What are the latest funding trends in AI inference chips?

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

Is AI inference-chip funding really booming right now?

AI inference-chip funding is clearly having a breakout year: more companies are raising, deals are happening much more often, and the typical round has become several times larger.

We count 48 financings so far, compared with 29 over the same part of both 2025 and 2024. The number of funded companies has climbed from 26 to 38, so the increase cannot be explained by a few startups repeatedly coming back for money. Audited disclosed capital has risen from $1.95 billion last year to $7.55 billion.

The jump in the median round is just as striking. It has gone from $25.1 million to $146.9 million in one year. That tells us the change reaches far beyond a handful of spectacular deals.

The pace is also unusually high against complete prior years. Current deal count has already passed the roughly 39 financings we recorded across all of 2025 and the 36 recorded in 2024. We should not extrapolate that pace mechanically through year-end, but as of today the market has already moved into a different funding range.

Funding measure	2026 YTD	2025 comparable	2024 comparable	2026 vs 2025
Deals	48	29	29	+65.5%
Unique funded companies	38	26	26	+46.2%
Audited disclosed capital	$7.55B	$1.95B	$1.68B	+286.5%
Median disclosed round	$146.9M	$25.1M	$29.0M	+486%
Average disclosed round	$188.9M	$81.4M	$76.4M	+132%
Largest audited round	$1.00B	$750M	$640M	+33.3%
Companies raising more than once	10	3	3	+233%
Full prior-year deal count	-	≈39	36	Already exceeded

Is one giant AI chip deal making the funding numbers look better than they really are?

No. The AI inference-chip funding jump survives the single-deal test easily, although almost all of the extra dollars are still coming from very large rounds.

The largest audited financing, SambaNova's $1 billion first close, accounts for only 13.2% of current capital. A year ago, the largest deal represented 38.4%. The top five financings now account for 38.1% of the market, versus 71.6% in the comparable 2025 period.

The concentration has moved from a few exceptional transactions to a whole class of nine-figure rounds. We count 26 financings above $100 million, compared with only five on a like-for-like audited basis last year.

But the outlier test also exposes the weak part of the market. Remove rounds above $100 million and only $466 million remains, below the $554 million recorded last year and the $570 million recorded two years ago. Remove deals above $50 million and current capital drops to $259 million, again below 2025.

We also corrected Etched's previously reported $500 million January financing back into 2025. Etched itself later disclosed that the round actually closed in December 2025, and TechCrunch reported the same timing. Even after making that less flattering adjustment, the 2026 surge remains very large.

Outlier test	2026 YTD	2025 comparable	2024 comparable
Audited disclosed capital	$7.55B	$1.95B	$1.68B
Capital excluding >$50M rounds	$258.6M	$317.8M	$166.7M
Capital excluding >$100M rounds	$466.3M	$554.4M	$570.2M
Rounds above $50M	29	8	10
Rounds above $100M	26	5	5
Largest-deal share	13.2%	38.4%	38.1%
Top-three share	27.8%	58.3%	52.6%
Top-five share	38.1%	71.6%	66.1%

Are more AI inference-chip startups getting funded, or are investors mostly backing the same winners again?

Both are happening, but most of the serious money is going back into companies that have already raised before.

Thirty-eight companies have secured financing so far, up from 26 during the same stretch last year. That is real expansion. At the same time, 43 of the 48 current financings are follow-ons, while only five are classified as first financings in our financing-history check.

Repeat fundraising has accelerated too. Ten companies raised more than once during the current period, compared with three last year. Etched raised $300 million and then $700 million within weeks. D-Robotics raised $150 million before a $400 million Series C. OLIX followed a $220 million financing with another $312 million round. Quadric added a second Series C close after its initial $30 million financing. The same companies coming back for a second round within weeks is a pattern; we cover who is getting funded, and by whom, in our report on AI chips.

First financings have moved much less: five today versus four last year. New companies are still entering the funding market, but investors are putting far more energy into scaling companies that have already cleared previous technical and financing hurdles.

Financing-history measure	2026	2025	2024
First financings	5	4	1
First-financing share	10.4%	13.8%	3.4%
Follow-on financings	43	25	28
Follow-on share	89.6%	86.2%	96.6%
Companies raising more than once	10	3	3
Disclosed first-financing capital	$6.6M	$99M	$1.1M
Median disclosed first financing	$3.3M	$23.3M	$1.1M
Median follow-on financing	$148M	$34.1M	$38M

Why have AI inference-chip rounds suddenly become so big?

AI inference-chip rounds have become huge because many of the companies attracting money now need to manufacture hardware, build complete systems and deploy them with real customers.

The median disclosed financing is around $147 million today, almost six times last year's $25 million. That change is too large to explain with one or two spectacular deals.

