What are the most valuable semiconductor startups?

Top startups in the semiconductor market ranked by valuation

Here is an updated table that ranks the top startups in the semiconductor market based on their latest reported or estimated valuations.

If you need to dig deeper and get more detailed data, please check our report covering the semiconductor market.

The semiconductor valuation ranking is currently led by CXMT by an enormous margin, followed by Kioxia, Cambricon, Astera Labs and Cerebras Systems.

CXMT is now worth roughly $589 billion on the public market. Kioxia is worth close to $200 billion, while Cambricon is around $103 billion. Astera Labs has recently traded around a $60 billion-plus market capitalization, and Cerebras is close to $50 billion.

After those five, the numbers drop quickly. YMTC is being valued in the tens of billions ahead of a possible listing, while GigaDevice and Montage Technology are each worth around $40 billion. MetaX, Moore Threads and several Chinese foundry or AI-chip companies also sit well above $20 billion.

Using the supplied dataset and midpoint values for ranges, the first five companies account for close to two-thirds of the roughly $1.5 trillion represented by all 97 companies. The top 10 account for close to four-fifths.

So someone scrolling through a list of almost 100 semiconductor companies could easily miss the main story. Most of the value is sitting in a very small number of companies.

Rank	Company	Approximate current valuation
1	CXMT	~$589B
2	Kioxia	~$190B–$200B
3	Cambricon	~$103B
4	Astera Labs	~$60B+
5	Cerebras Systems	~$49B
6	YMTC	~$41B–$49B
7	GigaDevice	~$42B
8	Montage Technology	~$40B
9	MetaX Integrated Circuits	~$35B
10	Hua Hong Semiconductor	~$25B–$34B depending on market reference
11	Moore Threads	~$31B
12	Tower Semiconductor	~$27B

Is CXMT really worth more than every other semiconductor startup?

Yes. CXMT currently sits so far ahead of the rest of the semiconductor ranking that its valuation has become an outlier rather than a normal startup comparison.

The Chinese DRAM maker came to market at an IPO valuation of roughly $85.5 billion. Reuters reported that its shares then jumped 466% on their first trading day, taking the company to almost $488 billion immediately after listing. Its market capitalization has since moved close to $590 billion.

That puts CXMT at roughly three times Kioxia's value and more than five times Cambricon's. CXMT alone is also worth roughly as much as a large block of the companies ranked immediately below the top five combined.

The speed of the repricing is the key part. CXMT did not slowly climb from an $80 billion private valuation to almost $600 billion through several funding rounds. Public investors repriced the company within weeks.

That makes CXMT a perfectly valid answer to “Which semiconductor company in this ranking is worth the most today?” It is a much less useful benchmark for what a private semiconductor startup can realistically raise at.

Why are semiconductor valuations so high right now?

Semiconductor valuations are unusually high today because AI infrastructure and memory demand are pushing the underlying chip market into a much bigger revenue cycle.

World Semiconductor Trade Statistics reported that global semiconductor sales reached about $702 billion in the first half of 2026, more than double the same period a year earlier. Memory sales grew 305% year over year.

Gartner has since raised its full-year semiconductor forecast to roughly $1.56 trillion. Its latest estimate puts memory revenue alone at about $837 billion, versus $220 billion in 2025. That would make memory more than half of semiconductor industry revenue this year.

The boom also reaches beyond memory. Gartner expects non-memory semiconductor revenue to rise from about $589 billion to $718 billion, with AI accelerators, networking chips, power management and optical interconnects all benefiting from the data-center buildout.

We can see that mix directly in the valuation ranking. CXMT, Kioxia and YMTC are memory companies. Cambricon, Cerebras and MetaX sell AI compute. Astera Labs sells connectivity chips. Ayar Labs and Lightmatter are attacking optical interconnects. Upscale AI and Enfabrica are working on networking.

Investors are paying up because several semiconductor bottlenecks are getting expensive at the same time.

How many semiconductor companies in this ranking are worth more than $1 billion?

At least 64 of the 97 semiconductor companies in our dataset are worth $1 billion or more based on their stated valuation or the lower end of their valuation range.

That is roughly two-thirds of the entire ranking.

Twenty companies are already worth at least $10 billion. Around one-third of the list reaches roughly $5 billion or more. By the time we get to companies around $1 billion, we are already well past the middle of the table.

Achronix and Phytium sit around the boundary because their estimated ranges extend above and below $1 billion. Kneron reaches $1 billion at the high end of its estimate. Black Sesame Technologies is also hovering near the threshold on the public market.

