What are the latest funding trends in Frontier AI Labs?

All funding deals in the Frontier AI Labs Market over the last years

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

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

Is Frontier AI funding really booming right now?

Frontier AI funding is genuinely much stronger today: more labs are raising, the typical round is much larger, and the improvement survives even after we strip out the giant financings.

We counted 65 comparable financings in the current YTD period, versus 29 in the same stretch of 2025 and 32 two years ago. The number of funded companies climbed even faster, from 21 last year to 55 now. So this is not simply the same handful of labs coming back for ever-bigger checks.

The headline dollar number is obviously extreme. Known capital jumped from $25.89 billion to $256.48 billion. Yet below $100 million, funding still increased from $722 million to $1.15 billion, while the median disclosed round rose from $100 million to $220 million.

That gives us a fairly clear answer. The market underneath the megadeals is stronger too; the giant rounds simply make an already-strong year look almost surreal.

Metric	2026 YTD	2025 YTD	2024 YTD	Change vs 2025
Deals	65	29	32	+124%
Funded companies	55	21	29	+162%
Known capital	$256.48B	$25.89B	$11.51B	+891%
Capital from rounds ≤$50M	$416M	$251M	$309M	+66%
Capital from rounds ≤$100M	$1.15B	$722M	$887M	+59%
Median disclosed round	$220M	$100M	$80M	+120%
Average disclosed round	$4.01B	$925M	$397M	+334%
Largest round	$122B	$13B	$6B	+838%

Is OpenAI making the Frontier AI funding market look stronger than it really is?

OpenAI makes the dollar boom look far bigger than the underlying improvement, but it does not create the improvement by itself.

OpenAI's $122 billion financing accounts for 47.6% of current capital. The top three rounds together absorb 84.6%. That is enough to make any comparison of raw funding totals misleading if we stop there.

The rest of the evidence points the same way, though. Capital from rounds of $100 million or less is up 59% year over year. Capital from rounds of $50 million or less is up 66%. The median round has doubled, and far more companies have managed to raise.

OpenAI itself also gives us a clue as to why the top of this market is getting so expensive. In its 2026 financing announcement, the company described durable access to compute as a strategic advantage rather than a simple operating expense. These days, the leading frontier labs are partly being financed like software companies and partly like enormous AI infrastructure projects.

So the $256 billion figure badly exaggerates the speed of the underlying venture-market expansion. It does not reverse its direction.

Are $100 million Frontier AI rounds basically normal now?

Yes. Nine-figure Frontier AI rounds have moved from exceptional to routine surprisingly quickly.

Forty-one of the 65 current financings are at least $100 million. Two years ago, only 34.4% of deals crossed that line. Last year, it was 51.7%. Today it is 63.1%.

The $220 million median is even more telling because it is much harder for a single megadeal to distort. Half of the disclosed rounds are now larger than that.

We are also seeing the middle of the market thin out. The $50 million-to-$100 million range represented a quarter of comparable deals in 2024. It is below 10% now. Frontier AI financing increasingly jumps from relatively small startup rounds to checks that already look like growth capital.

Deal size	2026 deals	2026 share	2025 share	2024 share
Under $5M	2	3.1%	0%	0%
$5M–<$20M	6	9.2%	10.3%	15.6%
$20M–<$50M	9	13.8%	17.2%	15.6%
$50M–<$100M	6	9.2%	17.2%	25.0%
$100M+	41	63.1%	51.7%	34.4%
Undisclosed	1	1.5%	3.4%	9.4%
All deals	65	100%	100%	100%

Can a new Frontier AI startup still get funded today?

Very much so. New Frontier AI labs are getting funded again, and some are starting with amounts that would once have looked like late-stage rounds.

We identified 19 current financings as first financings, compared with only three in the same period last year and eight in 2024. Their share of activity has moved back to 29.2% after falling to just 10.3% in 2025.

The size of these first rounds is the more unusual part. The current median is $200 million, versus $13 million last year. DeepSeek is an extreme example: its first outside financing raised more than $7 billion through an unusual limited-partnership structure. The Information reported that founder Liang Wenfeng himself committed roughly RMB20 billion, alongside Tencent, CATL, JD.com, NetEase, IDG Capital and China's national AI investment fund.

