What are the latest funding trends in AI chips?

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 really booming right now?

AI chip funding is booming in 2026, and the jump is big enough to show up in deal count, company count and round size at the same time.

We counted 22 qualifying financings through September 21, compared with eight over the same period in 2025 and 12 in 2024. Sixteen different companies raised money this year, up from seven a year earlier. Capital increased even faster, from at least $994 million to roughly $7.14 billion. Sixteen funded companies against seven a year ago is a big move, and we trace what changed this past year in our AI chips report.

The median round gives us the clearest sense of how different the market looks today. It has gone from $26 million in the comparable 2025 period to $263.5 million in 2026. A tenfold move in the median cannot be created by one spectacular announcement at the top of the table.

There are some unusual financing structures in the total, so we took the conservative route where necessary. Positron, for example, announced a structure that could reach $875 million, but only $375 million was firmly quantifiable for our main calculation. Several Asian financings also represent approved commitments or minimum amounts rather than perfectly clean cash-close figures. None of those adjustments comes close to changing the overall picture.

Metric	2026 YTD	2025 YTD	2024 YTD
Deals	22	8	12
Unique companies	16	7	11
Total capital	≈$7.14B	≥$994M	≥$1.29B
Median round	$263.5M	$26.0M	$35.2M
Average quantifiable round	$324.7M	$141.9M	$117.0M
Largest counted round	$1.0B	$750M	$640M
Companies raising more than once	6	1	1

Is one giant AI chip deal making 2026 look better than it really is?

No. Large rounds are driving the AI chip funding boom, but the money is spread across many companies rather than resting on one freak deal.

The contrast with 2025 is striking. Groq's $750 million round represented roughly three quarters of the capital raised during the comparable period last year. In 2026, the largest counted financing contributes only about 14% of the total.

Even the three biggest rounds account for around 38% of current capital, against more than 93% in 2025. Cerebras raised $1 billion, SambaNova completed a $1 billion first close, Etched announced financings of $300 million and $700 million, Rebellions raised $400 million, and Positron contributes $375 million under our conservative treatment.

The market is therefore much less exposed to the fortunes of a single company than it was last year.

But once we remove the big rounds, the picture changes fast. Capital from deals of $50 million or less totals only $53.5 million, below the comparable 2025 figure of at least $67 million. Even below $100 million, 2026 is only about 13% ahead of last year and remains below 2024.

So the boom is real. It is also overwhelmingly an upper-end boom.

Test	2026 YTD	2025 YTD	2024 YTD
Total capital	≈$7.14B	≥$994M	≥$1.29B
Capital excluding >$50M rounds	$53.5M	≥$67.0M	$153.2M
Capital excluding >$100M rounds	$134.2M	≥$118.6M	$153.2M
Top-1 share of capital	14.0%	75.5%	49.7%
Top-3 share of capital	37.8%	93.3%	78.8%
Top-5 share of capital	52.2%	98.2%	91.1%
Rounds above $100M	18	2	4

Have $100 million AI chip rounds basically become normal?

Pretty much. Nine-figure AI chip rounds have gone from exceptional events to the dominant type of financing in our 2026 sample.

Eighteen of the 22 deals this year are worth at least $100 million. That is nearly 82% of all transactions and about 98% of the capital we counted.

Only four financings sit below $100 million. There are two between $5 million and $20 million, one between $20 million and $50 million, and one between $50 million and $100 million. We found no qualifying round below $5 million.

The comparison with the previous two years makes the shift hard to miss. Only two of eight comparable 2025 deals reached $100 million, while four of 12 did so in 2024.

A nine-figure financing is no longer confined to the largest one or two AI chip companies. Among companies that have reached the scale-up phase, it has become a common outcome.

Why are AI chip startups raising so much money now?

AI chip companies are raising much larger rounds because more of them have reached the expensive part of the journey: manufacturing, systems production, data-center deployment and customer rollout.

Etched gives us one of the clearest examples. The company announced $300 million to accelerate production and customer deployments, alongside an 80,000-square-foot facility. Weeks later, it announced another $700 million as Jane Street became its first disclosed customer and began deploying an Etched rack in its own data center. Etched also said it had secured more than $1 billion in customer contracts.

