What’s funding like in the AI inference chips market right now?

All funding deals in the AI inference chips 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 on AI chip funding, valuations and who's ahead.

Is AI inference chip funding actually healthier now?

Yes. AI inference chip funding is healthier today, but almost all of that improvement is happening among companies that have already reached serious scale.

We counted 58 financings in the latest 12 months, up from 36 in the previous period. The number of funded companies climbed from 31 to 46, and audited equity funding rose from $2.87 billion to $9.64 billion. The typical disclosed financing also changed dramatically: the median went from $26 million to $146.7 million.

The weaker part of the market sits below those large rounds. Capital from deals below $50 million fell from $363 million to $298 million. Even when we widen the cutoff to $100 million, funding declined from $774 million to $562 million.

So the market has clearly improved, but the improvement has a very specific shape. More companies are getting financed, yet the big increase in available capital starts once a company can support a nine-figure round.

Market health	Current L12M	Previous 12M	Change
Deals	58	36	+61%
Funded companies	46	31	+48%
Audited capital	$9.64B	$2.87B	+236%
Median round	$146.7M	$26.0M	+464%
Companies raising more than once	11	4	+175%
Capital below $50M rounds	$298M	$363M	-18%
Capital below $100M rounds	$562M	$774M	-27%
Zero-deal months	0	1	-

Would AI inference chip funding still look stronger without the giant rounds?

Only partly. AI inference chip activity would still look stronger without the biggest deals, but the capital boom would disappear.

Deal count rose 61%, funded-company count rose 48%, and the median round increased more than fivefold. Those changes cannot be explained by one exceptional billion-dollar financing.

The dollar picture is different. Once rounds above $100 million are removed, current-period capital is 27% lower than in the previous 12 months. Below $50 million, the decline is 18%.

That split tells us more than the headline $9.64 billion total. The market has become busier, and a much larger group of companies can now attract very large checks. Smaller and mid-sized financings have not enjoyed the same recovery.

Are just a few AI chip companies taking all the money?

No. AI inference chip funding is actually much less dependent on the top few companies than it was a year ago.

The three largest deals accounted for 59% of capital in the previous period. Their share is now only 23%. The top ten fell from 86% to 50%. With the top three deals taking a far smaller share, we look at who is getting funded, and by whom, in our AI chips market report.

At the same time, large rounds collectively became far more important. Financings above $100 million generated 94% of current capital, compared with 73% previously.

Instead of two or three companies sitting far above everyone else, there is now a fairly deep group of heavily financed companies. Thirty-two current deals crossed $100 million, versus only six in the previous period.

Funding concentration	Current L12M	Previous 12M
Largest deal share	10.4%	26.2%
Top 3 share	22.8%	59.0%
Top 5 share	31.9%	68.6%
Top 10 share	50.2%	85.5%
Deals above $50M	36	11
Deals above $100M	32	6
Capital from >$50M deals	96.9%	87.3%
Capital from >$100M deals	94.2%	73.0%

Are $100 million AI inference chip rounds normal now?

They are getting surprisingly close to normal. More than half of current AI inference chip deals were $100 million or larger.

A year earlier, fewer than one in five transactions reached that level. Today, $100 million-plus rounds represent 55% of all deals in our dataset.

The shift happened across the whole distribution. Sub-$5 million rounds fell from 13.9% of activity to 5.2%. Deals between $5 million and $20 million fell from 19.4% to 12.1%, and the $20 million to $50 million bracket dropped from 19.4% to 8.6%.

This is why the jump in the median financing is so striking. A $146.7 million median means we are no longer looking at a handful of glamorous outliers sitting on top of a conventional venture market. Nine-figure financing has moved close to the center of the market.

Are investors still funding brand-new AI chip startups?

Yes, but brand-new AI chip startups are getting a much smaller slice of the current boom than companies that have already raised before.

We found eight first financings, compared with six previously. So there is no collapse in new-company formation.

The money tells a tougher story. The median first financing slipped from $12.75 million to $10 million, while the median follow-on jumped from $44 million to $148 million. Follow-ons now represent 50 of the 58 current deals.