Look at what companies are actually funding. Fractile said its $220 million round would accelerate getting its first chips and systems into customers' hands. Rebellions raised $400 million while expanding production of Rebel100 and rolling out RebelRack and RebelPOD infrastructure. OLIX raised $312 million as it moved toward its first racks. Neurophos, after raising $110 million, has since expanded its engineering operation and says it is preparing for commercialization.

Etched is further along the same curve. When it emerged publicly, the company said it had working silicon, more than $1 billion in signed customer contracts and production ramping to meet demand. Its later $300 million financing was explicitly tied to production and customer deployments.

These are expensive steps. Once an AI chip company moves beyond architecture and prototypes, it has to deal with tape-outs, fabrication, packaging, memory, boards, software, racks, manufacturing capacity and customer installations. The funding numbers increasingly look like those of an industrial hardware business rather than a conventional software startup market.

Is the smaller AI inference-chip funding market actually getting weaker?

Yes. Funding below the very large rounds looks noticeably softer today, which is easy to miss when the headline total is above $7 billion.

Aggregate capital below $100 million is down about 16% from last year's comparable period. Below $50 million, it is down roughly 19%.

The deal-size distribution has moved strongly upward as well. Only two current financings are below $5 million, four fall between $5 million and $20 million, five between $20 million and $50 million, and three between $50 million and $100 million. Meanwhile, more than half of all current deals sit above $100 million.

For founders, this changes how we should read the boom. A company trying to raise $20 million or $40 million is operating in a very different funding market from a proven chip company asking investors for $250 million to scale production. The record aggregate number mostly describes the second group.

Are early-stage AI inference-chip startups having a harder time raising?

Early-stage AI inference-chip startups are losing ground inside the funding mix, even though seed and Series A financing has certainly not disappeared.

Seed and Series A together represent 31.3% of current deals, down from 41.4% last year. At the other end, later-stage rounds have risen to 52.1% of activity from 37.9%.

Series C is where the shift becomes obvious. We count ten Series C financings today, compared with five during the same part of 2025 and three in 2024. Growth equity has also appeared as a meaningful category with four current transactions, while neither comparable period had any.

Some unusually large early rounds prevent us from saying early-stage capital has simply dried up. Neurophos raised $110 million at Series A, OLIX raised $220 million at Series A, and EUCLYD recently announced a Series A above €200 million. Those exceptional transactions make the pattern pretty clear too: investors will still fund young companies aggressively when they think the technical proposition is unusually strong.

Stage	2026 deals	2026 share	2025 share	2024 share
Seed	7	14.6%	17.2%	10.3%
Series A	8	16.7%	24.1%	27.6%
Series B	7	14.6%	17.2%	27.6%
Series C	10	20.8%	17.2%	10.3%
Series D+	4	8.3%	3.4%	10.3%
Growth equity	4	8.3%	0%	0%
Unknown	8	16.7%	20.7%	13.8%

Which part of the AI inference-chip market is gaining investor attention fastest?

Custom inference ASICs are the clearest category gaining share right now.

Only one custom-ASIC financing appeared during the comparable 2025 period. We count five this year. The category's deal share has moved from 3.4% to 10.4%, while its share of disclosed capital has jumped by roughly 15 percentage points to 15.7%.

Etched is obviously the largest contributor, with separate $300 million and $700 million rounds. But specialized architectures are attracting money beyond Etched. Taalas raised $169 million for model-specific silicon, and other companies are pursuing chips designed around particular inference workloads rather than general-purpose acceleration. Chips designed for particular inference workloads are drawing real money; we follow which technologies are gaining traction in our report covering AI chips.

The current median disclosed round for custom ASIC companies is around $235 million, which is remarkable for a category whose median company age is only about 2.6 years.

Investors appear increasingly willing to make a concentrated bet on specialization: if inference becomes a dominant AI workload, a chip optimized around a narrower set of model structures, memory patterns or latency requirements can potentially trade flexibility for much better economics.

Are data-center inference chips still where most of the money goes?

Yes. Data-center inference is still the biggest pool of AI inference-chip capital by a wide margin.

The category accounts for 18 current financings and roughly $4.6 billion, or about 61% of audited capital. SambaNova, Rebellions, OLIX, Fractile, Positron, EVAS Intelligence and several other large rounds keep data-center hardware firmly at the center of the market.

What has changed is its relative dominance. Data-center inference now represents 37.5% of deals, down from 41.4% last year, and its capital share has also fallen even though the absolute dollars have exploded.

Custom ASICs and large edge financings simply grew faster around it. Data-center inference remains the biggest funding market, but investors currently have more credible alternatives than they did a year ago.

