What are the most valued startups in the AI chip market?
Download our beautiful pitch about the AI chip market

In our AI chip market deck, you will find everything you need to understand the market
The AI chip market includes semiconductor startups building data-center accelerators, inference processors, photonic systems, memory architectures, and the infrastructure needed to run advanced artificial intelligence models.
We update this valuation ranking every month to reflect new funding rounds, public market movements, acquisitions, and credible valuation estimates.
The current list shows a highly concentrated market, with the largest Chinese GPU and accelerator companies accounting for a substantial share of total sector value.
And if you want to better understand this new industry, you can download our pitch covering the AI chip market.
A quick summary table
| Metric | Value |
|---|---|
| Most valuable AI chip startup | Cambricon, $128.6B |
| Second most valuable AI chip startup | Moore Threads, $45.8B |
| Median AI chip startup valuation | Approximately $400M |
| AI chip valuation captured by the top 10 | Approximately 84.0% |
| Top AI chip valuation versus median | Approximately 322 times |
| Median AI chip valuation-to-capital-raised ratio | Approximately 4.2 times |
| AI chip startups valued at $1B+ | 28 |

This market map, featured in our AI chip market deck, highlights top companies and startups in the AI chip market
Top startups in the AI chip market ranked by valuation
Here is an updated table that ranks the top startups in the AI chip market based on their latest reported or estimated valuations.
If you want more detaild about their fundraising activity, you can check our list of the startups who have raised the most funding in the AI chip market.
| # | Startup Name | What They Do | Current Valuation ($) | Valuation Confidence Level | Valuation Type | Evidence Status | Total Funding ($) | Funding Confidence Level |
|---|---|---|---|---|---|---|---|---|
| 1 | Cambricon | AI processors and accelerator systems | $128.6B | Full Confidence | Public Market Cap | Observed | $470M | Partial Confidence |
| 2 | Moore Threads | Chinese full-stack GPUs | $45.8B | Full Confidence | Public Market Cap | Observed | $1.8B | Strong Confidence |
| 3 | MetaX | Data-center training GPU platforms | $43.5B | Full Confidence | Public Market Cap | Observed | $915M | Partial Confidence |
| 4 | Kunlunxin | Baidu data-center AI accelerators | $16.0B–$23.0B | Partial Confidence | IPO or Listing Range Valuation | Estimated | $650M | Partial Confidence |
| 5 | Biren Technology | Data-center GPGPU accelerator chips | $13.4B–$16.1B | Partial Confidence | Public Market Cap | Observed | $1.8B | Partial Confidence |
| 6 | Cerebras Systems | Wafer-scale AI systems | $11.4B | Full Confidence | Public Market Cap | Observed | $8.8B | Strong Confidence |
| 7 | SambaNova Systems | Dataflow AI systems platform | $10.5B–$11.5B | Strong Confidence | Announced Private Round Valuation | Observed | $1.5B | Strong Confidence |
| 8 | Tenstorrent | RISC-V AI processor architecture | $8.0B–$10.0B | Partial Confidence | Proxy-Based Estimate | Estimated | $1.0B | Partial Confidence |
| 9 | Groq | LPU inference cloud chips | $7.0B–$9.0B | Partial Confidence | Implied Valuation from Raise | Implied | $2.4B | Partial Confidence |
| 10 | Etched | Transformer-specific inference chips | $4.8B–$5.2B | Strong Confidence | Announced Private Round Valuation | Observed | $800M | Partial Confidence |
| 11 | Lightmatter | Photonic AI interconnects | $4.0B–$5.0B | Partial Confidence | Announced Private Round Valuation | Estimated | $822M | Strong Confidence |
| 12 | Unconventional AI | Energy-efficient alternative AI computing | $4.5B | Full Confidence | Announced Private Round Valuation | Observed | $475M | Full Confidence |
| 13 | Iluvatar CoreX / Tianshu Zhixin | General-purpose data-center AI GPUs | $4.3B | Full Confidence | Public Market Cap | Observed | $807M | Strong Confidence |
| 14 | Lightelligence | Photonic computing and optical interconnects | $3.5B | Full Confidence | Public Market Cap | Observed | $566M | Partial Confidence |