There is one caveat worth keeping in mind: this does not mean the market contains 64 conventional private unicorns. Many of these companies are already public, and others have been acquired.

Still, the number says something important about the scale of the sector. Billion-dollar chip companies are no longer rare exceptions.

Valuation level	Approximate number in dataset
$100B+	3
$50B+	5
$25B+	About 12–13
$10B+	20
Around $5B+	About one-third of the dataset
Definitely $1B+	64
Below $1B	About one-third
Total companies	97

Are the biggest semiconductor startups even startups anymore?

Many of the most valuable names in the semiconductor startup ranking have already moved well beyond what most people would call a startup.

CXMT, Kioxia, Cambricon, Astera Labs and Cerebras are publicly listed. GigaDevice, Montage, MetaX, Hua Hong, Moore Threads, GlobalFoundries and several others further down the ranking are public too.

Another group has disappeared into larger companies. SoftBank completed its $6.5 billion acquisition of Ampere Computing. Marvell completed its purchase of Celestial AI after agreeing to about $3.25 billion of upfront consideration. Qualcomm paid about $1.4 billion for Nuvia. Analog Devices disclosed a roughly $1.5 billion acquisition of Empower Semiconductor.

We keep these companies because the ranking is useful for measuring value created by semiconductor businesses that came through the startup or scale-up ecosystem. But someone specifically searching for the most valuable independent private semiconductor startups should look at a narrower group.

That group includes companies such as YMTC, UNISOC, Tenstorrent, MatX, Ayar Labs, Lightmatter, Positron AI, Groq, Rebellions, d-Matrix and EnCharge AI.

Which private semiconductor startups are worth the most right now?

YMTC is currently the biggest private semiconductor company in the ranking, while most of the leading U.S. private chip startups sit somewhere between roughly $2 billion and $10 billion.

YMTC's potential listing has been discussed at a valuation of up to roughly 330 billion yuan, or around the mid-$40 billions in dollars. That puts it well ahead of the next private group.

UNISOC and Tenstorrent are both estimated in roughly the high-single-digit billions, although those estimates carry more uncertainty than an announced funding valuation.

MatX needs a freshness adjustment. Earlier financing evidence placed the company around $7 billion, but Reuters recently reported that MatX was seeking new funding at about a $4 billion valuation after Anthropic had previously discussed buying the company for roughly $7 billion. For a current ranking, we give that newer financing signal more weight than the older $7 billion figure.

Ayar Labs is easier to pin down. The company announced a $500 million Series E at a $3.75 billion valuation. Positron AI recently raised $875 million at a $5 billion post-money valuation. Groq announced a new financing that values it at $3.5 billion. Rebellions reached approximately $2.34 billion after a $400 million pre-IPO round.

The biggest private semiconductor companies are therefore valuable, but the gap between them and public-market outliers such as CXMT or Cambricon is enormous.

Private company	Current valuation anchor	Evidence type
YMTC	~$41B–$49B	IPO/listing evidence
UNISOC	~$8B–$10.5B	Estimate
Tenstorrent	~$8B–$10B	Estimate
Positron AI	$5B	Completed financing
MatX	About $4B current fundraising target	Recent reported financing
Lightmatter	$4.4B last announced round	Completed financing
Ayar Labs	$3.75B	Completed financing
Groq	$3.5B	Completed financing
Rebellions	~$2.34B	Completed financing
EnCharge AI	~$2.0B	Recent financing evidence
d-Matrix	~$2.0B	Private-market valuation
Upscale AI	$2.0B	Completed financing

Are AI chip startups taking over the semiconductor valuation ranking?

AI companies now dominate the newer part of semiconductor value creation, even though memory companies still own some of the largest absolute valuations.

Cambricon is worth more than $100 billion. Cerebras is close to $50 billion. MetaX and Moore Threads are worth tens of billions. Biren and Iluvatar also sit well into eight-figure billions.

The pattern gets even clearer among private companies. Tenstorrent, MatX, Positron, Groq, Rebellions, FuriosaAI, d-Matrix, EnCharge AI, SiMa.ai, Hailo and several others are all directly exposed to AI compute or inference.

Funding data backs this up. Crunchbase counted around $10.7 billion invested into semiconductor startups in seed through pre-IPO rounds during the first part of 2026 alone. That already put the market on pace to beat an unusually strong previous year.

AI's impact also spills into companies that do not build accelerators. Astera Labs benefits from connecting AI processors. Ayar Labs and Lightmatter move data optically. Upscale AI builds networking infrastructure. Memory suppliers benefit because modern AI servers need enormous amounts of DRAM, HBM and NAND.