We should not read the financing-status classification as a perfect reconstruction of every company's lifetime funding history. But the jump is far too large to dismiss. A new group of labs is clearly getting capital now.

Metric	2026 YTD	2025 YTD	2024 YTD
First financings	19	3	8
First-financing share	29.2%	10.3%	25.0%
First-financing capital	$14.92B	$2.02B	$1.25B
Median first financing	$200M	$13M	$30.5M
Follow-on financings	45	26	24
Follow-on share	69.2%	89.7%	75.0%
Follow-on capital	$241.49B	$23.87B	$10.27B
Median follow-on	$227.5M	$100M	$80M

Are Frontier AI investors only backing proven winners now?

No. Investors are putting enormous amounts into the established leaders while also reopening the door to much earlier companies.

Seed deals jumped from four in the previous comparable period to 19 now. Series A rose from three to nine. Together, Seed and Series A account for 43.1% of current transactions, almost back to their 46.9% share in 2024 after an unusually mature 2025 market.

The amounts going into those stages are also climbing. Median Seed financing rose from roughly $12 million to $40 million, while the Series A median went from $50 million to $100 million.

Later-stage companies still absorb nearly all the money. That is where the market has become extreme. But the financing pipeline itself currently reaches all the way from newly formed labs to the biggest incumbents.

Why does Frontier AI still feel like a late-stage funding market?

Because the dollars are overwhelmingly late-stage even though the deal count is not.

We count 28 early-stage deals and 28 later-stage deals now. By transaction count, the market is almost perfectly balanced. Later-stage financings, however, account for roughly $241 billion, or 94% of known capital.

Growth Equity shows the split most clearly. Only five deals fall into that bucket, yet they consume more than half of all current funding. Their median size is roughly $2.8 billion.

This is where the Frontier AI funding market gets unusual. There is still a functioning startup layer, but companies that reach the top tier can suddenly require capital measured in billions or tens of billions.

Is Language and Reasoning AI swallowing all the Frontier AI money?

Almost all the money, yes. Language and Reasoning models currently take 94.3% of Frontier AI capital despite accounting for only a third of the deals.

That gap has become wider this year. Language and Reasoning's deal share fell by 11 percentage points from the previous comparable period, while its capital share increased.

OpenAI's $122 billion financing is the biggest reason. Anthropic then added a $65 billion Series H, led by Altimeter, Dragoneer, Greenoaks and Sequoia, at a $965 billion post-money valuation. Anthropic said at the time that its run-rate revenue had passed $47 billion and that the money would help expand compute capacity as Claude demand grew.

DeepSeek adds another multi-billion-dollar financing to the same category. A relatively smaller group of Language and Reasoning labs is therefore pulling in amounts that the rest of the model market cannot come close to matching.

Meanwhile, Multimodal and Agentic Model Platforms are winning more actual deals. Money and startup formation are moving in different directions.

Category	Deals	Deal share	Capital	Capital share
Language & Reasoning	22	33.8%	$241.85B	94.3%
Multimodal	17	26.2%	$8.05B	3.1%
Audio & Speech	11	16.9%	$1.20B	0.5%
Agentic Model Platforms	8	12.3%	$1.56B	0.6%
Video Generation	6	9.2%	$3.74B	1.5%
Image Generation	1	1.5%	$76M	<0.1%
Total	65	100%	$256.48B	100%

Are agentic AI labs becoming a real new Frontier AI category?

Yes. Agentic Model Platforms currently look like the clearest genuinely new cluster forming inside Frontier AI.

We found eight financings and roughly $1.56 billion of capital in the current period, versus no comparable deals last year and just one two years ago.

The companies are young too. Where we have usable founding dates, the median age is roughly 0.8 years. That makes the jump harder to explain away as old AI companies simply adopting a fashionable label.

There are fresh technology developments behind the funding shift as well. Scale launched Scale Labs this year with research spanning agentic and multimodal systems. Alibaba has been pushing what it calls a full-stack agentic AI strategy, including newer Qwen models, cloud infrastructure and an enterprise agent platform. Moonshot's Kimi K3 is explicitly described by the company as a native multimodal agentic model.

The capital is still small next to Language and Reasoning. For now, the interesting part is the speed at which new companies and new model architectures are appearing.

Is multimodal AI actually gaining ground right now?