Investors are still willing to finance a new chip company. Once a company proves that it belongs in the market, though, the amount of capital available can increase enormously.

The biggest improvement in funding conditions is happening after the first institutional check.

Funding path	Current L12M	Previous 12M
First financings	8	6
First-financing share	13.8%	16.7%
First-financing capital	$879M	$106M
Median first financing	$10.0M	$12.75M
Follow-ons	50	30
Follow-on share	86.2%	83.3%
Follow-on capital	$8.76B	$2.76B
Median follow-on	$148M	$44M

Are investors playing it safer with AI chip companies now?

Yes. AI chip investors currently look more comfortable writing huge checks to companies with previous funding, products or customers than spreading smaller bets across the market.

Early-stage deals fell from 44.4% of activity to 34.5%, while later-stage deals rose from 36.1% to 51.7%. Money tilting toward later-stage deals changes where the openings are, and we map the biggest opportunities in our report covering AI chips.

The stage labels understate how far this has gone because some supposedly early rounds are enormous. EUCLYD, for example, raised more than €200 million in Series A financing with Samsung, Scaleup Europe Fund, Innovation Industries and other investors. That looks very different from the conventional idea of a young semiconductor Series A.

We see the same pattern in repeat fundraising. Eleven companies raised more than once during the latest 12-month window, compared with four previously.

As seen above, new startups are still getting funded. The more dramatic change is what happens once investors become convinced that a company has moved beyond the purely experimental phase.

What does an AI chip company need to prove to raise a huge round today?

For the biggest AI chip rounds today, a good architecture and a slide deck are no longer enough. The strongest companies are showing working hardware, customers, production progress or real deployment.

Etched is one of the clearest examples. The company said recently that it was validating rack-scale hardware with customers, had started production and had more than $1 billion in customer demand or contracts. It then raised $700 million from a group led by Jane Street after Jane Street tested the hardware and installed an Etched rack in its own data center. A customer testing the hardware before leading a $700 million round shows real traction; we show who is pulling ahead in our AI chips funding report.

Positron had already deployed more than 50 Atlas racks at Oracle before announcing its latest large financing. D-Robotics said cumulative shipments of its Sunrise chips had passed eight million units, while multiple robotics customers using its platform were already moving into mass production.

Public disclosure is too inconsistent to turn those examples into a clean market-wide milestone ratio. Still, they show what the top end looks like now: investors increasingly have something concrete to inspect before committing several hundred million dollars.

Are AI chip companies raising money to build products or to scale them?

At the top end of AI inference chips, companies are increasingly raising to manufacture, deploy and expand products that already exist.

SambaNova's $1 billion first close is tied to additional capacity, product development and enterprise or sovereign deployments. Positron is funding its Asimov silicon and Titan systems after putting Atlas hardware into the field. D-Robotics plans to expand its Sunrise chip family and the software platform around its robotics ecosystem.

Etched goes even further. The company describes work across chips, racks, factories, supply chains and fleet software as it tries to ramp production.

This helps explain why round sizes are climbing so quickly. Building the chip is only one part of the bill these days. A serious inference hardware company may also need advanced packaging, memory, boards, racks, cooling, software, manufacturing capacity and deployment teams.

Which part of the AI inference chip market is gaining the most momentum?

Custom inference ASICs are the clearest winner right now.

The category went from one deal in the previous period to six in the current one. Its share of deals rose from 2.8% to 10.3%, while its share of capital climbed from 0.5% to 17.5%. The current median round is around $300 million.

Etched gets most of the attention, but the category shift is bigger than one company. More investors are now prepared to finance silicon designed around specific inference workloads rather than assuming that general-purpose accelerators will own every part of the market.

Data-center inference is still far larger overall, and edge AI still produces many more deals. Custom ASICs stand out because breadth and check size improved at the same time.

Is data-center AI inference losing momentum?

No. Data-center AI inference is still the largest part of the market by far, although it no longer absorbs almost all of the money.

The category produced 24 current deals and 59% of audited capital. Its median financing rose from about $51 million to more than $230 million.