FuriosaAI has followed a similar path. LG AI Research adopted its RNGD accelerator after testing it on EXAONE models, and Furiosa moved RNGD into mass production in early 2026. The company says 4,000 accelerators had been manufactured and shipped by TSMC as it moved from customer evaluation toward commercial deployment. More recently, Furiosa has expanded through Samsung SDS and opened a Singapore hub while rolling RNGD out internationally.

Rebellions raised $400 million as it moved into scaled Rebel100 production, overseas expansion and preparation for a future IPO. SambaNova's $1 billion Series F first close came as the company pushed its RDU systems further into enterprise inference, with JPMorganChase among the customers highlighted in its financing announcement.

Once a chip company reaches this point, capital requirements climb quickly. Silicon production has to be financed alongside packaging, memory, servers, software, deployment teams, inventory and data-center infrastructure. The fundraising numbers are starting to resemble infrastructure financing because the businesses themselves are becoming more infrastructure-heavy.

Is AI inference where most of the AI chip money is going?

Yes. Inference is currently the main center of gravity in AI chip funding, both in deal activity and in the commercial milestones surrounding the largest rounds.

We classify 16 of this year's 22 financings as Inference Accelerators, representing approximately $5.09 billion. The category actually loses market share versus 2025 because new categories have grown around it, yet its absolute deal count rises from seven to 16.

The company evidence lines up unusually well with that funding pattern. Etched is building rack-scale inference clusters. FuriosaAI's RNGD is designed around LLM inference and is now being deployed with enterprise customers. Rebellions sells production-scale inference infrastructure rather than only standalone silicon. SambaNova presented fast enterprise inference as a central use case when it announced its latest financing. Rebellions and SambaNova sell whole systems rather than bare chips, so we compare which business models are winning in our report covering AI chips.

AI Compute Chips is also becoming important. Three transactions contribute roughly $1.58 billion, with a $500 million median round. That is 22% of capital from only 14% of deals, so the category is attracting unusually large checks.

We should not read too much into every category name, though. AI-chip companies increasingly combine silicon, systems and software, so an apparent move between labels can sometimes reflect classification rather than a genuine technological shift.

Category	2026 deals	2026 capital	Median round
Inference Accelerators	16	≈$5.09B	$265M
AI Compute Chips	3	≈$1.58B	$500M
Data Center GPUs	2	$305M	$152.5M
AI ASIC Platforms	1	$169M	$169M
Inference deal share	72.7%
AI Compute deal share	13.6%
AI Compute capital share	22.1%

Are early-stage AI chip startups getting squeezed out?

Early-stage AI chip funding looks weak next to the huge scale-up market developing above it.

Seed and Series A account for only three of the 22 financings we counted this year, compared with five of eight in the same period last year. Their share of deal activity has fallen from 62.5% to 13.6%.

Seed is especially thin. We found only one Seed financing in the current period, Agrani Labs' $8 million round. There were three comparable Seed deals in 2025.

Agrani itself shows why we avoid pretending these classifications are perfectly clean. A 2025 SEC Form D predates the public announcement of the $8 million Seed round, putting the financing close to the boundary between the two periods.

Still, the broader result does not depend on that one case. Series D+ alone attracted roughly $3.58 billion this year, while Seed attracted $8 million. Series B rounds have a $230 million median and Series C rounds a $300 million median.

There may be young chip companies raising quietly or announcing rounds later, so we would not turn a small sample into a sweeping claim about every semiconductor startup. The visible 2026 boom is clearly happening much further along the company lifecycle.

Stage	Deals	Deal share	Capital	Median
Seed	1	4.5%	$8M	$8M
Series A	2	9.1%	$330M	$165M
Series B	5	22.7%	≈$1.29B	$230M
Series C	3	13.6%	$822M	$300M
Series D+	5	22.7%	≈$3.58B	$700M
Growth Equity	3	13.6%	$560M	$147M
Unknown	3	13.6%	$547M	$169M

Are investors backing new AI chip startups or funding the same companies again?

Investors are overwhelmingly putting more money into AI chip companies that have already raised before.

Twenty-one of the 22 financings we classified this year are follow-ons. The number of first financings stays at one, exactly where it was in each of the previous comparable periods.

That becomes more revealing when we look at what happened around it. Total deal activity rose from eight transactions to 22, while first financings did not increase at all. Six companies also raised more than once during 2026, compared with one repeat fundraiser in each of the earlier periods.

Company age points in the same direction. The median funded company is about four years old today, versus 2.5 years in the 2025 sample. Series D+ companies have a median age of roughly nine years, while Growth Equity companies sit around six.

