AI Chip Funding Tracker (54 deals)

Our funding tracker covering the AI chip market

Below is the table listing all the deals. You may find our methodology at the end of this page.

And, if you want to gain 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 a $13 billion market today?

Yes. AI chip funding is running at roughly a $13 billion scale today, although almost all of that money sits in very large financings.

We tracked 52 distinct financings across 31 AI chip companies from January 2024 through the end of our research window. Together, they account for $12.99 billion of recorded capital. The median round is $136 million and the average reaches $259.7 million.

The median is the number worth paying attention to. A $136 million midpoint is huge for venture capital, so a couple of billion-dollar rounds cannot explain the size of this market on their own.

Large deals still dominate the dollars. Of the 50 financings where we could establish a usable amount, 29 were above $100 million. They captured 94.8% of known capital. Remove those rounds and only about $676 million remains.

Some headline amounts also need to be read carefully. Positron AI's recently announced $875 million financing combines a $375 million Series C with a Series C-1 of up to $500 million. SambaNova described its $1 billion Series F as a first close. Several Asian deals involve policy capital, staged commitments or currency conversions. We keep the publicly announced values where they are usable, but the $12.99 billion total is best read as a solid picture of market scale rather than cash settled to the last dollar.

How many AI chip companies are actually getting funded?

The AI chip funding market currently revolves around only 31 companies, and a relatively small repeat-fundraising group absorbs most of the money.

Seventeen companies appear once during the period. Fourteen come back for at least another financing, including five that appear three or more times.

Those repeat fundraisers collect $10.85 billion, or 83.6% of the entire market. Companies appearing only once account for the remaining $2.13 billion.

So 52 deals give an exaggerated impression of how many independent companies sit underneath the headline activity. The financial core is much tighter. Cerebras, Groq, SambaNova, Positron AI, Etched, Rebellions, FuriosaAI and MatX all return to investors as their programs get more expensive.

A normal company in this market is also far from lightly capitalized. Median cumulative financing during our window reaches $207 million per company.

Which AI chip startups have actually raised the most money?

A handful of AI chip companies hold an enormous share of the market, with Cerebras alone raising $2.10 billion during our research window.

The five largest companies by cumulative capital account for 55.0% of all recorded funding. Extend that to the top ten and the share reaches 80.4%. With five companies holding more than half the money, we cover who the top companies are today in our report covering AI chips.

Company-level concentration is noticeably stronger than deal-level concentration because the same successful platforms keep raising. Cerebras closes two billion-dollar rounds. Positron appears four times. Etched goes from $120 million to $300 million and then $700 million. Rebellions appears four times across Series B, its extension, Series C and a pre-IPO round.

Looking at companies gives a much clearer picture of where investor conviction sits than simply sorting individual announcements from largest to smallest.

Company	Financings in period	Capital recorded
Cerebras	2	$2.10B
Groq	2	$1.39B
SambaNova	2	$1.35B
Positron AI	4	$1.18B
Etched	3	$1.12B
Rebellions	4	$789M
Tenstorrent	1	$693M+
FuriosaAI	3	$669.9M
MatX	3	$620M
OLIX	2	$532M

What can an AI chip startup realistically expect to raise?

AI chip startups that reach institutional funding are raising unusually large rounds these days: the median financing in our research is $136 million.

There are no sub-$5 million financings in the audited cohort. Eight deals sit between $5 million and $20 million, and another eight between $20 million and $50 million. Those 16 financings together represent only around 2.4% of recorded capital.

Even the starting point is expensive. First financings have a median size of $25.5 million. Seed rounds overall have a median of $23.5 million.

Once companies move into serious silicon development, the numbers climb quickly. Neurophos raised $110 million at Series A. OLIX raised $220 million at Series A. MatX reached $500 million at Series B. Fractile raised $220 million at Series B.

Those examples sit far above normal software-startup economics. Designing silicon, paying specialized engineering teams, funding tape-outs and preparing hardware for customers bring substantial costs before large-scale commercial revenue arrives.

Which part of the AI chip market is getting the most funding now?

Inference accelerators are currently the clear center of AI chip funding, accounting for 35 financings and $8.13 billion of capital.

We found 19 companies in that category, covering several very different technical approaches. Groq uses its LPU architecture. Etched focuses on transformer-specific silicon. Fractile is attacking inference through a new memory and compute architecture. Positron has built memory-centric accelerator systems. Rebellions and FuriosaAI are developing data-center NPUs. Neurophos, Arago and Lumai are taking photonic routes.

