What’s funding like in the Foundation Model market right now?

All funding deals in the Foundation Model Market over the last 24 months

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 Foundation Model Market.

Is the Foundation Model market actually healthier today?

The Foundation Model funding market is healthier today on breadth and financing capacity, although the money itself has become more concentrated at the top.

We counted 75 financings in the latest 12 months, compared with 45 in the previous period. The number of funded companies rose from 36 to 57, and the median disclosed round jumped from $107.5 million to $242 million. Those three measures point in the same direction: more companies are getting financed, and the typical company that raises is getting a much bigger check.

The picture becomes less comfortable when we look at where the dollars went. A few frontier labs absorbed most of the capital, led by OpenAI and Anthropic. The market is broader, deeper and much more expensive than a year ago, but it is also increasingly dominated by companies that can raise at a scale almost no startup can match.

Market measure	Current L12M	Previous L12M	Change
Deals	75	45	+66.7%
Unique funded companies	57	36	+58.3%
Companies raising more than once	12	6	+100%
Median disclosed round	$242M	$107.5M	+125.1%
Average disclosed round	$3.57B	$1.87B	+90.6%
First financings	14	9	+55.6%
Zero-deal months	0	1	Better continuity

Does the Foundation Model funding boom survive without the giant rounds?

Yes, but only as a much smaller boom. Foundation Model funding still improved below the megadeal tier, while the headline 212% increase massively overstates how much the wider market grew.

Audit-adjusted funding increased from about $82.3 billion to $256.8 billion. Once we remove financings above $100 million, the increase falls to roughly 31%. Remove rounds above $50 million and it drops again to about 17%.

That distinction changes the story. Smaller and mid-sized funding did get stronger, so the current cycle has real depth underneath OpenAI and Anthropic. But there is no evidence that capital availability across the ordinary part of the market tripled.

The median round gives us another useful check. It more than doubled even though medians are far less sensitive to giant financings than total capital. The market genuinely improved; the 212% headline is simply the wrong number to use when describing the experience of most companies.

Are more Foundation Model companies getting funded now?

Yes. Foundation Model funding is reaching far more companies now, and the increase is large enough to call it a real expansion in market breadth.

We found 57 funded companies in the latest period, up from 36. Twelve companies completed multiple financings, twice as many as in the previous year, and every month contained at least one deal.

That is very different from a market where two frontier labs raise huge amounts while activity elsewhere dries up. The funding pool remains heavily concentrated by dollar value, but the number of companies successfully reaching investors has widened sharply.

The monthly rhythm shows the same thing. The current period averaged 6.25 deals per month versus 3.75 previously, while the busiest month reached 12 financings instead of seven. Foundation Model fundraising has become a much more continuous market.

Is Foundation Model funding becoming too concentrated?

Foundation Model funding is extremely concentrated today, and concentration among the three biggest rounds actually increased.

The single largest financing represented 47.5% of current-period capital, roughly in line with 48.6% one year earlier. But the top-three share climbed from 74.5% to 84.5%. OpenAI's $122 billion round, Anthropic's $65 billion Series H and Anthropic's other very large financing account for most of that jump.

This creates an unusual market shape. Dozens of companies can still raise, yet the economic gap between the frontier and everyone else keeps widening. A startup raising $200 million now counts as a substantial Foundation Model financing and still looks tiny beside the biggest labs.

As seen above, deal breadth and capital breadth tell two different stories here. Participation is spreading while financial firepower is concentrating.

Concentration measure	Current L12M	Previous L12M
Largest deal share of capital	47.5%	48.6%
Top 3 share	84.5%	74.5%
Top 5 share	88.6%	86.0%
Top 10 share	93.2%	95.1%
Deals above $50M	57	30
Deals above $100M	48	23
Capital from >$50M rounds	99.85%	99.59%
Capital from >$100M rounds	99.56%	98.96%

Are $100 million Foundation Model rounds becoming normal?

Increasingly, yes. A $100 million Foundation Model financing now looks much closer to a standard serious round than an exceptional one.

Nearly two-thirds of current deals reached at least $100 million, compared with just over half in the previous period. At the same time, the $20 million–$50 million bracket shrank from 17.8% of all deals to 8%.

The market is effectively moving up a weight class. Model training is expensive, inference is expensive, top researchers are expensive, and access to large amounts of compute can itself become a strategic constraint. Companies trying to compete near the frontier increasingly need a balance sheet that resembles an infrastructure business more than a traditional software startup.

We see the same change surprisingly early in company life. The median early-stage financing reached $80 million, up from $32 million. These days, knowing that a company is at seed or Series A tells us much less about its capital needs than it would in an ordinary SaaS market.