AI is now the main force shaping semiconductor startup valuations, even when “AI chip” is not the most accurate description of what the company sells.

Why are memory-chip companies worth so much today?

Memory-chip companies are worth so much right now because AI servers are consuming more memory just as memory prices and industry revenue have surged.

That shows up immediately at the top of the ranking. CXMT is first. Kioxia is second. YMTC is worth tens of billions. GigaDevice also has major memory exposure.

The industry numbers are even more striking. Gartner currently expects memory revenue to reach about $837 billion this year, compared with roughly $220 billion in 2025. Its forecast has DRAM revenue rising about 247% and NAND flash around 372%.

WSTS is already seeing the same pattern in actual sales data: memory revenue was up 305% year over year in the first half.

AI servers are one reason. Large accelerator clusters need far more memory capacity and bandwidth than ordinary servers, particularly HBM around accelerators and large amounts of DRAM throughout the system.

Memory is also highly cyclical, so we should not assume that every dollar of today's valuation survives the next supply cycle. But these companies are currently operating inside one of the strongest memory markets the semiconductor industry has seen.

Is China dominating semiconductor startup valuations?

China currently dominates the upper end of this semiconductor valuation ranking, especially in memory, AI accelerators and domestic chip manufacturing.

CXMT is number one. Cambricon is around $103 billion. YMTC, GigaDevice, Montage, MetaX and Moore Threads all sit high in the table. Hua Hong, Biren, Iluvatar, Rockchip, Nexchip and Loongson add more Chinese names further down.

The interesting part is the range of businesses involved.

CXMT makes DRAM. YMTC makes NAND. Cambricon, MetaX, Moore Threads, Biren and Iluvatar target accelerated computing. Hua Hong and Nexchip manufacture chips. Montage works on memory-interface and server silicon. GigaDevice sells memory and microcontrollers.

That breadth tells us more than simply counting Chinese AI-chip companies. China has created valuable businesses across several layers of the semiconductor stack.

Domestic substitution helps explain part of this. Chinese customers and policymakers have strong incentives to build alternatives to imported memory, CPUs, GPUs and manufacturing capacity. Public investors are currently attaching very large values to companies that can fill those gaps.

High valuation does not automatically mean global technical leadership. Nvidia, TSMC, SK Hynix and other incumbents still hold enormous advantages in particular markets. What the ranking clearly shows is that China's semiconductor ecosystem now attracts huge amounts of market value.

How do American semiconductor startups compare with Chinese chip companies?

Chinese semiconductor companies currently lead on absolute valuation, while American startups are much more concentrated around the AI data-center stack.

China has CXMT near $590 billion and Cambricon above $100 billion, plus a deep group of public companies worth tens of billions. That creates a scale that the current U.S. private semiconductor market does not match.

The American picture looks different. Astera Labs and Cerebras have reached very large public valuations. Among private companies, Positron, Lightmatter, Ayar Labs, MatX, Groq, d-Matrix and EnCharge AI cluster mostly between about $2 billion and $5 billion using the freshest direct financing evidence.

Their focus is also narrower. Many American newcomers are trying to fix one expensive AI infrastructure problem: inference, optical interconnects, networking, power efficiency or data movement.

China's valuable semiconductor companies stretch across memory, fabrication, CPUs, GPUs, sensors, connectivity and microcontrollers.

The two ecosystems are creating semiconductor value in different ways. China is building across a broad domestic supply chain, while much of the current U.S. startup boom revolves around making large AI systems faster or cheaper.

Why are photonics and chip-interconnect startups suddenly worth billions?

Photonics and chip-interconnect startups are reaching multi-billion-dollar valuations because moving data between AI processors has become almost as important as adding more processors.

Ayar Labs raised $500 million at a $3.75 billion valuation to expand production of co-packaged optical technology. Lightmatter reached a $4.4 billion valuation after raising $400 million and now reports about $850 million in total funding.

Celestial AI provides an even clearer strategic benchmark. Marvell agreed to pay approximately $3.25 billion upfront for the company, with potential additional consideration tied to future revenue milestones.

Astera Labs shows what can happen once connectivity moves from startup story to public-market growth story. Its market capitalization has recently moved above $60 billion, up dramatically from its valuation around the time of its public listing.

The technical reason is straightforward. AI clusters now combine thousands or tens of thousands of processors, and those processors constantly need to exchange data. Electrical connections consume power and become harder to scale as clusters get larger.

As compute keeps growing, investors are increasingly willing to pay billions for companies that help all those expensive chips communicate efficiently.