We should treat the first-financing classification as directional because private funding histories are not always perfectly observable. Even with that limitation, the gap is too large to ignore. Investors are currently committing far more money to companies that have already cleared several technical and financing hurdles.

Are AI chip investors becoming more selective?

AI chip investors look extremely aggressive once a company proves itself, while the bottom of the funding funnel remains much tighter.

That combination explains several things that might otherwise seem contradictory. We have more funded companies, more transactions and much larger rounds, yet Seed activity is lower and first financings are virtually absent.

Capital below $50 million is also weaker than last year. Meanwhile, companies that reach the other end of the funnel can raise $300 million, $500 million or even $1 billion. Late-stage money piling in while smaller rounds weaken is why we look at which investors are making the biggest bets in our AI chips funding report.

Commercial evidence is increasingly visible around those companies. Etched has working silicon, a delivered rack and signed customer contracts. FuriosaAI has moved from sampling to mass production and enterprise deployment. Rebellions now sells rack-scale infrastructure and is preparing for an eventual IPO. SambaNova is selling inference infrastructure into large enterprises.

Investors are paying heavily for proof right now. Once a company demonstrates that its silicon can become a deployable product, the next phase can be funded on a scale that was rare only a year or two ago.

Why are corporate and strategic investors showing up in so many AI chip rounds?

Strategic investors are becoming much more common because AI chip companies increasingly need help with manufacturing, memory, infrastructure and deployment as well as money.

After excluding cases where the supposed strategic relationship was too vague to defend, we find strategic participation in 13 of 22 current deals, or 59.1%. The comparable share was 37.5% in 2025 and 41.7% in 2024. The median strategic-backed financing has also climbed from about $125 million last year to $350 million today.

The relationships are often directly connected to the product. AMD participated in Cerebras' $1 billion Series H. SK Hynix participated in Etched's $300 million round as Etched built a system around memory-heavy inference workloads. Jane Street went even further: it led Etched's subsequent $700 million financing while becoming the company's first disclosed customer and deploying its hardware.

FuriosaAI's ecosystem shows the same pattern from another angle. TSMC manufactures RNGD, the accelerator uses HBM memory, LG AI Research has deployed the product, and Samsung SDS is now making RNGD available through an enterprise AI infrastructure service.

At this stage, a strategic relationship can solve an actual operating bottleneck. In some of the biggest rounds, that can make corporate capital more useful than a purely financial check.

Are new investors moving into AI chip funding too?

Yes. The group willing to lead major AI chip rounds has widened considerably this year.

We found 27 unique disclosed leads or co-leads in 2026, compared with 13 in the comparable 2025 period. The disclosure rate barely changed, so the jump cannot be explained simply by better reporting.

Current leads include General Atlantic, Tiger Global, Sequoia, Accel, Founders Fund, Vista Equity Partners, NEA and the Korea National Growth Fund. Jane Street is the only disclosed lead or co-lead appearing in more than one current transaction in our analysis.

That is a useful detail. The extra capital is not coming from a tiny club repeatedly syndicating every major deal. A broader set of growth investors, venture firms, sovereign or public vehicles, corporates and customers is now willing to finance AI compute.

Some firms that appeared in 2025 are absent this year, but after only a few quarters we would not call that a retreat. Funding schedules are too uneven. The more defensible takeaway is simply that the pool of institutions prepared to write very large AI-chip checks has grown.

Is Asia catching up with North America in AI chip funding?

Asia is gaining ground quickly in AI chip funding, although North America still attracts most of the capital.

North America accounts for ten current deals and roughly $4.73 billion, or about two thirds of all capital. Its position remains dominant, especially with Cerebras, SambaNova, Etched, MatX and Positron among the companies raising large rounds.

What has changed is the gap. North America represented around 84% of comparable capital in 2025. Asia-Pacific now accounts for nine deals and approximately $1.66 billion, with its deal share up almost 16 percentage points and its capital share up more than ten points.

South Korea has become particularly interesting because private financing and industrial policy are now moving together. Rebellions' $400 million pre-IPO round included the Korea National Growth Fund. The country's Financial Services Commission has also approved roughly KRW800 billion of direct equity investment around FuriosaAI, explicitly linking the support to next-generation AI chips using HBM4 and the development of a domestic AI-semiconductor ecosystem.