Is edge AI chip funding booming too?

Edge AI chip funding is getting much bigger per company, while its share of overall deal activity has actually slipped.

Edge accelerators account for 20 current deals, the largest category by transaction count. Yet their share fell from 48.3% last year to 41.7% today.

The companies that do get funded are raising much more. D-Robotics closed $150 million and later $400 million. Axelera AI raised more than $250 million. SiMa.ai recently raised $150 million to expand in humanoids, automotive and drones. NanoCoreChip raised roughly $149 million, while Acrab raised $130 million. Edge chipmakers drawing rounds of $130 million to $400 million show where money sees openings, and we look at the biggest opportunities in our AI chips funding report.

Axelera also offers a useful commercialization clue. When it announced its latest financing, the company said it was already shipping to its 500th customer across areas including industrial manufacturing, robotics, defense, retail and agritech.

So edge AI has not suddenly produced a larger share of deals. The bigger change is that several edge-chip companies are now being financed like serious scale-up businesses.

Which AI inference-chip deals best show how different this year has become?

The deals that define the current AI inference-chip market are no longer isolated $500 million exceptions. There is now a dense group of companies raising several hundred million dollars.

SambaNova's $1 billion first close sits at the top, followed by Etched's $700 million financing. Then come D-Robotics and Rebellions at $400 million each, Positron's $375 million confirmed Series C component, OLIX at $312 million, Etched again at $300 million and EVAS Intelligence at roughly the same level.

That sequence is more informative than simply naming the biggest transaction. A few years ago, one $600 million round could dominate the market statistics. These days, a $200 million financing can sit surprisingly far down the annual leaderboard.

Company / financing	Audited amount used	Stage	Main category
SambaNova Systems	$1.00B	Series D+	Data-center inference
Etched	$700M	Series D+	Custom inference ASIC
D-Robotics	$400M	Series C	Edge accelerators
Rebellions	$400M	Growth equity	Data-center inference
Positron	$375M confirmed	Series C	Data-center inference
OLIX	$312M	Series B	Data-center inference
Etched	$300M	Series C	Custom inference ASIC
EVAS Intelligence	≈$299M	Series B	Data-center inference
Axelera AI	>$250M	Series C	Edge accelerators

Is Asia taking over AI inference-chip startup funding?

Asia-Pacific has become the center of AI inference-chip deal activity, although North America still attracts much more money per transaction.

Nearly 65% of current financings are in Asia-Pacific, compared with 55% last year and 38% two years ago. The region's pipeline now includes Rebellions, D-Robotics, HyperAccel, NanoCoreChip, DEEPX, EVAS Intelligence, Sunrise and a long tail of Chinese and Korean chip startups.

North America tells a different story. It has only 23% of current deals but receives 45% of disclosed capital, helped by huge financings for SambaNova, Etched and Positron.

Europe has also changed in an interesting way. Its share of deals remains modest, but its current median disclosed round is roughly $226 million. OLIX, Fractile, Axelera AI and EUCLYD have all raised amounts that would have looked exceptional for European AI hardware very recently.

As of now, Asia has the broadest startup pipeline, North America remains the most capital-heavy market, and Europe is producing fewer but much larger financings.

Why are corporate and strategic investors showing up so often in AI inference chips?

Strategic investors are back in a big way because AI inference hardware increasingly touches the companies that make memory, servers, devices, robots and the services that consume inference compute.

Strategic investors appear in 43.8% of current deals, up from 27.6% during the comparable 2025 period. That is close to 2024 levels, so we see it more as a rebound than an entirely new phenomenon.

The relationships are also becoming more concrete. SK Hynix invested in Etched as the company scaled its hardware. Jane Street subsequently led another Etched financing after becoming a customer. Samsung co-led EUCLYD's financing. Meituan participated in D-Robotics. Intel Capital appears in SambaNova, alongside a broader product relationship between Intel and SambaNova. A trading firm, a memory maker and a delivery platform are all writing checks, and we follow which investors are making the biggest bets in our report on AI chips.

The typical strategic-backed round has become much larger too, with a median around $231 million in our calculations.

This increasingly looks like industrial ecosystem financing. Some investors can potentially provide much more than capital: manufacturing relationships, memory, semiconductor IP, distribution, cloud infrastructure, devices or a direct workload for the chip.

Are more investors willing to lead big AI chip rounds now?

Yes. The pool of investors willing to lead AI inference-chip financings has expanded dramatically.

We count 54 distinct disclosed leads or co-leads so far, compared with 26 over the comparable 2025 period. That is faster growth than the number of funded companies.

And the lead list is unusually dispersed. Innovation Industries is the only disclosed lead or co-lead appearing more than once in the current period.

