AI Chip Funding: A Quarter-by-Quarter Analysis

All funding deals in the AI chip 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 chip funding actually getting stronger right now?

AI chip funding is stronger today for companies already reaching scale, while funding for smaller and newly financed startups remains thin.

We tracked at least $10 million in Q2 2025, $2.06 billion in Q3, $887.5 million in Q4, $2.99 billion in Q1 2026 and $1.17 billion in Q2. Q2 2025 only contains one qualifying financing, so the useful comparison really starts in Q3.

From there, funding swings much more violently than deal count. Capital dropped 57% in Q4, jumped 237% in Q1 and fell 61% in Q2. The number of deals moved from six to five, then nine and seven. Check size explains far more of the movement than the number of companies raising money.

The smaller end tells a very different story. Once we remove rounds above $50 million, quarterly capital comes to $33.5 million in Q3, $10 million in Q4, $8 million in Q1 and $45.5 million in Q2. Huge checks are getting written, but that strength still has very little depth underneath it.

Period	Deals	Total funding	Funding excluding >$50M	Median round
Q2 2025	1	≥$10M	≥$10M	≥$10M
Q3 2025	6	$2,060.1M	$33.5M	$88.3M
Q4 2025	5	$887.5M	$10.0M	$71.0M
Q1 2026	9	$2,987.0M	$8.0M	$230.0M
Q2 2026	7	$1,172.2M	$45.5M	$147.0M
Q3 → Q4	–16.7%	–56.9%	–70.1%	–19.6%
Q4 → Q1	+80.0%	+236.6%	–20.0%	+223.9%
Q1 → Q2	–22.2%	–60.8%	+468.8%	–36.1%

Was Q1 2026 really a broad AI chip funding boom?

Q1 2026 was a huge quarter for AI chip scale-ups, but the boom barely reached the smaller end of the market.

We counted nine financings and $2.99 billion in announced capital. Eight of those nine rounds were above $100 million. Cerebras raised $1 billion, MatX $500 million, Rebellions $400 million, SambaNova $350 million and Positron AI $230 million. OLIX, Taalas and Neurophos also crossed $100 million.

Cerebras clearly helped lift the total, although the median round of $230 million shows that Q1 ran much deeper than one giant transaction. A $500 million Series B for MatX and a $230 million Series B for Positron would have been major events in almost any quarter.

The weak spot was startup formation. Agrani Labs' $8 million Seed round was the only financing below $100 million and the only first financing we identified. Nearly every dollar went into businesses that investors had already backed before.

Calling Q1 an AI chip funding boom works for the scale-up market. It says much less about how easy it was for a new chip startup to raise its first serious round.

Are new AI chip startups still getting funded?

New AI chip startups are getting squeezed out of the recent funding mix, with investors putting far more money behind companies that already have financing history and products further along.

The clearest change comes after Q4 2025. First financings represented 60% of Q4 deals, then fell to 11% in Q1 2026 and disappeared from Q2. Follow-on rounds accounted for all seven Q2 financings.

Stage moved in the same direction. Later-stage deals rose from 20% of the Q4 total to 56% in Q1 and 86% in Q2. FuriosaAI, Fractile, Moffett AI, HyperAccel and AIGCIC were all raising additional capital to develop, manufacture or scale businesses already in motion. Money going to companies already in motion changes who is getting funded, and by whom, which we follow in our AI chips funding report.

Company age gives us an interesting nuance. Median age went from two years in Q4 to four years in Q1 and stayed at four in Q2. Investors are backing more mature financing stories without necessarily moving toward decade-old semiconductor companies. Several of today's heavily funded challengers are still very young businesses.

Our financing-history check is deliberately conservative, so we would not treat every "first financing" label as a complete reconstruction of a company's private capital history. The fall across two consecutive quarters is large enough that the direction is still hard to miss.

Why did early-stage AI chip funding suddenly look so strong in Q4 2025?