China contributes through companies including AIGCIC, Moffett AI and Sunrise, so the Asian increase is broader than one Korean government program.

Europe remains much smaller in transaction count, with three current financings, but its measured capital rises to roughly $752 million. For now, the geography story is gradual diversification rather than a change in global leadership.

Has 2026 already beaten the previous full years for AI chip funding?

Yes. AI chip funding through the current period has already passed both 2025 and 2024 full-year totals in our research, without needing any annualized forecast.

We counted 22 deals through September 21. There were 15 in all of 2025 and 16 in all of 2024. Deal counts already beat the full years of 2025 and 2024; we follow how much went in, quarter by quarter, in our AI chips report.

Capital tells an even stronger story. The roughly $7.14 billion counted so far represents about 221% of the approximately $3.23 billion recorded for the whole of 2025 and around 326% of the at-least-$2.19-billion total for 2024.

We would not extrapolate the current pace mechanically into the final months of the year. Venture financing can arrive in bursts, particularly in a market where individual rounds now reach hundreds of millions of dollars.

The comparison already says enough: 2026 has become a much larger funding year than either of the previous two years in the market we track.

Is the AI chip market moving from experimentation to industrial scale?

Yes. The clearest change in AI chip funding today is the move from financing chip development toward financing production, systems and commercial rollout.

Several independent parts of the analysis point there. Later-stage transactions dominate the deal mix. Follow-on rounds account for almost every financing. The typical check is now in the hundreds of millions. Strategic investors participate in a majority of current deals.

The company milestones make the shift much more concrete. Etched is building manufacturing and deployment capacity while delivering hardware to customers. FuriosaAI has entered mass production and is expanding commercial deployment internationally. Rebellions is scaling Rebel100 and selling complete rack and cluster products. SambaNova is pushing RDU infrastructure into enterprise inference. Cerebras has continued raising large amounts while expanding AI infrastructure and data-center capacity.

The biggest rounds are increasingly attached to companies that have crossed from proving the chip into figuring out how to manufacture, deploy and support it at scale.

That is why 2026 feels so different from the earlier funding cycle. The expensive part of semiconductor commercialization is now showing up directly in financing data.

What is the biggest weakness in the AI chip funding boom?

The weakest part of today's AI chip funding market is still the creation and financing of the next startup cohort.

Record capital can easily hide that point. We count roughly $7.14 billion overall, yet only $53.5 million remains once rounds above $50 million are excluded. There is one Seed deal and one first financing. Those numbers have barely participated in the boom.

There are also a few places where apparent precision would be misleading. Positron's financing contains a potential additional component that we do not treat as committed capital. Some rounds in China and Korea involve staged, approved or minimum amounts. Company-category boundaries are also getting fuzzier as vendors move from individual chips toward complete systems.

Those issues deserve caution, but they do not change the main findings. Deal volume is much higher, the median round is dramatically larger, later-stage financing dominates, and large rounds are spread across far more companies than last year.

The open question is what comes next. If Seed and first-financing activity eventually rise too, the current scale-up cycle could begin feeding a broader startup wave. We do not see that part yet.

What is actually changing in AI chip funding?

AI chip funding is much stronger in 2026, with more transactions, more funded companies and a roughly tenfold increase in the median round versus the comparable 2025 period.

The capital boom is concentrated at the top of the market. Eighteen of 22 deals exceed $100 million, while funding below $50 million is actually lower than last year.

Large rounds are also much more widely distributed. Last year's market was heavily distorted by one financing, whereas the largest current deal represents only around 14% of 2026 capital.

Investors are concentrating money on companies that already have funding histories, working products and increasingly visible commercial traction. Follow-on rounds dominate the market, while Seed and first-financing activity remain thin.

Inference is currently where the deepest commercial momentum appears. Etched, FuriosaAI, Rebellions and SambaNova all connect large funding events with production, deployments or enterprise inference activity.

Strategic capital is becoming more important as the companies move toward manufacturing and deployment. Memory suppliers, semiconductor companies, infrastructure groups and customers now appear directly in many large rounds.

Geographically, North America still leads by a wide margin, but Asia is becoming much more important. South Korea in particular is combining private capital with direct industrial-policy support for domestic AI-semiconductor companies.

The strongest current interpretation is a scale-up and industrialization cycle. The part to watch most carefully is the early-stage pipeline, because the extraordinary growth at the top has not yet produced a comparable increase in new startup financing.