Its relative capital share fell sharply from 83%, but that happened because custom ASICs, edge AI and automotive chips started attracting serious amounts of capital too.

A year ago, the market was much more dependent on one part of the inference stack. Today, money is spreading across several technical approaches while data-center companies themselves are raising larger rounds.

Category	Current deals	Current capital share	Previous deals	Previous capital share	Current median
Data-center inference	24	59.1%	17	83.1%	$230.5M
Edge AI accelerators	23	19.1%	15	15.9%	$57.0M
Custom inference ASICs	6	17.5%	1	0.5%	$300M
Embedded / industrial	3	0.2%	3	0.6%	$7.9M
Automotive AI chips	2	4.1%	0	0%	$198.5M
Largest category	Data center	-	Data center	-	-
Fastest rotation	Custom ASICs	-	-	-	-

Is edge AI funding finally getting serious?

Yes. Edge AI funding is getting much larger even though edge chips are not taking a bigger share of total deal activity.

Edge accelerators produced 23 current deals versus 15 previously, but their share of transactions stayed around 40% because the whole market grew quickly.

The bigger change is financing size. The median edge round rose from about $20.5 million to $57 million. Median edge rounds nearly tripling suggests investors back that technology, and we follow which technologies are gaining traction in our AI chips funding report.

Recent financings show how broad the end markets have become. D-Robotics raised $400 million around robotics infrastructure. SiMa.ai raised $150 million to push further into humanoids, automotive systems and drones. Acrab raised $130 million around an edge-agent AI compute platform.

Edge AI is moving well beyond tiny embedded chips. Robotics, vehicles and other physical AI systems are giving investors reasons to fund much larger platforms.

Is China now dominating AI inference chip funding?

China dominates AI inference chip deal count, but North America still dominates the size of the checks.

China produced 23 current transactions, up from ten in the previous period. Across Asia-Pacific, the region's share of deals climbed from 52.8% to 62.1%.

North America moved the other way on volume, accounting for 25.9% of deals. Yet it still captured almost half of all current capital, and its median financing reached roughly $275 million.

There are two different races going on. China is producing more funded companies and more transactions. North America is concentrating enormous amounts of money into fewer companies.

Europe adds a third model. Its deal share fell slightly, yet median round size reached about $225.5 million thanks to financings such as EUCLYD and OLIX.

Why are European AI chip rounds suddenly so big?

European AI chip funding is getting bigger because a few companies have crossed into genuine infrastructure-scale financing.

Europe accounted for only 10.3% of current deals, down from 13.9%, so there has been no explosion in startup count. Its capital share nevertheless rose from 3.6% to 12.8%.

EUCLYD raised more than €200 million in Series A funding around an architecture combining programmable compute, memory design and data-center systems. OLIX raised $312 million for specialized inference silicon and rack-scale systems.

Those rounds show that European AI chip companies can now attract several hundred million dollars when the pitch moves beyond a standalone processor and into the economics of AI infrastructure, memory, efficiency and sovereign compute.

For now, Europe's momentum is much more about deal size than breadth.

Are corporate investors becoming more important in AI chips?

Absolutely. Corporate and strategic investors are much more involved in AI inference chip financings today.

Strategic participation appeared in 48.3% of current deals, up from 25% in the previous period. Almost all of those current strategic-backed deals were follow-ons. Corporate investors in nearly half of current deals is a big shift; we cover which investors are making the biggest bets in our report on AI chips.

The names make the pattern easy to understand. Samsung backed EUCLYD. SK Hynix invested in Etched before Jane Street later led another round. Dell Technologies Capital participated in SiMa.ai. Meituan backed D-Robotics. Arm appeared in multiple current financings.

These companies sit close to the actual hardware ecosystem: memory, processor IP, servers, cloud infrastructure, vehicles, robotics or end customers.

Strategic capital is showing up most heavily once integration and deployment become central problems. The median current strategic-backed round reached about $240.5 million.

Is AI inference chip funding controlled by the same few investors?

No. The investor pool has become much wider.