Q4 2025 produced a genuine burst of new AI chip companies, although one extraordinary Seed round made the early-stage dollar figures look much bigger than the underlying wave.

Three of the quarter's five deals were first financings. Majestic Labs raised $71 million at Series A, SUNMMIO announced more than approximately $56.5 million at Seed and Mastiska raised $10 million at Seed.

Then came Unconventional AI. Its $475 million Seed financing represented roughly 54% of all Q4 capital. With that one round included, early-stage deals captured 69% of quarterly funding.

A $475 million Seed bears little resemblance to the economics of an ordinary company-formation round. The useful part of Q4 is the rise in the number of newly financed companies. That wave faded quickly: one first financing remained in Q1 and none appeared in Q2.

Is inference hardware taking over AI chip funding?

Inference hardware currently has the strongest momentum in AI chip funding, and its rise now shows up in both the number of deals and the size of the checks.

Inference accelerators were already common in Q3 2025, with five of six financings, yet they attracted only 47% of capital because Cerebras alone raised $1.1 billion in the broader AI Compute Chips category.

The money then caught up with the deal activity. Median inference financing increased from $51.6 million in Q3 to $230 million in Q1 2026. By Q2, inference accelerators represented six of seven deals and approximately $1.09 billion of funding. Funding that swings this much from Q3 to Q2 is easier to follow quarter by quarter, as we show in our AI chips funding report.

Several different companies drove that quarter. FuriosaAI had the largest announced financing at approximately $532 million, followed by Fractile at $220 million, Moffett AI at around $147 million, Sunrise at more than approximately $147 million and HyperAccel with an approximately $32.6 million commitment.

Investors are funding a whole group of inference specialists now, rather than one company carrying the category by itself.

Quarter	Category	Deals	Capital share	Median round
Q3 2025	Inference Accelerators	5	46.6%	$51.6M
Q3 2025	AI Compute Chips	1	53.4%	$1.10B
Q4 2025	Inference Accelerators	3	38.5%	$56.5M
Q4 2025	AI Compute Chips	1	53.5%	$475M
Q1 2026	Inference Accelerators	5	43.9%	$230M
Q1 2026	AI Compute Chips	2	50.2%	$750M
Q2 2026	Inference Accelerators	6	93.1%	$147M
Q2 2026	AI Compute Chips	1	6.9%	$80.7M

Why are inference chip rounds getting so much bigger?

Inference chip rounds are getting bigger because these companies are now raising for production, systems and customer deployments rather than mainly for early chip development.

FuriosaAI is a good example. When the company announced its $125 million Series C in 2025, it said the money would help scale RNGD production and develop its next-generation chip as commercial demand grew. The financing came after LG AI Research selected RNGD to run EXAONE inference workloads.

d-Matrix raised $275 million later that year around the expansion of its Corsair inference platform and large deployments. Positron's $230 million Series B in Q1 2026 brought in Arm and customer Jump Trading after Jump had already deployed and tested its Atlas systems. SambaNova paired a $350 million financing with new inference hardware and a multi-year Intel collaboration.

The fresh announcements after our June 30 cutoff make the pattern even harder to dismiss. SambaNova subsequently announced a $1 billion first close, OLIX raised $312 million, Etched announced $700 million after shipping its first rack to Jane Street, and Positron announced an $875 million Series C following a 50-plus-rack deployment at Oracle Cloud Infrastructure. We excluded those later financings from the quarterly calculations, but they are useful when asking whether the scale-up trend is still alive today.

Chip startups at this stage need money for wafers, memory, packaging, servers, software, supply chains and production capacity. Once customers start deploying the hardware, a financing round can quickly move from tens of millions into the hundreds of millions.

Did broader AI compute chips suddenly lose momentum?

Broader AI Compute Chips had a very weak Q2 2026 compared with inference specialists, although one quarter is too short to call the category finished.