Which semiconductor startups have increased their valuations the fastest?

Positron AI is one of the clearest recent examples of a semiconductor startup whose valuation has jumped extremely quickly.

Positron announced a $230 million Series B earlier this year at a valuation above $1 billion. The company then raised an $875 million Series C at a $5 billion post-money valuation. That is roughly a fivefold valuation increase within the same year.

Rebellions has also moved quickly. Its latest $400 million pre-IPO financing valued the company around $2.34 billion and came only months after a $250 million Series C. The company says more than three-quarters of all the capital it has ever raised arrived within a six-month period.

Lightmatter provides an earlier example. Its $400 million Series D valued the company at $4.4 billion, four times its previous valuation according to its own announcement.

Upscale AI reached $2 billion after a $190 million financing extension and has now raised roughly $500 million in total.

There are cases where the latest signal goes the other way. MatX had previously been associated with a valuation around $7 billion, yet Reuters recently reported that the company was seeking fresh capital at about $4 billion. That is why we prefer the newest credible evidence over repeating an older headline valuation forever.

Does raising more money make a semiconductor startup more valuable?

No. Semiconductor funding helps companies survive expensive development cycles, but the amount raised tells us surprisingly little about the valuation investors will eventually give the business.

Positron is currently worth $5 billion after raising a little over $1 billion in total. Lightmatter is worth $4.4 billion with about $850 million raised. Ayar Labs is worth $3.75 billion after raising $870 million.

Groq is a useful counterexample. The company's newest financing values it at $3.5 billion, while the amount of capital flowing into the business has become very large. Mythic has also raised hundreds of millions of dollars while its latest valuation evidence remains far below that level.

At the distressed end, companies can raise huge sums and still lose most of their value if commercial adoption never catches up with technical spending.

We would never treat valuation divided by cumulative funding as an investment return. The capital enters at different prices and investors own different percentages after every round. It still gives us a rough way to compare how much market value a business has created relative to the capital it has consumed.

Company	Current valuation anchor	Reported funding	Valuation / funding
Positron AI	$5.0B	~$1.18B	~4.2x
Lightmatter	$4.4B	~$850M	~5.2x
Ayar Labs	$3.75B	$870M	~4.3x
Rebellions	~$2.34B	~$850M	~2.8x
Upscale AI	$2.0B	$500M	4.0x
EnCharge AI	~$2.0B	~$454M	~4.4x
d-Matrix	~$2.0B	~$450M	~4.4x
Groq	$3.5B	Multi-billion-dollar cumulative capital base	Lower than several peers

Why do semiconductor startups need such huge funding rounds?

Semiconductor startups need unusually large funding rounds because building a competitive chip company now requires far more than designing one good processor.

A company has to design the architecture, tape out the chip, pay foundries, package it, test it, write software, build developer tools, validate the product with customers and often reserve expensive manufacturing capacity before meaningful revenue arrives.

The numbers reflect that burden. Positron just raised $875 million in one round. Ayar Labs raised $500 million. Rebellions raised $400 million. Lightmatter raised $400 million. Groq has recently raised another $1 billion across two financings disclosed only a few months apart.

Crunchbase found around $10.7 billion flowing into semiconductor startups during the first part of this year alone.

This creates a harsh filter. A software startup can sometimes reach paying customers with a few million dollars. A semiconductor company can spend hundreds of millions before the market has really proven whether the chip will sell at scale.

That is why semiconductor rounds look enormous even before the companies themselves become enormous.

Are public semiconductor stocks making this valuation ranking look more extreme?

Yes. Public semiconductor companies make the top of the ranking much larger and much more volatile than the private-market portion.

Astera Labs is a good example. The company has recently moved between roughly $44 billion and more than $60 billion within a short period because its shares trade every day.

Cerebras has also swung sharply since its listing, including periods when its market capitalization moved above $60 billion before returning closer to $50 billion.

Private companies do not work that way. Positron can remain officially worth $5 billion until another financing, secondary transaction or acquisition gives us better evidence. Its economic prospects may change every week, but its last directly observed private valuation does not.

CXMT makes the distortion especially obvious. Using the dataset values, that single public company represents close to 40% of the combined midpoint valuation of all 97 names.

For readers interested specifically in venture-backed startups, public-market names should therefore be mentally separated from private companies. For readers asking how much value semiconductor companies have created overall, keeping both groups is useful.

Are semiconductor acquisitions producing billion-dollar exits?

Yes. Strategic buyers have already paid several billion dollars for semiconductor startups when the technology fills an important gap in a larger computing platform.