Yes. Multimodal AI is attracting far more Frontier AI deals even though the megadeals make its funding share look smaller.

We count 17 current financings, giving Multimodal 26.2% of deals. Its deal share has risen by almost nine percentage points from last year. Absolute capital has also more than tripled to roughly $8.1 billion.

Yet Multimodal's share of total dollars fell. That sounds contradictory until we remember what happened at the top of Language and Reasoning.

The cleaner reading is that investors are funding more multimodal labs and writing substantial checks to them, but nobody outside the biggest language-model companies can keep pace with today's biggest rounds.

As we saw above, deal share is the more useful measure of startup formation here than capital share.

Is standalone image-generation AI fading?

Standalone image-generation funding certainly looks weaker now, although the sample is small enough that we would not call the category finished.

There were four comparable image-generation deals in 2024, two last year and one in the current period.

The broader product market gives that decline some context. Image capability is increasingly bundled into multimodal systems rather than treated as a separate frontier-model stack. A company can now compete on text, vision, audio, video and agentic behavior in the same model family.

That makes pure image-generation labs harder to interpret as their own funding category. The three-year drop is worth watching, but one or two financings could still move the numbers quickly.

Is audio AI having its own funding boom?

Audio and Speech AI is getting much broader investor attention, but the checks themselves are not getting bigger.

Current deal count reached 11, versus four in the comparable 2025 period. The median financing, however, slipped from roughly $38 million to $30 million, and the category accounts for only about 0.5% of total capital.

Investors are willing to fund more audio companies these days without treating the category like the next capital-intensive frontier race.

It is a healthier pipeline, but a very different kind of expansion from what we see in Language and Reasoning or even Multimodal.

Is Asia becoming a serious second center for Frontier AI funding?

Yes. Asia-Pacific is now the clearest geographic challenger to North America when we look at funded labs rather than just the giant dollar totals.

APAC deal count climbed from seven in the previous comparable period to 20 now. The region represents 30.8% of current deals, and its median financing is close to $293 million.

DeepSeek is obviously part of that shift, but it cannot explain a near-tripling of deal activity by itself. Moonshot is also competing at the frontier, and Alibaba is building across proprietary chips, cloud infrastructure, foundation models and agentic applications. Alibaba's 2026 annual report said AI-related products had reached 30% of Cloud Intelligence Group external revenue in its final fiscal quarter, while the company continues to expand AI infrastructure across Asia and Europe.

North America still dominates funding dollars because OpenAI, Anthropic and other U.S. labs raise at enormous scale. The map of funded companies is becoming considerably less one-sided.

Is Europe falling behind in Frontier AI funding?

Europe is raising more Frontier AI capital and funding many more labs, even though its global capital share has fallen.

The region went from three comparable financings last year to 11 now. Absolute capital reached roughly $6.8 billion.

Mistral is the clearest example of what European frontier funding looks like today. The company recently raised €3 billion at a post-money valuation above €21 billion, with Samsung Electronics leading and EQT's Scaleup Europe Fund and PSG Equity co-leading. Mistral said the financing would expand frontier research, compute capacity, infrastructure and international deployment.

The company has kept moving since the round. Its current work includes Mistral Large 4, an open-weight model trained in European infrastructure, which gives the funding a visible technology program behind it rather than making it look like a purely defensive "European champion" bet.

Europe's lower capital share mostly reflects the extraordinary denominator created by U.S. megadeals. Its own funding activity is considerably stronger than it was a year ago.

Why are strategic investors suddenly all over Frontier AI funding?

Strategic capital is now part of the basic financing structure of Frontier AI because access to compute, chips and cloud capacity has become almost as important as the cash itself.

Strategic investors participate in 72.3% of current deals, versus 41.4% in the previous comparable period. The median strategic-participation round has actually fallen from $404 million to $305 million, so corporate involvement is spreading beyond a few giant transactions.

Anthropic gives us the clearest current example. The company has agreed to secure up to five gigawatts of Amazon compute and committed more than $100 billion over ten years to AWS technologies. It separately expanded its relationship with Google and Broadcom for multiple gigawatts of next-generation TPU capacity. Amazon also announced an additional $5 billion investment in Anthropic, with the possibility of another $20 billion tied to commercial milestones.