Funding is spread across a real group of competing companies rather than resting on one extraordinary transaction. The largest single financing in the category contributes only around one eighth of its capital.

Recent deals make the direction especially clear. Positron's latest financing is tied to its next-generation inference silicon and production ramp. Etched raised $700 million after getting its first rack into Jane Street. OLIX raised $312 million as it moved toward first rack deliveries. Investors are spending heavily on companies that claim they can make model serving faster, cheaper or less power-hungry.

AI chip category	Deals	Companies	Capital	Median round
Inference Accelerators	35	19	$8.13B	$136M
AI Compute Chips	7	5	$3.94B	$500M
Server AI Processors	3	3	$341M	$71M
Data Center GPUs	3	3	$305M	$152.5M
AI ASIC Platforms	3	2	$242.2M	$38M
Training Accelerators	1	1	$25M	$25M

Why are broader AI compute startups raising even bigger rounds?

Broad AI compute companies are raising some of the biggest checks in the entire chip market, with a $500 million median financing in that category.

There are only seven AI Compute Chip financings in our research, yet they account for $3.94 billion, or 30.4% of all capital. That is a huge amount of money for five companies.

Cerebras is the obvious heavyweight, with a $1.1 billion Series G followed by a $1 billion Series H. Tenstorrent raised more than $693 million. MatX went from a $25 million Seed to roughly $95 million at Series A and then $500 million at Series B. Unconventional AI arrived with a $475 million Seed round. Billion-dollar rounds for a single company stand out, so we follow where the biggest rounds are going in our AI chips funding report.

Cerebras also shows where this money goes once a company reaches scale. Its Series G announcement specifically mentioned processor design, packaging, system design, supercomputers, U.S. manufacturing capacity and data-center capacity.

A company building a broader compute platform can end up financing several layers of infrastructure at the same time. That pushes funding needs well beyond the cost of designing the processor itself.

Why do AI chip startups need hundreds of millions of dollars?

AI chip startups increasingly need hundreds of millions because getting working silicon is only one part of the job; investors are now financing manufacturing, systems, racks, supply chains and customer deployments as well.

Fractile's $220 million Series B gives a good view of the middle of that process. The company said the money would help bring its first chips and systems to customers while it works on the memory-bandwidth, latency and cost problems around frontier-model inference.

Positron is further along. Its latest financing is linked to the tape-out of its Asimov silicon and the production ramp of Titan systems after more than 50 racks of its Atlas generation had already been deployed. Positron already has dozens of racks deployed, and we compare who is ahead, and who is falling behind, in our AI chips market report.

Etched gives us an even more concrete example. The company raised $300 million to scale production and customer deployment, then raised another $700 million shortly afterward. Jane Street had tested the hardware and received the first rack. Etched said the next phase includes factories, global supply chains and fleet software.

OLIX's latest round follows the same pattern. The company says its $312 million Series B funds the path toward first customer deliveries as well as the manufacturing and supply-chain commitments needed to scale frontier inference hardware.

Once a design starts working, the spending problem gets bigger. Successful hardware has to be produced, packaged, installed, supported and delivered in quantities that matter to real customers.

How far along are AI chip startups when the big checks arrive?

The biggest AI chip checks usually arrive several years into company building, with the median company aged four years when it raises.

Seed and Series A financings happen at a median company age of about two years. Series B moves to four years, Series C to 5.5 years and Series D+ to nine years.

The pattern is unusually clean. Early capital often arrives while teams are still developing architecture and first-generation hardware. By B and C, investors increasingly expect working silicon, systems development or visible customer validation. The oldest companies in our sample are funding full-scale commercialization.

MatX shows how quickly the curve can move for an unusually well-backed team. Founded in 2022, it raised $25 million in 2024, roughly $95 million later that year and $500 million in 2026. Cerebras sits at the other end of the timeline: founded in 2015, it was already a mature hardware platform when it raised another $2.1 billion across two rounds in our period.

The founding-year analysis is slightly less clean than the funding totals because corporate histories can get messy after spin-outs or reorganizations. We excluded Sunrise from the age calculations for that reason rather than force conflicting founding dates into one answer.

How fast do AI chip rounds grow from Seed to late stage?