Are investors still backing new Foundation Model startups?

Yes. Investors are still creating new funded Foundation Model companies, and the really striking change is how much money some of those newcomers receive immediately.

First financings increased from nine to fourteen. Their share of total deals edged down from 20% to 18.7%, so existing companies still dominate activity, but startup formation itself did not weaken.

The median first financing tells the more interesting story. It rose from $40 million to about $167 million, more than quadrupling in one year. Meanwhile, the median age at first financing fell from about 0.77 years to 0.39 years.

Investors are therefore willing to write far larger checks to selected teams with very little corporate history. In this part of AI, founder pedigree, research talent, model capability and access to compute can sometimes unlock institutional-scale funding before a company has spent years building conventional operating proof.

Follow-on rounds remain the larger part of the market, but new Foundation Model startups are hardly being shut out. Some are entering with unusually large balance sheets from day one.

Are established Foundation Model companies taking most of the money?

Yes. Existing Foundation Model companies still absorb the overwhelming majority of capital, even though more new startups are getting funded.

Follow-ons represented 80% of deals in both periods. Current follow-on capital reached roughly $243 billion, and the median follow-on increased from $140 million to around $293 million.

The stable 80% share is useful because it shows that the market's expansion did not fundamentally change who dominates fundraising. New entrants increased in absolute number, while incumbents scaled their rounds even faster.

OpenAI and Anthropic are the extreme cases, but the same pattern appears further down the market. Investors are repeatedly coming back to companies that have already shown technical progress, recruited scarce teams or established a clearer position in the model ecosystem.

Which Foundation Model category is gaining the most momentum now?

World and robotics models currently have the strongest all-round funding momentum because more companies are raising and the rounds themselves are getting much bigger.

Deals in World & Robotics Models increased from seven to eighteen. Their share of the market rose from 15.6% to 24%, while their median financing jumped from $105 million to $260 million. Total funding for the category moved from roughly $1.2 billion to about $7.6 billion.

That acceleration lines up with what companies are actually building these days. Skild AI announced a $1.4 billion Series C to develop a general-purpose intelligence system that can control different robots. World Labs raised $1 billion for spatial intelligence and world models. Runway raised $315 million and said the capital would support the next generation of world models and new applications beyond its original media focus.

Nvidia's announcements offer another useful piece of context. The company has spent 2026 pushing Cosmos world models, Isaac simulation and GR00T robot foundation models into robotics, manufacturing and industrial workflows. Its partner list now includes large industrial robot companies as well as younger AI labs.

Several independent financings, higher median round sizes and more visible deployment infrastructure are moving together here. That's why the World & Robotics Models increase looks more durable than a category jump driven by one isolated megadeal.

Foundation Model category	Current deal share	Previous share	Current median round	Previous median
Language Models	29.3%	20.0%	$650M	$1.00B
World & Robotics Models	24.0%	15.6%	$260M	$105M
Video Generation	14.7%	17.8%	$86M	$56.9M
Audio & Speech	13.3%	17.8%	$46.5M	$21.5M
Image Generation	10.7%	4.4%	$210.5M	$35M
Multimodal Models	8.0%	24.4%	$426.5M	$356M
Biggest deal-share gain	Language Models	+9.3 pp	-	-
Second-biggest gain	World & Robotics	+8.4 pp	-	-

Is image-generation AI attracting serious money again?

Yes. Image-generation Foundation Models are attracting far larger rounds now, and the category has expanded well beyond the tiny funding footprint it had one year earlier.

Deal count rose from two to eight, pushing image generation from 4.4% to 10.7% of financings. The median round climbed from $35 million to $210.5 million.

The commercial context is also changing. Higgsfield, one of the companies operating across image and video creation, said in late September that it had crossed a $1 billion annualized revenue run rate, calculated from its most recent four weeks of revenue. We should treat an annualized run rate differently from trailing revenue, but the scale is still notable for a young generative-media business.

Visual models are also spreading into advertising, 3D creation, world generation and simulation. That gives investors more ways to underwrite a large round than the earlier thesis of simply generating better standalone images.

Did investors lose interest in multimodal Foundation Models?

Investor interest in multimodal Foundation Models became much narrower: deal share collapsed while a few companies continued to absorb gigantic amounts of capital.

Multimodal Models fell from 24.4% of financings to only 8%. Their typical round still grew, with the median rising from $356 million to $426.5 million, but the number of funded companies did not keep pace with other categories.

OpenAI completely changes the capital view because its $122 billion financing sits inside this segment. Looking only at dollars would make multimodal AI appear far stronger and broader than the deal count suggests.