AI Compute Chips repeatedly absorbed an outsized share of funding earlier in the period. Cerebras contributed $1.1 billion in Q3 2025. Unconventional AI raised $475 million in Q4. Cerebras and MatX together brought in $1.5 billion during Q1 2026.

The category then fell to one Q2 transaction, AIGCIC at approximately $80.7 million, representing 6.9% of quarterly capital.

That drop is enormous, but this category has always been lumpy. A Cerebras-sized financing can change the shape of an entire quarter. At the same time, inference has now built enough breadth across FuriosaAI, Fractile, Moffett AI, Sunrise and HyperAccel that the Q2 rotation deserves to be taken seriously.

For now, capital is moving heavily toward inference. Whether that becomes a lasting decline in every other type of AI processor is still open.

Is AI chip funding moving from North America to Asia?

AI chip funding swung sharply toward Asia-Pacific in Q2 2026, but we need another quarter or two before calling that a lasting geographic shift.

North America accounted for 93% of capital in Q3 2025, 93% again in Q4 and 79% in Q1 2026. We then recorded no North American deal in Q2. North America going from 93% of capital to no deals shows how much can move in a year, as our AI chips report covers.

Asia-Pacific moved to six of seven Q2 deals and 81% of capital. FuriosaAI and HyperAccel contributed South Korean financings, while Moffett AI, Sunrise and AIGCIC added Chinese activity. The shift involved several companies rather than depending entirely on FuriosaAI's large financing.

The timing also fits a broader change in Asian AI infrastructure investment. China continues to fund domestic accelerator alternatives, while South Korea has begun using large state-backed financing programs alongside private semiconductor capital.

Q2 gives us one clear geographic break after three quarters dominated by North America. The rotation was real for that quarter; persistence remains open.

Why is South Korea suddenly so important in AI chip funding?

South Korea has become one of the most interesting AI chip financing hubs because government-backed capital is now arriving at the same time local chip companies need much larger scale-up rounds.

Rebellions raised $400 million in Q1 2026. Korea's Financial Services Commission then described KRW250 billion of direct strategic-fund investment within a broader KRW600 billion financing package aimed at mass production and next-generation NPU development.

FuriosaAI followed with an approximately KRW800 billion financing plan approved in Q2. The Korean regulator linked the program to next-generation AI semiconductors using HBM4 and to the expansion of the domestic AI semiconductor ecosystem. Around KRW370 billion was approved through the Advanced Strategic Industry Fund as part of the broader package.

Private capital was already active before these programs became so large. FuriosaAI had closed $125 million in 2025, and Rebellions raised $250 million that same year with investors including Arm, Samsung Ventures and Pegatron VC.

What changed in 2026 is the amount of industrial money available once these companies reached mass-production and global-expansion stages. South Korea can now combine venture capital, strategic corporate investors and public industrial financing around the same group of AI chip companies.

Are strategic investors changing AI chip funding?

Strategic investors are becoming much more closely tied to customers, suppliers and deployment partners, even though their share of total deals still jumps around from quarter to quarter.

We found strategic investors in 33% of Q3 deals, 60% in Q4, 56% in Q1 and 43% in Q2. Those percentages do not form a clean upward curve.

The names behind the rounds tell us more. MatX brought in Alchip and Marvell, companies directly connected to semiconductor production and infrastructure. Arm invested in Positron, while Jump Trading joined the same round after becoming a customer. Intel Capital participated in SambaNova around a wider infrastructure collaboration. FuriosaAI's Q2 financing involved Naver and Cisco, and HyperAccel brought in KRAFTON. Strategic names keep showing up beside chip startups, so we follow which investors are making the biggest bets in our report covering AI chips.

Lead investors, meanwhile, remain fairly diverse. We counted ten unique disclosed leads in Q3, eight in Q4 and thirteen in Q1, with very little repetition. Q2 had weaker disclosure, so the apparent drop to five disclosed leads tells us little about whether the investor pool actually narrowed.