| 15 | Celestial AI | Photonic AI interconnect fabric | $3.3B | Full Confidence | Acquisition Value | Observed | $520M | Strong Confidence |
| 16 | Enflame Technology | Cloud AI training chips | $2.6B–$3.0B | Partial Confidence | IPO or Listing Range Valuation | Estimated | $746M | Strong Confidence |
| 17 | Rebellions | AI inference accelerator infrastructure | $2.2B–$2.4B | Strong Confidence | Announced Private Round Valuation | Observed | $850M | Strong Confidence |
| 18 | d-Matrix | Digital in-memory inference chips | $1.9B–$2.1B | Strong Confidence | Announced Private Round Valuation | Observed | $450M | Partial Confidence |
| 19 | Habana Labs | Data-center AI training accelerators | $2.0B | Full Confidence | Acquisition Value | Observed | $120M | Strong Confidence |
| 20 | Axelera AI | Energy-efficient AI accelerator platforms | $1.5B–$2.2B | Partial Confidence | Implied Valuation from Raise | Implied | $380M | Strong Confidence |
| 21 | Rivos | RISC-V AI server chips | $1.5B–$2.0B | Low Confidence | Proxy-Based Estimate | Estimated | $370M | Partial Confidence |
| 22 | Taalas | Model-specific AI inference chips | $1.1B–$1.8B | Partial Confidence | Implied Valuation from Raise | Implied | $219M | Strong Confidence |
| 23 | Preferred Networks | Integrated AI chips and models | $1.0B–$1.5B | Partial Confidence | Comparables-Based Estimate | Estimated | $308M | Partial Confidence |
| 24 | Positron AI | Energy-efficient AI inference systems | $1.0B–$1.2B | Strong Confidence | Announced Private Round Valuation | Observed | $305M | Full Confidence |
| 25 | OLIX Computing | Photonic AI inference processors | $1.0B–$1.2B | Strong Confidence | Announced Private Round Valuation | Observed | $250M | Partial Confidence |
| 26 | Fractile | High-speed AI inference processors | $950M–$1.1B | Strong Confidence | Announced Private Round Valuation | Observed | $238M | Strong Confidence |
| 27 | Mythic | Analog in-memory inference processors | $800M–$1.2B | Partial Confidence | Implied Valuation from Raise | Implied | $290M | Partial Confidence |
| 28 | Vastai Technologies | Vision and video AI processors | $800M–$1.2B | Partial Confidence | Comparables-Based Estimate | Estimated | $378M | Partial Confidence |
| 29 | Enfabrica | AI networking and memory fabric | $900M–$1.0B | Strong Confidence | Acquisition Value | Observed | $240M | Full Confidence |
| 30 | Openchip | RISC-V AI and HPC processors | $760M–$850M | Strong Confidence | Active Raise Valuation | Implied | $178M | Partial Confidence |
| 31 | FuriosaAI | Data-center AI inference accelerators | $700M–$800M | Strong Confidence | Announced Private Round Valuation | Observed | $246M | Partial Confidence |
| 32 | SiMa.ai | Physical AI inference accelerators | $600M–$900M | Partial Confidence | Implied Valuation from Raise | Implied | $355M | Strong Confidence |
| 33 | Neurophos | Photonic data-center AI processors | $550M–$920M | Strong Confidence | Implied Valuation from Raise | Implied | $117M | Strong Confidence |
| 34 | Hailo | Edge AI accelerator chips | $600M–$850M | Low Confidence | Comparables-Based Estimate | Estimated | $340M | Strong Confidence |
| 35 | Tachyum | Universal AI and HPC processors | $550M–$900M | Low Confidence | Implied Valuation from Raise | Estimated | $300M | Partial Confidence |
| 36 | EnCharge AI | Analog in-memory AI processors | $500M–$830M | Strong Confidence | Implied Valuation from Raise | Implied | $144M | Full Confidence |
| 37 | Recogni | Data-center generative AI inference | $500M–$750M | Partial Confidence | Comparables-Based Estimate | Estimated | $176M | Strong Confidence |
| 38 | DEEPX | Low-power edge AI processors | $500M–$650M | Partial Confidence | Announced Private Round Valuation | Estimated | $103M | Strong Confidence |
| 39 | Kneron | Edge AI neural processors | $450M–$700M | Low Confidence | Comparables-Based Estimate | Estimated | $195M | Partial Confidence |
| 40 | Q.ANT | Photonic data-center processors | $400M–$670M | Strong Confidence | Implied Valuation from Raise | Implied | $80M | Strong Confidence |
| 41 | Cornami | Many-core AI security processors | $350M–$650M | Partial Confidence | Comparables-Based Estimate | Estimated | $164M | Partial Confidence |
