Which edge AI startup is growing the fastest?

Why is the edge AI startup race much easier to judge now?

The edge AI startup race is finally producing enough real commercial data for us to separate companies that are shipping from companies that are still mostly promising.

For years, comparing startups such as Axelera AI, DEEPX, SiMa.ai and Hailo meant comparing funding rounds, benchmark results, chip tape-outs and design wins. Those indicators told us who had credible technology, but very little about who was actually growing fastest.

That has changed. Axelera AI says it now works with more than 500 customers and has been adding hardware and deployment partners. DEEPX recently disclosed 77 commercial purchase orders worth more than $13 million after one year of DX-M1 mass production. SiMa.ai is shipping its second-generation Modalix platform and has spent the past few months turning it into production-ready drone and industrial systems with partners such as ARK Electronics and AVerMedia. Hailo, meanwhile, has reached the point where Microchip has signed an agreement to acquire it.

The freshest developments reinforce that move toward commercialization. Axelera's latest announced collaborations include ASRock Industrial and PwC Middle East. SiMa.ai has recently announced several physical-AI and autonomous-drone integrations. DEEPX says more than 60% of its first-year DX-M1 orders arrived during the final four months of that period.

We now have something much more useful than a collection of funding announcements: customer growth, purchase orders, production availability and channels that can put these chips into actual machines.

What does "fastest-growing edge AI startup" actually mean?

For an edge AI startup, we think the best measure of growth is how quickly customers are moving from evaluation to production and spending money.

Revenue would normally settle the question. Private chip companies rarely disclose enough revenue, however, and comparing the few numbers that do appear would create a distorted ranking.

We therefore give the most weight to commercial purchase orders, recognized revenue where available, production shipments and clearly defined customers. Distribution and hardware integrations matter too, especially when they make products immediately orderable rather than merely creating another development partnership. Funding sits further down the list. A huge round can accelerate growth, but raising money is very different from selling chips.

This distinction produces several different leaders. Axelera has the broadest disclosed customer footprint. DEEPX has the clearest recent acceleration in commercial orders. SiMa.ai is expanding its production ecosystem quickly but gives us little customer-volume data. Hailo appears more commercially mature than most of the group, although its pending acquisition makes it a weaker fit for an independent-startup ranking.

What we measure	Current leader	Why
Disclosed customer reach	Axelera AI	More than 500 customers
Commercial order acceleration	DEEPX	77 purchase orders worth $13M+
Recent ecosystem expansion	Axelera AI / SiMa.ai	Rapid stream of production partnerships
Mature installed business	Hailo	100+ current customers and 10,000+ developers
Capital available to scale	Axelera AI	$250M+ latest round

Is Axelera AI the fastest-growing edge AI startup right now?

Yes. Axelera AI currently has the strongest overall case because customer growth, production availability, partner expansion and financing are all moving in the same direction.

The cleanest starting point is customer reach. A company presentation in late 2025 showed more than 150 customers and a commercial pipeline above €300 million. A few months later, Axelera announced that it was shipping to its 500th global customer. Even if we use 150 as the starting point, the disclosed customer count increased by roughly 233%.

We would be more cautious if that figure stood alone. It does not. Axelera's Metis-based products are commercially available, its accelerator technology appears inside systems from established embedded-computing vendors, and its partner list now stretches across hardware, software, distribution and integration.

The expansion has continued lately. ASRock Industrial introduced systems using Axelera acceleration with up to 214 TOPS of additional inference capacity. Axelera's newest Middle East collaboration with PwC is aimed directly at helping regional organizations move AI workloads into practical deployments. Earlier additions included companies such as Advantech, SECO, Prodrive Technologies and a much wider group of distributors and software partners.

Axelera also raised more than $250 million after this commercial ramp had already begun. The company says total capital attracted since 2021 now exceeds $450 million across equity, grants and venture debt. That money gives Axelera more room to manufacture products, support customers and keep widening its channel while smaller rivals are still financing the transition into volume production.

Taken together, Axelera has the broadest growth profile we can currently verify.

If you want more recent data on this point, please see our latest edge AI market report.

Are Axelera AI's 500 customers really as impressive as they sound?

Axelera AI's 500-customer figure is genuinely impressive, although we should read it as evidence of commercial reach rather than assume 500 large production deployments.

