middle of the market than the $2.18 billion average.

Even that middle is unusual. Only two known-size financings came in below $5 million. Fourteen were between $5 million and $20 million, 26 between $20 million and $50 million, and another 24 between $50 million and $100 million.

Then the distribution changes sharply. Eighty-one disclosed deals were at least $100 million.

For a founder, that means a $30 million or $50 million round can be substantial in absolute terms while still looking fairly modest inside Frontier AI. The companies competing directly at the general-purpose model frontier often need hundreds of millions before they even reach what another software sector would call mature scale.

Deal size	Deals	Share of known-size deals	Known capital
Under $5M	2	1.4%	$5.7M
$5M–<$20M	14	9.5%	$154.9M
$20M–<$50M	26	17.7%	$820.3M
$50M–<$100M	24	16.3%	$1.61B
$100M+	81	55.1%	$318.06B
Undisclosed exact size	7	-	Not included
All financings	154	-	$320.65B+

How much Frontier AI funding is left once we remove the mega-rounds?

Very little. If we remove every financing above $50 million, only about $1.28 billion of known Frontier AI capital remains.

Rounds above $50 million account for 99.6% of disclosed funding. Even if we raise the cutoff to $100 million, the remaining pool is only $3.39 billion.

Smaller venture rounds exist, but they barely move the dollar totals. The apparent size of Frontier AI depends almost entirely on companies that can raise nine figures or more.

The divide these days is less about whether a lab can raise money at all. It's about whether the company can repeatedly finance massive compute, research and deployment programs.

Which Frontier AI categories are getting most of the money?

Language and Reasoning Models currently absorb almost everything: 93.1% of known Frontier AI capital went to that category.

We counted 56 Language and Reasoning financings across 26 companies, representing $298.49 billion. Yet the category accounts for only 36.4% of all deals.

Multimodal Models are a distant second at $11.59 billion. Video Generation attracted $5.64 billion, Audio and Speech $2.48 billion, Agentic Model Platforms $1.78 billion and Image Generation $661 million.

The gap shows up in the size of individual rounds as well. The median Language and Reasoning financing is $500 million. Multimodal is $221 million. Video is $70 million and Audio is $45 million.

General-purpose model development is running on a much more expensive financial model than specialized AI.

Frontier AI category	Deals	Companies	Known capital	Median round
Language & Reasoning	56	26	$298.49B	$500M
Multimodal	32	24	$11.59B	$221M
Video Generation	22	15	$5.64B	$70M
Audio & Speech	26	15	$2.48B	$45M
Agentic Model Platforms	10	8	$1.78B	$166.9M
Image Generation	8	5	$661M	$58M
Total	154	92	$320.65B+	$100M

Why are language and reasoning AI labs raising such huge amounts?

The biggest Language and Reasoning labs now have to finance compute and physical infrastructure on top of research, products and hiring.

OpenAI made the point unusually clearly in its 2026 funding announcement: durable access to compute is becoming a strategic advantage across research, products and delivery. The company has also said that its Stargate infrastructure program has already moved beyond its original 10 GW U.S. capacity target ahead of schedule.

Anthropic shows the same dynamic. Its $30 billion financing was described as supporting frontier research, products and infrastructure expansion. When the company raised another $65 billion later in 2026, it again said the money would help expand compute.

The latest capacity commitments put those financing numbers in perspective. Anthropic has announced agreements involving up to 5 GW with Amazon, another 5 GW with Google and Broadcom, and additional GPU capacity through SpaceX. OpenAI is building multi-gigawatt data-center campuses with partners including Oracle and SoftBank.

Stanford's 2026 AI Index points in the same direction: frontier-model development keeps requiring much larger underlying compute systems even as individual chips become more efficient.

At that scale, a giant financing pays for much more than a larger research team. It supports the industrial system required to train and serve the models.

Are multimodal, video, audio and image AI getting funded the same way?

No. Specialized Frontier AI categories can attract serious money, but their financing is much less extreme and often depends heavily on one or two breakout companies.

Multimodal is the deepest category outside Language and Reasoning. Its 24 companies raised $11.59 billion, and the largest individual financing contributes around 22% of the category total. That still leaves a meaningful base of capital spread across several companies.

Video is much more top-heavy. Kling AI's roughly $2.8 billion financing alone represents close to half of the category's known funding.