Multimodal funding today looks concentrated rather than weak. Investors are still prepared to fund the leaders on a massive scale, while new financing activity has rotated more visibly toward language models, world models, robotics and visual systems.

Are Foundation Model investors moving toward later-stage companies?

Only a little. The bigger change today is that both young and mature Foundation Model companies need much more capital.

Early-stage deals accounted for 36% of current financings versus 42.2% previously. Later-stage share was actually slightly lower, moving from 51.1% to 49.3%, while unknown-stage transactions increased from 6.7% to 14.7%.

The round sizes are much clearer. Early-stage median funding increased from $32 million to $80 million, and later-stage median funding went from $304 million to $400 million. First financings became dramatically larger too, as discussed above.

Investors have not simply retreated into late-stage safety. They are still making aggressive early bets, but the amount of money needed to build a credible Foundation Model company has risen throughout the stack.

The rise in transactions without a conventional stage label is also worth watching. Strategic investments, unusual private placements and very large non-standard financings fit less neatly into the old Seed-to-Series-D ladder.

Funding-stage measure	Current L12M	Previous L12M
Early-stage deal share	36.0%	42.2%
Later-stage deal share	49.3%	51.1%
Unknown-stage share	14.7%	6.7%
Early-stage median round	$80M	$32M
Later-stage median round	$400M	$304M
Median first financing	$166.9M	$40M
Median follow-on	$292.6M	$140M
Overall median	$242M	$107.5M

Are Foundation Model companies raising again faster?

No. Foundation Model companies are actually taking a little longer to come back for another round, but the next check tends to be much larger.

Among companies where we could observe an earlier financing in the same market, the median gap between rounds increased from 6.2 months to eight months. The proportion raising again within 12 months fell from 70% to 62.5%.

At the same time, the median step-up from the previous financing increased from 1.67x to 1.88x. More than 44% of current observable follow-ons were over twice the size of the previous round, compared with 36.8% one year earlier.

Investors are not just recycling companies through fundraising every few months. Successful labs are returning somewhat later and getting larger pools of capital when they do.

Larger rounds may naturally lengthen the fundraising interval too. A company that has just raised several hundred million dollars has more runway before it needs another financing, even in a compute-heavy market.

Is China catching up in Foundation Model funding?

China is catching up fast in the number of funded Foundation Model companies, although the United States still dominates the market by dollars.

Chinese deal count increased from six to twenty. Across Asia-Pacific, market share by deal count rose from 17.8% to 34.7%, while North America's share fell from 71.1% to 50.7%.

Capital moved too. Chinese companies raised roughly $19.7 billion in the latest period versus around $1 billion previously, with large financings involving companies such as DeepSeek, Kling, Moonshot AI, ShengShu and VAST/Tripo.

Recent developments make this harder to dismiss as a temporary burst. Reuters reported that DeepSeek completed an approximately $7.4 billion maiden outside financing before exploring another round. Kling AI also raised close to $3 billion in a large financing backed by Chinese and international investors.

Still, North America captured almost 90% of Foundation Model capital in our calculations. OpenAI and Anthropic make the dollar gap enormous.

China has built a much deeper field of funded competitors. The US advantage remains overwhelming when we measure the amount of capital available to the very largest labs.

Is Europe keeping up with the Foundation Model funding race?

Europe is producing more funded Foundation Model companies, but it is still far behind the US and China in financing scale.

European deals increased from four to eleven. That is meaningful progress in breadth, yet Europe's share of capital stayed around the same level because a large part of regional funding remains concentrated in Mistral.

This puts Europe in an awkward position. The region has credible research talent and more funded companies than a year ago, but competing at the very top now requires financing measured in billions or even tens of billions of dollars, alongside secure access to compute.

Runway's decision to open a London research hub and commit substantial investment to its European operations shows that the region still attracts important AI research activity. It does not close the financing gap visible in the funding data.

For now, Europe looks more competitive in company formation and research than in the financial scale available to its biggest Foundation Model challengers.

Are Nvidia and other strategic investors taking over Foundation Model funding?

Strategic investors have become a normal part of Foundation Model financing, with corporate participation now showing up in roughly two-thirds of deals.

We identified a disclosed strategic investor in 50 of 75 current financings, compared with 22 of 45 one year earlier. The participation rate rose from 48.9% to 66.7%.

Nvidia appears repeatedly, but the pattern goes far beyond one semiconductor company. AMD-related entities, Amazon, Microsoft, Salesforce, Samsung, Tencent and Alibaba-linked investors also appear across the market.