SoftBank paid $6.5 billion for Ampere Computing. Marvell agreed to roughly $3.25 billion of upfront consideration for Celestial AI. Analog Devices disclosed a roughly $1.5 billion purchase of Empower Semiconductor. Qualcomm paid approximately $1.4 billion for Nuvia.

All four companies owned technology that could slot directly into a much larger platform: Arm server processors, optical AI connectivity, integrated power management and high-performance CPU cores.

Below the billion-dollar level, exit values fall quickly. Historical deals such as pSemi, Aquantia and Annapurna Labs were worth hundreds of millions rather than billions.

The pattern is fairly clear. Semiconductor companies can produce spectacular exits when a buyer sees the technology as strategically difficult to reproduce internally. Years of technically impressive silicon by themselves do not guarantee one.

Has AI inference become the hottest semiconductor startup category?

AI inference has become one of the busiest areas for semiconductor startup funding and valuation, with several companies now worth between roughly $2 billion and $5 billion.

Positron sits at $5 billion. Groq is valued at $3.5 billion. Rebellions is around $2.34 billion. EnCharge AI is around $2 billion. d-Matrix is also around $2 billion, while FuriosaAI sits in a similar valuation neighborhood based on available financing evidence.

These companies are not all building the same architecture. Groq emphasizes extremely fast inference. d-Matrix uses in-memory computing. EnCharge AI is pursuing analog in-memory acceleration. Positron focuses on memory-efficient inference infrastructure. Rebellions sells vertically integrated inference systems.

They are all chasing the same basic economic problem: once AI models are trained, somebody has to run them every time a user sends a prompt.

Positron's fundraising gives us the strongest current signal. The company moved from a valuation above $1 billion to $5 billion within months as investors backed its attempt to lower inference costs. Groq now says it serves more than six million developers across 13 data centers, while Rebellions has raised $650 million in only six months to expand production-scale inference infrastructure.

Training still attracts enormous spending, but inference is now big enough to support an entire group of multi-billion-dollar chip companies.

Which semiconductor businesses are getting the highest valuations?

Memory, AI compute and AI infrastructure currently produce the largest semiconductor valuations, while smaller edge-AI and IoT chip companies generally sit much lower down the table.

Memory owns the first two positions through CXMT and Kioxia, with YMTC also worth tens of billions.

AI compute produces Cambricon, Cerebras, MetaX, Moore Threads, Biren and Iluvatar near the top, followed by a large group of private inference and accelerator companies.

Manufacturing still matters too. Hua Hong, Tower Semiconductor, GlobalFoundries, Nexchip, Powerchip and Vanguard International Semiconductor all reach multi-billion or tens-of-billions valuations because foundry capacity remains strategically valuable.

Then comes the infrastructure around the processors. Astera Labs has become one of the most valuable companies in the entire ranking through data-center connectivity. Lightmatter, Ayar Labs and Celestial AI show how much optical interconnects are now worth.

Further down, many edge-AI, IoT and experimental architecture companies remain below $1 billion. Kneron, Atmosic, Rain AI, Mythic, GreenWaves and GrAI Matter Labs all sit far below the leaders.

The market is currently rewarding companies attached to the most expensive bottlenecks: memory, compute, fabrication and data movement.

What does the semiconductor startup valuation ranking actually tell us today?

The semiconductor valuation boom is real, but most of the value is piling into a small number of companies exposed to memory, AI compute and the infrastructure needed to connect huge AI systems.

The first striking result is scale. We have 97 companies in the dataset, yet the five largest account for close to two-thirds of the combined midpoint valuation. CXMT alone represents an extraordinary share.

The second result is how much AI has changed the map. Processor startups attract obvious attention, but some of the strongest valuations now belong to companies making memory, networking chips and optical interconnects. Gartner's forecast that AI data-center semiconductors will take a growing share of the industry's revenue fits what we are already seeing in company valuations.

China is the third major part of the story. Its semiconductor companies occupy an unusually large share of the top ranking across memory, AI chips, foundries and other categories.

Private U.S. startups are still creating major outcomes. Positron at $5 billion, Lightmatter at $4.4 billion, Ayar Labs at $3.75 billion and Groq at $3.5 billion would have been huge semiconductor valuations in almost any earlier startup cycle. They simply look smaller today because several public chip companies have reached tens or hundreds of billions.

Our final judgment is clear: semiconductor startups and scale-ups are currently creating exceptional amounts of value, but this boom is highly concentrated. The companies closest to AI's biggest physical bottlenecks are pulling away from the rest of the market.