| 42 | VSORA | High-performance AI inference chips | $310M–$500M | Partial Confidence | Implied Valuation from Raise | Implied | $66M | Partial Confidence |
| 43 | Achronix | FPGA acceleration chips and IP | $300M–$500M | Low Confidence | Revenue or ARR Multiple Estimate | Estimated | $122M | Partial Confidence |
| 44 | Mobilint | Edge AI NPU chips | $310M–$480M | Partial Confidence | Implied Valuation from Raise | Implied | $62M | Partial Confidence |
| 45 | Nervana Systems | Deep-learning training processors | $350M–$400M | Strong Confidence | Acquisition Value | Observed | $28M | Partial Confidence |
| 46 | MemryX | Edge AI accelerator chips | $290M–$430M | Partial Confidence | Implied Valuation from Raise | Implied | $63M | Partial Confidence |
| 47 | MatX | LLM-focused AI accelerators | $250M–$450M | Partial Confidence | Announced Private Round Valuation | Estimated | $105M | Partial Confidence |
| 48 | Cornelis Networks | AI and HPC network fabric | $250M–$400M | Partial Confidence | Implied Valuation from Raise | Implied | $85M | Partial Confidence |
| 49 | Baya Systems | Chiplet interconnect system IP | $240M–$400M | Partial Confidence | Implied Valuation from Raise | Implied | $36M | Full Confidence |
| 50 | Kinara | Programmable edge AI NPUs | $307M | Full Confidence | Acquisition Value | Observed | $54M | Partial Confidence |
| 51 | Sapeon | Data-center AI inference processors | $250M–$300M | Strong Confidence | Acquisition Value | Implied | $62M | Partial Confidence |
| 52 | Quadric | On-device AI processor IP | $220M–$320M | Strong Confidence | Implied Valuation from Raise | Implied | $72M | Strong Confidence |
| 53 | Ventana Micro Systems | RISC-V data-center processor chiplets | $180M–$320M | Low Confidence | Acquisition Value | Estimated | $53M | Strong Confidence |
| 54 | Oxmiq Labs | Licensable GPU and AI architecture | $180M–$290M | Strong Confidence | Implied Valuation from Raise | Implied | $60M | Strong Confidence |
| 55 | Vertical Compute | Three-dimensional AI memory chiplets | $170M–$300M | Partial Confidence | Implied Valuation from Raise | Implied | $65M | Partial Confidence |
| 56 | GSI Technology | Compute-in-memory accelerator processors | $234M | Full Confidence | Public Market Cap | Observed | $84M | Partial Confidence |
| 57 | DeePhi Tech | FPGA deep-learning acceleration | $230M | Strong Confidence | Acquisition Value | Observed | $50M | Strong Confidence |
| 58 | Ceremorphic | Energy-efficient AI supercomputing chips | $140M–$280M | Low Confidence | Comparables-Based Estimate | Estimated | $50M | Full Confidence |
| 59 | Salience Labs | Photonic AI connectivity chips | $150M–$230M | Partial Confidence | Implied Valuation from Raise | Implied | $42M | Strong Confidence |
| 60 | Blaize | Edge AI computing platform | $175M | Full Confidence | Public Market Cap | Observed | $158M | Partial Confidence |
| 61 | Flex Logix | Embedded FPGA and inference IP | $120M–$200M | Partial Confidence | Acquisition Value | Estimated | $82M | Strong Confidence |
| 62 | AI Analog Inference / Sagence AI | Analog in-memory inference processors | $120M–$180M | Partial Confidence | Comparables-Based Estimate | Estimated | $58M | Partial Confidence |
| 63 | NeuReality | AI inference server processors | $100M–$180M | Partial Confidence | Implied Valuation from Raise | Implied | $70M | Partial Confidence |
| 64 | Arago Computing | Photonic AI inference processors | $104M–$173M | Strong Confidence | Implied Valuation from Raise | Implied | $26M | Full Confidence |
| 65 | Snowcap Compute | Superconducting AI compute systems | $92M–$153M | Strong Confidence | Implied Valuation from Raise | Implied | $23M | Full Confidence |
| 66 | Lemurian Labs | AI compute software architecture | $100M–$120M | Partial Confidence | Comparables-Based Estimate | Estimated | $37M | Full Confidence |
| 67 | NEUCHIPS | Recommendation inference accelerator ASICs | $80M–$130M | Low Confidence | Comparables-Based Estimate | Estimated | $20M | Strong Confidence |
| 68 | Extropic | Thermodynamic probabilistic AI processors | $60M–$100M | Partial Confidence | Comparables-Based Estimate | Estimated | $14M | Full Confidence |
| 69 | Netrasemi | Edge AI system-on-chips | $74M | Strong Confidence | Announced Private Round Valuation | Observed | $15M | Partial Confidence |