AI chip round sizes climb very quickly with maturity, moving from a $23.5 million Seed median to $696.5 million at Series D and beyond.

Series A has an $83 million median. Series B reaches $172 million, followed by $262.5 million at Series C. By the time a company is raising at D+ in this market, a several-hundred-million-dollar financing feels normal.

The capital split makes the same point from another angle. Seed and Series A represent 41.3% of classified financings but only 14.1% of the corresponding capital. Later stages absorb the overwhelming majority of the dollars.

The expensive part of AI-chip company building comes after the first institutional backing. Every successful technical milestone opens another, larger spending requirement.

Stage	Deals	Capital recorded	Median round
Seed	9	$646.5M	$23.5M
Series A	10	$1.00B	$83M
Series B	9	$1.47B	$172M
Series C	6	$1.97B	$262.5M
Series D+	9	$6.02B	$696.5M
Growth Equity	3	$559.9M	$147M
Unknown	6	$1.32B	$124.9M

Where does AI chip funding get really hard?

The clearest pressure point in AI chip venture funding appears between Series B and Series C, where the number of financings falls from nine to six.

The early ladder looks surprisingly intact. We count nine Seed rounds, ten Series A rounds and nine Series B rounds. Series C is where the visible cohort starts thinning.

The nine Series D+ financings should be read differently because several belong to an older generation of companies that entered this period already mature. Cerebras, Groq, SambaNova and Tenstorrent did not work their way from Seed to late stage inside our research window.

What changes around Series C is the quality of proof investors can demand. A promising architecture and a credible engineering team can raise early capital. Later rounds increasingly sit alongside production plans, customer deployments, working systems or strategic partners.

For founders, that makes the B-to-C jump particularly important. The company is starting to be judged as a future semiconductor business rather than primarily as an ambitious technical project.

Are investors still backing brand-new AI chip startups?

Yes, but new AI chip companies are a small part of the funding market right now: only six of the 52 financings we tracked are classified as first financings.

The other 46 are follow-ons. First financings account for $600.5 million, while companies that had already raised before absorb more than $12 billion during the period. With nearly all the money going to companies that already raised, we follow who is getting funded, and by whom, in our AI chips funding report.

Unconventional AI is the spectacular exception. The company launched with a $475 million Seed financing. Agrani Labs raised $8 million, Arago $26 million and Mastiska $10 million. MatX also entered our window with a $25 million first financing before moving rapidly into much larger rounds.

There is another useful clue here: we found no financing below $5 million. Database coverage may miss some angel or pre-seed activity, so we would not claim that tiny AI-chip rounds never happen. But the $23.5 million Seed median makes it hard to dismiss the pattern as a database artifact.

These days, the visible funding market is far more focused on pushing already-capitalized companies toward production than on financing hundreds of cheap experiments.

How important are strategic investors in AI chip funding?

Strategic investors are now deeply involved in AI chip funding, appearing in 27 of the 52 financings we tracked.

Rounds with an identified strategic investor have a $297 million median size. The median falls to $37.5 million when no strategic is identified. We would not treat that gap as a clean causal effect because later-stage companies both raise more money and attract more corporates, but the association is striking.

Strategic participation also becomes much more common as companies mature. It appears in two of nine Seed rounds and seven of nine Series D+ rounds.

The investor names explain why. Arm can bring processor IP and deep semiconductor relationships. SK Hynix sits in the memory supply chain. Cisco sells the networking infrastructure around AI clusters. Samsung touches memory, foundry and other semiconductor layers. Intel is both an investor and commercial collaborator in several parts of the AI infrastructure market.

Sometimes the customer becomes part of the financing story too. Jane Street led Etched's $700 million round after testing the company's hardware and receiving its first rack. Jump Trading participated in Positron after deploying its systems.

For hardware companies approaching production, strategic capital can come with something almost as useful as money: a place in the supply chain or a credible path to deployment.

Which investors keep coming back to AI chip deals?

There is surprisingly little repeat leadership among AI chip investors: 43 financings disclose a lead, but those rounds involve 69 different lead-investor names.

Only four leads appear more than once in our review.

Atreides Management is the standout, appearing on three financings with combined round sizes above $2 billion. Jane Street appears twice across $1.2 billion of rounds. Valor Equity Partners appears twice across $926.6 million, while Korea Investment Partners leads two much smaller rounds totaling around $70 million. Jane Street leading twice and Valor doing the same; we follow which investors are making the biggest bets in our report on AI chips.