OpenAI has made a similar point from its own side, describing durable compute access as a strategic advantage. Mistral talks explicitly about financing compute capacity and infrastructure alongside model research.

When corporate investors show up in Frontier AI today, we increasingly have to look beyond the cap table. The cloud contract, chip supply, deployment channel or infrastructure commitment may be just as important as the equity check.

Are hyperscalers crowding traditional VC firms out of Frontier AI?

No. Traditional investors are still very active, and the group leading Frontier AI rounds has become much broader.

We identified 86 unique disclosed lead or co-lead investors in the current period, compared with 23 last year. Fifteen appear more than once.

Sequoia stands out with four current lead or co-lead appearances after none in the previous comparable period. GIC, Coatue, D. E. Shaw Ventures, Dragoneer, MGX, Greycroft, GV, Cambium Capital and Alibaba Cloud each appear twice. Andreessen Horowitz remains active as well.

Lead disclosure is not perfect, so an investor missing from the current list should not be treated as having left the market. Still, a near-fourfold expansion in disclosed lead names is too large to write off as reporting noise.

The funding market currently has more corporate involvement and more financial-investor breadth at the same time.

Is Frontier AI funding getting more concentrated or more competitive?

Frontier AI is becoming more competitive at the startup level and more concentrated at the winner level.

The top three rounds account for 84.6% of current capital, well above the 71.5% share one year earlier. Yet no single deal dominates the market quite as much as the biggest deal did in either prior comparable period.

Underneath those enormous rounds, many more companies are getting financed, new entrants have returned and the lead-investor pool has widened sharply.

That leaves a weird market structure. There are more credible attempts to build frontier AI companies than a year ago, while the amount of money separating the largest labs from everybody else has widened enormously.

For founders, those two realities can coexist. Getting a first serious round currently looks more possible than it did in 2025. Staying in the top frontier group may require a completely different class of capital.

Are Frontier AI investors betting on much younger companies now?

Not across the whole market. Frontier AI has a lot more Seed activity now, but the typical funded company is roughly the same age as last year's.

Among deals with usable founding dates, median company age is about 2.7 years today versus about three years in the previous comparable period. The more striking contrast is with 2024, when the median was only around 1.1 years.

The stage progression also looks sensible. Current Seed companies are around 0.8 years old at the median, Series A companies about 2.5 years, Series B companies about three years and Series D-plus companies close to four years.

Agentic Model Platforms are the youthful exception, with a median under one year.

We are more cautious with this part of the analysis. Founding-date coverage is weaker for the current period, and several raw records contained malformed spreadsheet-style values that we excluded rather than trying to reconstruct them.

Has Frontier AI already beaten the previous full funding years?

Yes. Frontier AI has already passed the full-year deal counts of both 2024 and 2025, even before the current year is complete.

The previous full year contained 41 deals and roughly $30.6 billion of known capital. The current YTD period is already at 65 deals. Full-year 2024 had 46.

The funding comparison is much more extreme, although we should not pretend that capital arrives smoothly throughout the year. OpenAI and Anthropic alone make mechanical annualization fairly meaningless.

The useful point is simpler. As of now, Frontier AI does not need a strong finish to qualify as a much busier financing year. It has already cleared both previous full-year deal totals.

So what is actually changing in Frontier AI funding?

Frontier AI funding is stronger across the market today, but the biggest change is the widening gap between getting into the race and staying at the very front of it.

More labs are getting financed. New-company formation has returned. Seed and Series A activity has recovered. Agentic and Multimodal companies are taking a larger share of deals, while Asia-Pacific and Europe are producing far more funded labs than a year ago.

At the same time, the leading Language and Reasoning companies now raise amounts that belong in a different financial universe. OpenAI, Anthropic and several other top labs have turned compute access, cloud partnerships and strategic infrastructure into core fundraising issues. As pointed out above, that is why raw capital has risen far faster than ordinary-round funding.

Investor behavior has changed with it. Corporate participation has spread much further down the market, yet traditional lead-investor participation has broadened too. Frontier AI currently attracts more kinds of capital, from more investors, across more companies.

The strongest current takeaway is the breadth of the expansion rather than the headline $256 billion figure. The main caution is concentration: a handful of labs still absorb nearly all the money. We have a bigger Frontier AI funding market now, but an increasingly steep hierarchy inside it.