Image Generation is even smaller. Black Forest Labs raised $300 million in its 2025 Series B, around 45% of all the Image Generation capital we recorded. Recraft's $30 million Series B shows how different the normal scale remains for another serious company in the same category.

Audio has more depth in company count but smaller checks. ElevenLabs raised $180 million in its 2025 Series C, Cartesia raised $64 million in Series A and Hume AI raised $50 million in Series B. Those are major venture rounds, yet they remain modest next to general-purpose frontier-model financings.

A category total of several billion dollars can describe very different markets. Sometimes the money is spread across many credible companies. Sometimes one outsized round does most of the work.

How young are Frontier AI labs when they start raising serious money?

Very young. The median Frontier AI company is only two years old when it raises a qualifying financing.

We could establish usable founding years for 85 of the 92 companies. Of those, 62 were founded in 2023 or later.

The 2023 cohort alone contains 29 companies. Another 11 were founded in 2024, 17 in 2025 and five in 2026.

Some of the most striking financings came almost immediately after the company was created. Safe Superintelligence was founded in 2024 and announced a $1 billion raise that same year. Thinking Machines Lab was founded in 2025 and raised $2 billion in its first observed financing.

Few venture markets allow teams to move from formation to hundreds of millions, sometimes billions, in funding before building a conventional multi-year startup financing history. Frontier AI does.

Is Frontier AI still an early-stage startup market?

Frontier AI is currently young in terms of companies but already late-stage in terms of where the money goes.

Seed and Series A represent 61 of the 131 financings for which we have a conventional stage classification. Nearly half of classified deals happen near the bottom of the ladder.

Those early rounds receive only about 5% of classified capital.

Series D+ and Growth Equity move the opposite way. Together they account for 31 financings, but they absorb most of the money in the staged portion of the market.

The check progression is also unusually steep. The median moves from $30.5 million at Seed to $50 million at Series A, $135 million at B, $300 million at C, $500 million at D+ and $750 million at Growth.

New labs keep arriving, while the largest checks go to the companies investors already see as capable of scaling aggressively.

Stage	Deals	Companies	Known capital	Median round
Seed	40	38	$6.66B	$30.5M
Series A	21	20	$9.30B	$50M
Series B	30	28	$13.60B	$135M
Series C	9	9	$10.47B	$300M
Series D+	15	8	$120.98B	$500M
Growth Equity	16	9	$151.61B	$750M
Unknown	23	17	$8.03B	$100M

Why are so few Frontier AI companies showing up at Series C?

Series C is currently the thinnest obvious part of the Frontier AI funding ladder, with only nine financings compared with 30 at Series B.

The check required at the same time becomes much larger. Median financing rises from $135 million at Series B to $300 million at Series C.

We should not read that as a literal survival rate because these deals do not form one single startup cohort. Some labs were already mature when our financing window began, 23 rounds do not have a clean conventional stage, and several Chinese financings use labels that map poorly onto the U.S. Seed-to-Growth sequence.

Still, the gap is hard to ignore. Plenty of emerging model companies can convince investors to back a substantial Series B. The group that moves into sustained late-stage financing is much smaller.

After that narrow middle, the market opens up again for proven leaders. We count 15 Series D+ transactions and 16 Growth Equity financings. Once a lab reaches that level, capital availability can expand dramatically.

Are investors funding new Frontier AI labs or mostly backing companies they already know?

Investors are mostly putting more money into Frontier AI companies that have already raised before.

Follow-on financings represent 118 of the 154 deals, or 76.6%. They account for $302.02 billion of known capital.

First financings represent 35 transactions and $18.55 billion.

The first-round group still produces spectacular exceptions. Safe Superintelligence's early financing and Thinking Machines Lab's $2 billion seed show that a newly formed company can jump straight into very large checks when the founding team carries enough credibility.

But the broader capital flow favors repeat winners. Anthropic, Moonshot AI, Mistral, Safe Superintelligence and several Chinese labs returned to investors multiple times during the period.

New labs continue to appear, but the capital advantage belongs overwhelmingly to companies with an established financing history.

Do strategic investors really matter in Frontier AI?

Strategic investors matter a lot in Frontier AI: they appear in 90 of the 154 financings we studied, and those deals are much larger on average.