Those figures refer to the total size of rounds they led or co-led, not each investor's personal check.

The company side is much more concentrated. Capital keeps returning to the same hardware platforms, while the lead-investor base remains dispersed. Large rounds can bring together crossover funds, specialist VCs, corporate investors, sovereign or policy capital and major financial institutions.

There is no obvious five-firm venture club controlling this market today. The recurring nodes are easier to see among the companies being financed.

Where is AI chip funding concentrated geographically?

North America currently dominates AI chip funding by dollars, even though it accounts for less than half of the deals.

North America represents 25 of the 52 financings and $9.85 billion, or 75.8% of total capital. The United States alone contributes around $8.95 billion across 22 financings.

Asia-Pacific is much more active than its capital share might suggest. We found 18 financings there, worth $2.30 billion. South Korea is particularly concentrated: nine financings come from only three companies, led by Rebellions and FuriosaAI. China contributes eight deals across six companies.

Europe records six financings and about $803 million. The United Kingdom produces most of that amount through Fractile, OLIX and Lumai.

The geography splits in two ways. The United States has both a relatively broad company base and several of the largest late-stage platforms. South Korea has serious financing depth around a much smaller number of domestic AI-chip companies.

Country or group	Deals	Companies	Capital recorded
United States	22	12	$8.95B
South Korea	9	3	$1.53B
Canada	3	2	$900M
United Kingdom	5	3	$777M
China	8	6	~$763M
Other countries	5	5	~$69M

Does AI chip funding happen steadily throughout the year?

AI chip funding comes in bursts, with long quiet stretches broken by months when several major companies raise almost at once.

Across the 32 complete months from January 2024 through August 2026, the median month has only one financing and $106.5 million of capital.

February 2026 looks completely different. Six companies raised $2.469 billion in the same month. Cerebras brought in $1 billion, MatX $500 million, SambaNova $350 million, Positron $230 million, OLIX $220 million and Taalas $169 million.

Because six separate companies contributed, the spike cannot be explained away as one oversized transaction.

We see the same basic rhythm elsewhere in the period. Several months contain little activity, and then investors write very large checks across different architectures within a short window. Funding conviction appears to arrive in waves, especially once a group of companies reaches similar technical or commercial milestones.

What is the biggest bottleneck for AI chip startups right now?

The hardest part of AI chip fundraising today is joining the small group of companies that can keep raising nine-figure rounds after their technology starts leaving the lab.

Our research repeatedly points to that transition. Early rounds are expensive but still relatively available across the cohort. The ladder becomes tighter around Series C. Repeat fundraisers absorb most of the capital. Strategic investors become increasingly common at later stages. A funding ladder that tightens around Series C is a real obstacle, and we dig into the biggest unsolved problems in our AI chips report.

The latest deals also show what investors want by then. Etched had hardware in Jane Street's data center. Positron had dozens of deployed racks before raising for its next silicon generation. Rebellions had moved into production-scale infrastructure. FuriosaAI's financing was tied to scaling RNGD production and developing its next chip.

A technically interesting processor can get a startup into the race. Staying there requires manufacturing, software, systems, customers and a supply chain that can support real deployment.

That is where the market currently separates promising chip companies from the handful of platforms able to attract another $200 million, $500 million or $1 billion.

What should we actually take away from the funding data?

AI chip funding has become a genuinely huge private-capital market. A $136 million median financing tells us more than the headline billions do: large checks are normal here.

The market is also much narrower than the deal count suggests. A few dozen companies generate all the activity, and repeat fundraisers form the financial backbone of the sector.

Inference is where competition is deepest today. Many architectures are being funded, from conventional digital accelerators to photonics, memory-centric designs and model-specific silicon, because the cost and speed of running AI models have become major commercial problems.

The really expensive stage comes after early technical validation. Tape-outs lead to production commitments, working chips lead to systems, and working systems lead to racks, factories, supply chains and data-center deployments. Capital requirements rise with that progression.

Strategic investors increasingly sit inside that process. Memory companies, processor-IP suppliers, infrastructure vendors and customers are showing up alongside financial investors because scaling an AI-chip company now depends on an industrial ecosystem.

The biggest open question is breadth. Capital is abundant for the companies that make it into the top tier, while first financings and small experimental rounds remain scarce. The next phase of the market will depend on how many new companies can cross that gap and become credible production-scale competitors.