| 70 | Lumai | Free-space optical AI accelerators | $50M–$85M | Partial Confidence | Implied Valuation from Raise | Implied | $10M | Full Confidence |
| 71 | NeuroBlade | Processing-in-memory analytics accelerators | $40M–$90M | Low Confidence | Acquisition Value | Estimated | $110M | Strong Confidence |
| 72 | Rain AI | Compute-in-memory AI processors | $25M–$75M | Low Confidence | Proxy-Based Estimate | Estimated | $38M | Partial Confidence |
| 73 | GEMESYS | Brain-inspired edge AI chips | $36M–$60M | Partial Confidence | Implied Valuation from Raise | Implied | $9M | Partial Confidence |
| 74 | SEMRON | Three-dimensional in-memory AI chips | $32M–$55M | Partial Confidence | Implied Valuation from Raise | Implied | $8M | Strong Confidence |
| 75 | Opticore | Optical AI processing units | $30M–$50M | Strong Confidence | Implied Valuation from Raise | Implied | $15M | Strong Confidence |
| 76 | Akhetonics | All-optical general-purpose processors | $25M–$42M | Partial Confidence | Implied Valuation from Raise | Implied | $9M | Strong Confidence |
| 77 | Morphing Machines | Reconfigurable many-core processor IP | $25M–$40M | Strong Confidence | Announced Private Round Valuation | Estimated | $14M | Partial Confidence |
| 78 | Ubitium | Universal RISC-V processor architecture | $18M–$30M | Partial Confidence | Implied Valuation from Raise | Implied | $4M | Full Confidence |
| 79 | Flow Computing | Parallel CPU acceleration IP | $15M–$25M | Low Confidence | Implied Valuation from Raise | Implied | $4M | Full Confidence |
| 80 | Esperanto Technologies | Many-core RISC-V inference chips | $10M–$30M | Low Confidence | Acquisition Value | Estimated | $63M | Partial Confidence |
| 81 | Wave Computing | Dataflow AI processor architecture | $5M–$25M | Low Confidence | Proxy-Based Estimate | Estimated | $203M | Partial Confidence |
| 82 | Luminous Computing | Photonic AI supercomputer processors | $5M–$15M | Low Confidence | Acquisition Value | Estimated | $123M | Partial Confidence |
| 83 | Zettascale | Reconfigurable dataflow AI accelerators | $4M–$8M | Low Confidence | Implied Valuation from Raise | Implied | $1M | Strong Confidence |
| 84 | Untether AI | At-memory AI inference processors | $0M–$5M | Low Confidence | Proxy-Based Estimate | Estimated | $152M | Strong Confidence |
| 85 | AlphaICs | Edge AI processor architecture | $0M–$5M | Low Confidence | Proxy-Based Estimate | Estimated | $11M | Strong Confidence |
| 86 | Vathys | Deep-learning processor architecture | $0M–$3M | Low Confidence | Proxy-Based Estimate | Estimated | $3M | Low Confidence |
| 87 | deepsilicon | Quantized AI inference hardware | $0M–$2M | Low Confidence | Proxy-Based Estimate | Estimated | $1M | Low Confidence |

This chart, featured in our AI chip market deck, compares the 2026 size of the AI chip market with other markets of similar size
Key valuation trends in the AI chip market
Insights
- The top 10 AI chip companies represent about 84% of the dataset’s total midpoint valuation, showing that market value remains heavily concentrated among a small number of scaled accelerator developers.
- Cambricon’s $128.6 billion market capitalization is about 322 times the median AI chip startup valuation, illustrating the unusually wide gap between public market leaders and emerging semiconductor companies.
- Moore Threads is valued at $45.8 billion after raising about $1.8 billion, while Cerebras is valued at $11.4 billion after absorbing $8.8 billion, showing that capital raised does not translate directly into market value.
- Habana Labs generated a $2 billion acquisition outcome from only $120 million of funding, equal to almost 17 times invested capital and one of the strongest realized outcomes in the AI accelerator market.
- Photonic AI companies cover almost the full valuation spectrum, from Celestial AI at $3.3 billion to Opticore near $40 million, which suggests broad investor interest but very different levels of commercial maturity.
- Celestial AI and Enfabrica show that strategic value is moving beyond processor cores. AI interconnect, networking, and memory infrastructure can attract billion-dollar outcomes by solving data movement bottlenecks.