General Atlantic led SambaNova's $1 billion first close. Jane Street led Etched's $700 million financing after becoming a customer. Sequoia led Etched's earlier $300 million Series C. Accel, Factorial Funds and Founders Fund led Fractile's $220 million financing. Hummingbird led OLIX's earlier round, while its later $312 million financing brought in additional large investors.

A year ago, much of the sector could still look like specialist semiconductor investing. Today, the investor base reaches growth funds, major generalist venture firms, crossover investors, strategic corporates, trading firms and state-backed capital.

Why are such young AI chip companies already raising hundreds of millions?

AI inference-chip startups are reaching huge rounds unusually quickly because investors are compressing the traditional hardware funding timeline for companies that hit strong technical or commercial milestones early.

The median company receiving financing is about 3.2 years old, down from 4.5 years last year. At the same time, funding has shifted toward later stages. Those two movements together are unusual.

OLIX is only about two years old and has already raised $220 million followed by $312 million. Etched was founded in 2022 and has moved through several enormous financings after producing working silicon and signing more than $1 billion in customer contracts. Positron, founded in 2023, went from a $230 million Series B to a large Series C within months. D-Robotics became a standalone company in 2024 and has already completed two major rounds this year.

Investors are not necessarily waiting five or six years for a chip company to mature before providing growth-scale capital anymore. A working architecture, manufacturing path, credible customer testing or early deployments can push a young company up the funding ladder very quickly.

The risk, of course, is that fundraising milestones can move faster than manufacturing and commercial reality. We will only know later which of these accelerated funding trajectories were justified.

Is today's AI inference-chip funding market actually healthy for startups?

Today's AI inference-chip funding market is extremely strong for proven scale-up candidates and much less generous for everyone else.

The good news is real: the number of funded companies has expanded, lead-investor participation has broadened across many institutions, strategic money has returned, and large rounds now span several chip architectures and regions.

The harder part is visible lower down the funding stack. As seen above, sub-$100 million capital has not kept pace, early-stage deals represent a smaller share of activity, and first financings have barely increased.

That creates a demanding market. Investors are paying a lot more attention to inference hardware, but they appear especially willing to pay for evidence: functioning silicon, design wins, manufacturing progress, systems ready for customers, or actual deployments.

Neurophos' recent expansion after its $110 million Series A, Rebellions shipping rack-scale products, Axelera reporting hundreds of customers and Etched disclosing signed contracts all fit that pattern. The biggest checks are increasingly going to companies that can tell a commercialization story rather than only a chip-design story.

What should we watch next in AI inference-chip funding?

The next big question for AI inference-chip funding is whether today's heavily financed companies turn their capital into repeatable customer deployments quickly enough.

We already know investors are prepared to fund the sector. The uncertainty now sits on the commercial side. Money is no longer the open question, so we turn to the biggest unsolved problems in our AI chips funding report.

Etched has disclosed more than $1 billion in signed customer contracts and says it is ramping production. Rebellions is shipping rack-scale systems and recently announced another infrastructure partnership in Japan. Neurophos has expanded its team and Silicon Valley presence to push toward commercialization. SiMa.ai is directing new capital toward physical-AI applications such as humanoids, vehicles and drones.

SambaNova is also worth following closely. Its latest financing came alongside customer and product announcements around enterprise inference, and the company has continued to argue that inference economics increasingly depend on reducing power and infrastructure costs.

If companies like these turn large funding rounds into repeat purchases and sizable deployments, the current wave will look like the emergence of a new commercial chip cohort. If deployments lag badly behind the capital raised, today's funding numbers will eventually look much more speculative.

For now, fundraising itself is no longer the question. Execution is.

What is actually changing in AI inference-chip funding?

AI inference-chip funding is stronger on almost every measure that reflects market activity, but the improvement becomes far more selective once we look at where the money actually lands.

The biggest shift is the normalization of huge scale-up rounds. At the same time, custom inference ASICs are gaining share, Asia is producing a much denser startup pipeline, strategic investors are returning, and younger companies are reaching large rounds faster than before.

The weak spot remains the lower half of the financing market. A record year for $200 million and $500 million rounds does not automatically make this an easy year for a startup trying to raise its first institutional round.

What is changing	Our read today
Overall funding market	Clearly stronger in both deals and funded companies
Size of successful rounds	The biggest structural change; nine-figure rounds have become common
Smaller and mid-sized funding	Weaker than the headline boom suggests
Investor preference	Increasingly tilted toward companies with prior financing and tangible proof points
Hottest category shift	Custom inference ASICs are gaining both deal and capital share
Geographic shift	Asia leads in breadth; North America still leads in capital intensity
Strongest thing to watch next	Whether large rounds turn into repeatable deployments and customer demand