The change worth focusing on is the role strategic investors play. They increasingly arrive when a chip company needs help manufacturing hardware, integrating systems, accessing infrastructure or turning technical progress into real deployments.

Did AI chip funding actually weaken in Q2 2026?

Q2 2026 cooled sharply from Q1 in headline dollars, while the mix of funding became even more concentrated around established inference companies in Asia.

Total capital fell 61% from $2.99 billion to $1.17 billion, deal count dropped from nine to seven and the median round moved from $230 million to $147 million. Q1 had set an unusually high bar, so part of that decline reflects the comparison itself.

Underneath the headline total, smaller financing improved. Capital from rounds of $50 million or less rose from $8 million in Q1 to $45.5 million in Q2. That still leaves the market heavily skewed toward big checks, but it gives us little evidence of a collapse across every deal size.

The bigger changes came from who received the money. Inference accelerators jumped from 56% to 86% of deals. Asia-Pacific moved from two of nine Q1 financings to six of seven in Q2. Every Q2 deal was a follow-on, and later-stage financings dominated the quarter.

Q2 looks more like a sharp rotation after an exceptional Q1 than the start of a straightforward funding downturn.

Which AI chip funding trends still look real today?

The durable AI chip funding story today is increasingly about scale: more follow-on capital, more later-stage rounds and much bigger financing for inference hardware.

The follow-on shift has survived two consecutive quarter changes. First financings fell from 60% in Q4 to 11% in Q1 and then zero in Q2. Later-stage deals moved from 20% to 56% and then 86%.

Inference also looks increasingly durable. Its deal count was already strong in 2025, then round sizes rose and the category took most Q2 capital. The large inference financings announced after our cutoff give us fresh evidence that investors are still willing to put hundreds of millions behind companies approaching real deployment.

Other changes remain much less settled. Asia's Q2 surge needs more time. Strategic-investor participation moves up and down. Q4's burst of newly financed companies vanished quickly. Q1's almost $3 billion total has already proved too volatile to use as a normal quarterly run rate.

As seen above, the absence of new first financings in Q2 may be the most important weakness to watch. A market can finance today's winners extremely well and still create problems for itself if too few new companies make it into the pipeline behind them.

Change we tracked	How we read it today
More follow-on financing	Persistent
More later-stage financing	Persistent
Inference gaining deal breadth	Persistent
Inference attracting much larger rounds	Increasingly persistent
Round sizes rising every quarter	Already reversed once
Asia-Pacific taking share from North America	Real but still emerging
Strategic-investor participation steadily rising	Too volatile
Strategic investors moving closer to deployment and supply chains	Increasingly visible
More new companies entering the market	Q4 spike
Small and midsize funding broadening	Still weak
Q1's ~$3B pace becoming normal	Unsupported so far

What is the funding trend actually telling us?

AI chip funding has become much better at financing scale than at financing breadth. Investors can currently mobilize $200 million, $500 million or $1 billion around companies they believe can reach production, secure customers or become strategically important.

The market also looks more mature than it did in late 2025. Follow-ons and later-stage rounds now dominate recent activity, while the flow of newly financed companies has weakened sharply.

Inference is where the momentum looks strongest. Investors were already funding many inference companies in 2025, and they are now giving those companies much larger checks. Recent post-period financings at SambaNova, Etched, OLIX and Positron suggest that appetite has continued beyond the five quarters used for our core calculations. Inference companies are getting much larger checks, and we look at who the likely future winners are in our report on AI chips.

The geographic picture is moving faster than we expected, especially in South Korea. Q2's Asia-heavy mix coincided with large Korean industrial-financing programs, giving companies such as Rebellions and FuriosaAI access to a mix of private, strategic and public capital.

The weak spot remains the bottom of the pipeline. The market has proved that it can finance companies once they reach the scale-up stage. Whether enough new AI chip startups can reach that point is now a more interesting question than whether investors can find another billion dollars for an established contender.