The median round with a disclosed strategic investor is $292.6 million. The median without one is $50 million.

Look at who these investors are. NVIDIA supplies the accelerators. Amazon and Google provide huge amounts of cloud capacity. AMD makes AI chips. Cisco supplies networking infrastructure. Tencent and Alibaba combine cloud infrastructure, distribution and their own AI ecosystems.

Sakana AI offers a useful smaller-scale example. Its roughly $200 million Series A included NVIDIA as well as NEC, KDDI and Fujitsu, and Sakana explicitly announced a collaboration with NVIDIA covering research and infrastructure.

Mistral's €1.7 billion Series C was led by ASML, one of the world's most important semiconductor-equipment companies, with NVIDIA also participating. The relationship was described as a strategic partnership rather than simply another financial investment.

These companies can bring something that a purely financial investor cannot: direct access to the infrastructure and industrial relationships needed to keep scaling.

Why does NVIDIA keep appearing in Frontier AI deals?

NVIDIA has become unusually embedded in Frontier AI financing because the company sits directly on the industry's biggest constraint: compute.

We counted NVIDIA as a strategic investor in 32 financings after normalizing investor names. Its presence spans five of the six major model categories and stretches from early-stage rounds to mature growth financings.

xAI's 2024 Series C is a good illustration. NVIDIA and AMD both joined the $6 billion round while xAI was rapidly expanding Colossus, its giant GPU cluster. Thinking Machines later brought NVIDIA, AMD and Cisco into its financing. Safe Superintelligence has also developed a direct NVIDIA relationship around compute.

The freshest evidence goes well beyond startup investing. In August 2026, NVIDIA announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR aimed at mobilizing more than $500 billion of third-party capital for AI compute infrastructure.

NVIDIA itself described the idea as turning compute into an investable asset class. Chip supply, data-center finance and startup funding are becoming increasingly connected.

Which investors keep leading Frontier AI rounds?

Andreessen Horowitz is currently the most recurrent financial lead in our Frontier AI funding data, appearing as lead or co-lead in 11 deals.

Khosla Ventures follows with six. Sequoia Capital, Lightspeed Venture Partners and NEA each appear five times. General Catalyst and Menlo Ventures appear four times.

The market nevertheless has a very long investor tail. Among the 110 financings where a lead is disclosed, we identified 131 normalized lead investors, and only 28 show up more than once.

Plenty of investors are willing to participate in a Frontier AI deal. Far fewer repeatedly take responsibility for leading one, especially as round sizes escalate.

The combined round values below refer to the full financing rounds those firms led or co-led, not to the amount each investor personally contributed.

Lead investor	Led/co-led deals	Combined value of those rounds
Andreessen Horowitz	11	$124.70B
Khosla Ventures	6	$854.7M
Sequoia Capital	5	$66.93B
Lightspeed Venture Partners	5	$17.78B
NEA	5	$478.4M
General Catalyst	4	$787.7M
Menlo Ventures	4	$363M
ICONIQ	3	$108.00B
NVIDIA	3	$5.19B
Alibaba Group	3	$2.30B

Is Frontier AI funding basically a U.S. market?

The United States clearly dominates Frontier AI funding today, although the gap is much bigger in dollars than in startup activity.

North America accounts for 85 financings across 57 companies and $284.49 billion of known capital. The United States alone represents 79 deals, 55 companies and $283.17 billion.

Asia-Pacific still has substantial activity. We found 44 deals across 21 companies, including 34 transactions involving 15 Chinese companies. Yet the entire region accounts for only $25.86 billion.

Europe has 22 deals, 13 companies and $10.14 billion of disclosed funding. Mistral alone has a large effect on that figure.

China, Europe and Japan clearly have serious frontier-model activity. The U.S. advantage is primarily the extraordinary depth of capital available to finance the largest labs.

Do monthly Frontier AI funding totals tell us much about the market?

Not much on their own. Frontier AI monthly funding can jump by more than 100 times without any comparable change in the number of companies raising money.

Across the 33 completed months we studied, the median month contained four deals and about $1.2 billion of known capital.

The busiest month had 15 financings and roughly $9.93 billion of funding.

The biggest month by dollars had only four deals and reached $123.36 billion because one enormous transaction dominated the period. Capital in that month was 102.8 times the median monthly total.