We identified 95 disclosed lead or strategic investors in the current period, compared with 45 previously. Only a small number appeared repeatedly.

The current boom therefore looks quite different from one driven by a closed circle of specialist venture firms. Large asset managers, corporate investors, semiconductor companies, sovereign capital and traditional venture investors are all appearing in financings.

There is one limitation worth keeping explicit: public investor lists are incomplete for many rounds. We can measure disclosed participation, but we cannot claim to have a complete syndicate map.

Even with that constraint, the direction is clear. Investor breadth has increased sharply.

Why do so many young AI chip companies already look late-stage?

AI inference chip companies can look surprisingly young on paper because several of today's startups inherited technology, teams or businesses that existed before the legal company did.

The median company age at financing fell from about 4.2 years to 3.1 years, even while later-stage deals became more common.

D-Robotics is a good example. The company was created from Horizon Robotics' AIoT and robotics activities, so its incorporation date does not represent the age of all the technology behind it. Sunrise similarly came out of SenseTime's chip operation.

That makes raw company age less useful in this market than it would be for a typical software startup.

When we assess maturity today, customer deployment, hardware status, manufacturing progress and previous funding history usually tell us more than the date on the incorporation documents.

Are AI chip companies raising money faster than before?

We do not have enough evidence to say that AI chip companies generally come back to market faster, even though repeat fundraising has clearly increased.

The current period contains many more companies with multiple financings, and Etched provides an extreme example: a $300 million financing was followed shortly afterward by another $700 million round. With rounds landing this close together, we follow how much money went in, quarter by quarter, in our AI chips funding report.

But a proper market-wide answer requires the previous financing date for every follow-on company. Those dates cannot be established consistently enough from public disclosures to calculate a reliable median interval.

So we can confidently say repeat financing is more common inside the latest 12-month window. We cannot claim that the typical company now raises again six months faster, nine months faster or any other precise amount.

That distinction is worth keeping because semiconductor rounds often include extensions, additional closes and tranches that can make superficial timing comparisons misleading.

What would be the most misleading headline about AI inference chip funding right now?

“AI inference chip funding tripled” would be accurate and still miss the most interesting part of the story.

Current audited capital is more than three times the previous level, but smaller rounds did not rise with it. At the same time, the market became much broader in companies and much less dependent on its three biggest deals.

The real shift is the emergence of a large upper tier. Dozens of companies can now raise sums that would have looked extraordinary one year earlier.

And the companies attracting that money increasingly resemble infrastructure businesses. Etched is working across silicon, racks, manufacturing and software. Positron sells complete inference systems. SambaNova combines proprietary processors with full-stack enterprise deployments. D-Robotics links chips to an entire robotics development platform.

These days, the central funding question is less about whether investors believe alternative AI inference hardware can exist. The tougher question is which companies can prove they are ready to manufacture, integrate and deploy it at scale.

What really changed over the last 12 months?

AI inference chip funding became healthier, but mainly for companies that have already shown enough progress to raise at scale. Deal count and company count both increased strongly, while smaller-dollar capital declined.

The biggest shift is the new depth of the nine-figure market. Large rounds are no longer confined to two or three obvious winners, and top-three concentration dropped sharply even as $100 million-plus deals took a much bigger share of total capital.

Existing companies benefited much more than brand-new entrants. First financings continue, but the financing gap between formation and follow-on stages has widened dramatically.

Custom inference ASICs have the clearest category momentum, while data-center inference remains the largest segment and edge AI is attracting substantially larger checks around robotics, automotive and physical AI.

Investors also appear to be asking for more concrete proof before funding the biggest rounds. Working hardware, customer deployment, production readiness and ecosystem relationships show up repeatedly among the companies raising hundreds of millions of dollars.

The market is moving further into commercialization and industrial scale. Today, the strongest inference-chip companies are being financed to build factories, systems, racks, supply chains and deployment infrastructure around their silicon.

The headline increase in total funding is real, but it is easy to read it incorrectly. AI inference chips have become a much bigger funding market, while the bar for receiving meaningful capital has also moved higher.