Their involvement makes commercial sense. Model companies need accelerators and cloud infrastructure, while hardware and cloud companies benefit when model training and inference demand grows. Industrial businesses can help physical-AI companies reach real deployments, and media companies have a direct interest in visual and audio models.

OpenAI's recent financing illustrates how closely these relationships now overlap. The company described capital, compute and distribution as interconnected requirements and brought in Amazon, Nvidia and SoftBank as major investors. Anthropic has also signed huge compute arrangements with Amazon and SpaceX while raising enormous equity rounds.

Strategic capital is now part of how the market is built, rather than a side feature attached to occasional financings.

Do Foundation Model startups need more proof before investors write a big check?

Foundation Model investors do not appear to require more years of company history before writing large checks. The bar has shifted toward proving that a team can compete technically and operate at enormous scale.

The clearest clue is the first-financing cohort. Companies receiving their first round are younger than before, yet their median financing is dramatically larger. That is hard to reconcile with a simple story of investors becoming more conservative.

Follow-on capital shows the other side. Older and better-established companies receive the bulk of total dollars, so technical validation and market position still carry huge rewards.

We cannot responsibly give a precise percentage for how many companies had revenue, major customers or specified technical milestones at financing because those fields are not disclosed consistently enough across the market. Where public information exists, though, investors repeatedly emphasize model performance, research teams, compute access, adoption and strategic distribution.

The Foundation Model funding bar is unusual. Some teams can raise huge sums very early, but only if investors believe they have a credible path into the small group of labs capable of competing at frontier scale.

Are Foundation Model companies raising to research, or to scale businesses now?

The largest Foundation Model rounds are increasingly paying for research and commercial scale at the same time.

OpenAI's current financing language is very explicit about that combination. The company ties capital to research, products, distribution and durable access to compute. Anthropic said its $65 billion Series H would expand compute, safety and interpretability research, products and partnerships.

The same mix appears lower down the market. World Labs says its $1 billion raise will accelerate spatial intelligence and world models across creative work, robotics and scientific applications. Runway said its $315 million financing will fund new world-model pre-training while bringing those systems into additional products and industries.

These companies are still spending heavily on model development. What has changed is the commercial ambition sitting alongside that research. Large Foundation Model companies now need enough money to train the next system, serve existing users, acquire compute, hire globally and open new distribution channels all at once.

That helps explain why round sizes have moved so far beyond normal software financing.

What's the biggest mistake people make when reading Foundation Model funding numbers today?

The biggest mistake is treating the roughly $257 billion raised in the latest 12 months as proof that every part of the Foundation Model market became three times stronger.

The market did improve. Deals grew 67%, funded companies grew 58%, the median round more than doubled, first financings increased and funding happened every month. Those are meaningful changes that survive the megadeal test.

But the top of the market has entered its own financial universe. OpenAI alone closed $122 billion in committed capital, while Anthropic raised another $65 billion in its Series H. Once rounds above $50 million are stripped out, capital growth falls to about 17%.

The useful conclusion is sharper than simply saying AI funding is booming. Foundation Model financing is broadening at the company level while the amount of money needed to lead the market is pulling away at extraordinary speed.

That combination is what has changed most over the last year.

What really changed over the last 12 months?

The Foundation Model market is healthier today because many more companies are getting financed, the typical round is considerably larger and fundraising activity is more continuous than a year ago. The improvement survives once the very biggest transactions are removed, although it becomes far more modest.

Capital concentration is now the market's defining tension. More companies are getting funded while an even greater share of the money accumulates around a few frontier labs. OpenAI and Anthropic operate at a financing scale that increasingly separates the leaders from the rest of the industry.

New startups still have access to capital. First financings increased, young companies are raising sooner, and the median opening round has jumped sharply. At the same time, existing winners continue to absorb the vast majority of dollars through much larger follow-ons.

World and robotics models currently show the cleanest category momentum. Their increase comes from more deals, higher round sizes and much more total capital rather than one isolated transaction. Image-generation funding has also been re-rated upward, while multimodal funding has become much narrower and more concentrated.

The geography of Foundation Model funding is widening. China has built a far deeper financed ecosystem, Europe has added companies, and North America's deal-share dominance has fallen. The US still controls most capital because its largest labs raise vastly more money.

Strategic investors are now part of the basic funding architecture of the sector. Semiconductor companies, cloud providers, industrial groups and major technology platforms increasingly combine equity investment with compute, distribution or commercial partnerships.

The most misleading headline would simply say Foundation Model funding tripled. A more accurate reading is that the market got broader, rounds got much bigger and the frontier became dramatically more expensive to compete in.