Headlines such as "AI funding tripled this month" are especially unhelpful here. A monthly jump can simply mean that one of the largest labs happened to close its financing during that period.

Deal count gives a better view of how broad fundraising activity actually was.

Can a specialized Frontier AI lab still compete without raising billions?

Yes. A specialized Frontier AI company can still build a serious position with tens or hundreds of millions rather than billions, especially in voice, images and vertical multimodal systems.

ElevenLabs is the clearest example in Audio and Speech. Its $180 million Series C valued the company at $3.3 billion and funded further audio research, product development and international expansion. Cartesia's $64 million Series A and Hume's $50 million Series B show another viable scale for advanced voice-model companies.

The image market tells a similar story. Black Forest Labs pushed much higher with its $300 million Series B, but Bria raised $40 million and Recraft $30 million while building distinct commercial approaches around visual generation.

Liquid AI raised $250 million in Series A to expand efficient general-purpose models, including edge and on-premise use cases. Sakana AI raised about $200 million while pursuing a different research architecture from the giant U.S. labs.

These companies are hardly small. But they demonstrate a viable alternative: develop a differentiated model or deployment advantage, focus on a narrower technical problem and avoid funding a direct compute race with OpenAI or Anthropic.

What is Frontier AI funding still missing today?

Frontier AI currently has plenty of new-company formation and plenty of mega-rounds, but the middle of the market remains surprisingly thin.

Series C is the clearest example. Agentic Model Platforms are another. We found ten financings in that category, including seven Seed rounds and one Series A, but no transaction confidently classified as Series B or later.

Image Generation also lacks a mature financing ladder. Its eight deals across five companies include no classified Series C, D+ or Growth financing.

Repeat fundraising is far from universal as well. Sixty of the 92 companies appear only once during the period.

Some of these gaps will probably close simply because so many companies are young. The 2024, 2025 and 2026 founding cohorts have barely had time to build financing histories. But for now, Frontier AI looks much stronger at creating new labs and financing a small number of proven giants than at producing a thick layer of companies in between.

Is Frontier AI funding turning into infrastructure finance?

For the biggest labs, Frontier AI funding is increasingly behaving like infrastructure finance alongside venture and growth equity.

OpenAI's Stargate program originally set out to invest $500 billion in U.S. AI infrastructure. The company has since said it moved past the initial 10 GW capacity goal ahead of schedule and continues to secure new sites, power and data-center capacity.

Anthropic is assembling its own multi-provider compute portfolio across Amazon, Google, Broadcom, Microsoft, NVIDIA and SpaceX. The company's 2026 fundraising announcements explicitly connect new capital with infrastructure expansion.

NVIDIA has now taken the logic one step further by bringing large asset managers and private-capital firms directly into AI compute financing.

At the upper end, the money increasingly supports chips, power, networking, long-term cloud commitments and physical data-center capacity alongside model research.

A normal software company mainly needs capital to hire and sell. Today's largest Frontier AI labs also need to secure something closer to industrial capacity. Their funding rounds are starting to reflect that.

What should we actually take away from Frontier AI funding?

Frontier AI is one of the largest private funding markets we have ever seen, but most companies do not live in the financial world implied by the headline total. The median financing is around $100 million, while a tiny number of labs have pushed the market into tens or hundreds of billions.

The real center of gravity is general-purpose Language and Reasoning AI. That category has captured more than nine dollars out of every ten we tracked, largely because the leading labs now have to finance enormous amounts of compute and infrastructure.

At the same time, Frontier AI remains a remarkably young market. Most companies with known founding dates were created in 2023 or later. Investors are already willing to write nine-figure checks to teams that have existed for only a year or two.

The difficult part is staying in the race. New labs keep appearing, but three quarters of financings are follow-ons and serious scale capital goes disproportionately to companies that have already raised before. The thin Series C layer captures that gap particularly well.

Strategic investors now play a central role because capital alone is not enough. NVIDIA, Amazon, Google, AMD, Cisco and other industrial players can bring the chips, cloud capacity, networking and distribution that model developers need.

The result is a market with two speeds. Most Frontier AI companies still operate inside a recognizable venture-capital system. A small group at the very top has moved into something much closer to large-scale technology and infrastructure financing.