- Positron AI, OLIX Computing, and Fractile reached valuations near or above $1 billion with less than $310 million raised each, highlighting investor demand for specialized AI inference alternatives.
- The median AI chip company is valued at approximately 4.2 times total funding, but capital efficiency varies sharply according to architecture, commercialization requirements, and whether the company focuses on training or inference.

This chart, featured in our AI chip market deck, shows how Nvidia is leading in AI chips
A few word about our methodology
As you can see, we built a database that ranks startups in the AI chip market based on their current valuation.
Estimating AI chip startup valuations is not always straightforward. Many semiconductor companies do not publicly disclose their valuation, and the available information can vary widely depending on the company and its stage.
To build this ranking, we applied a structured valuation methodology and cross-checked information across multiple reliable sources.
Whenever possible, we relied on direct disclosures. These include announced valuations from completed funding rounds, public filings for listed AI chip companies, or official acquisition prices.
When an AI chip company is publicly listed, we use its current market capitalization as the reference valuation.
If an AI chip company was acquired and no independent valuation can reasonably be estimated today, we use the acquisition price as the main reference point.
When an AI chip startup recently raised capital but the valuation was not disclosed, we estimate the implied valuation using typical dilution levels for that stage of fundraising.
In some cases, we also estimate valuations using operating metrics such as revenue, customer deployments, chip shipments, cloud usage, or commercial contracts, combined with valuation multiples from comparable AI semiconductor companies.
When direct financial data is not available, we may rely on carefully selected comparable startups and other signals such as hiring growth, investor quality, product benchmarks, manufacturing progress, customer traction, or strategic partnerships.
All estimates follow a strict evidence hierarchy. Recent funding rounds with announced valuations carry the most weight, followed by public market values, acquisition prices, strong operating metrics, and comparable company analysis.
We also carefully evaluate the age of every data point. Recent information carries more weight, while older data is treated cautiously and adjusted conservatively when necessary.
Whenever information is uncertain or incomplete, we clearly distinguish between confirmed facts and reasonable inferences.
Because AI chip valuation data is not always fully public, each startup in the ranking is assigned a confidence level based on the reliability, recency, and consistency of the available evidence.
Full confidence means the valuation is supported by strong and recent evidence. Strong confidence means the estimate is well supported but includes minor inference. Partial confidence means the estimate relies more heavily on indirect signals. Low confidence means available information is limited or inconsistent.
When confidence is lower, we take a more conservative approach by widening the valuation range. This helps reflect the uncertainty and increases the probability that the true valuation falls within the estimated range.
This reflects how we conduct all our research, including the work behind our report covering the AI chip market.
In a world where LLMs hallucinate and unreliable information is everywhere, our goal is simple: provide data you can trust.
If you want the full detail on a specific valuation estimate, feel free to contact us and we will gladly explain.
Finally, know that we update the dataset once per month, so come back here if you need fresh information.

This chart, featured in our AI chip market deck, shows annual funding in AI chip startups
Related blog posts
- The most recent news in the AI chip market
- The most recent funding news in the AI chip market
- The latest update in the AI chip market
- The evolution of funding activity in the AI chip market
- The main fundraising trends in the AI chip market
Who is the author of this content?
NEW MARKET PITCH TEAM
We track new markets so founders and investors can move fasterWe build living “market pitch” documents for emerging markets: from AI to synthetic biology and new proteins. Instead of digging through outdated PDFs, random blog posts, and hallucinated LLM answers, our clients get a clean, visual, always-updated view of what’s really happening. We map the key players, deals, regulations, metrics and signals that matter so you can decide faster whether a market is worth your time. Want to know more? Check out our about page. You can also follow us on Instagram or on Facebook.
How we created this content 🔎📝
At New Market Pitch, we kept seeing the same problem: when you look at a new market, the data is either missing, paywalled, or buried in 300-page reports that feel like they were written in the 80s. On the other side, LLMs and random blog posts give you confident answers with no sources, and sometimes they just make things up. That’s not good enough when you’re about to invest real money or launch a company.
So we decided to fix the experience. For each market we cover, we build a structured database and update it on a regular basis. We track funding rounds, fund memos, M&A moves, partnerships, new products, policy changes, and the real activity of startups and incumbents. Then we turn all of that into a clear “market pitch” that shows where the opportunities are and how people actually win in that space.
Every key data point is checked, sourced, and put back into context by our team. That’s how we can give you both speed and reliability: fast coverage of new markets, without the usual guesswork.