Customer counts get messy in semiconductors. One company may count a buyer of a development kit. Another may wait until a customer has signed a production contract. A third may talk about design wins, evaluations or deployed companies. Those categories can look identical in a headline while representing wildly different amounts of revenue.

Axelera says it has shipped to more than 500 customers across manufacturing, retail, defense, public safety, agritech, robotics and security. The European Innovation Council also described the company as deployed across more than 500 customers when discussing its latest financing. Axelera's current media materials still use the 500-plus figure, so this does not look like a one-off number that quietly disappeared.

There is also evidence beyond the customer count. Advantech says it has integrated Axelera's Metis processor into its products for manufacturing automation and intelligent infrastructure. SECO has incorporated Metis into its edge AI product line. Axelera's current partner page includes dozens of companies spanning hardware manufacturers, distributors, model providers and systems integrators.

That makes the 500 figure more convincing. We still do not know how many of those customers are buying hundreds or thousands of units, which prevents us from turning the number into a revenue estimate.

The useful interpretation is fairly simple: Axelera has already built a much wider commercial funnel than most private edge-AI chip startups, and enough of that funnel has moved into shipping products for us to take the growth seriously.

Is DEEPX actually growing faster than Axelera AI on real orders?

Yes. If we judge the race purely by recent purchase-order acceleration, DEEPX currently has the stronger growth curve.

DEEPX began mass production of its DX-M1 edge AI processor in 2025. After roughly seven months, the company reported 27 commercial purchase orders across eight countries. Its latest first-year update puts that total at 77 orders worth more than $13 million across more than ten countries and regions.

The pace inside that year is the interesting part. According to DEEPX, more than 60% of those orders arrived during the final four months. ChosunBiz reported the same concentration when covering the company's latest figures.

So DEEPX did much more than accumulate 77 contracts at a steady rate. The commercial curve steepened after customers had more time to test the production chip.

Earlier reporting gives us another view of the acceleration. DEEPX had secured only two orders during the first several months of its commercialization period. That became 27, then more than 30, and eventually 77. By April, management was talking about roughly $6.6 million in purchase orders. The latest figure exceeds $13 million.

This is unusually useful data for a private semiconductor startup because we can watch the funnel turn into signed commercial business rather than infer demand from partnerships.

Axelera remains ahead in overall scale and breadth. DEEPX, though, is currently giving us the fastest hard order growth.

If you want more recent data on this point, please see our latest edge AI market report.

How big is DEEPX's $13 million order book really?

DEEPX's $13 million order book is still small beside established semiconductor companies, but for a startup completing its first year of mass production, the speed of the ramp is more interesting than the absolute number.

More than $13 million spread across 77 purchase orders gives a simple average of roughly $169,000 per order. We should not treat that as a typical contract size because a few large orders can pull the average upward, yet it gives us the right order of magnitude. DEEPX has moved beyond a business made entirely of $500 development boards.

There are already examples of larger deployments. At an April press conference, CEO Lokwon Kim said Baidu had placed an initial order for roughly 40,000 chips. DEEPX has also been working with hardware companies and distributors including AAEON, Advantech, ASRock, MSI, Avnet and WPG.

The geographic mix makes the $13 million more credible. The latest company figures cover South Korea, the United States, China, Europe, Taiwan, Hong Kong, Japan, Singapore and India. More than half of the commercial activity has come from outside DEEPX's home market according to earlier reporting.

The weak point is still obvious: purchase orders are not recognized revenue, and $13 million remains a modest base. One or two large accounts can also move a young chip company's numbers dramatically.

Even with those limits, almost doubling disclosed order value from roughly $6.6 million to more than $13 million within a few months is hard to dismiss. DEEPX looks like it is entering the steep part of its commercialization curve.

DEEPX commercial ramp	Purchase orders	Disclosed value
Early production ramp	2	Not disclosed
After roughly seven months	27	Later reached about $6.6M
After one year	77	More than $13M

Why are DEEPX orders suddenly coming in so much faster?

DEEPX's order growth is accelerating because hundreds of earlier chip evaluations are finally feeding into a product that customers can buy in volume.

DEEPX spent years putting pre-production silicon in front of potential customers. The company has said more than 300 organizations evaluated its prototypes, while later comments from management put the broader evaluation pool above 400 companies.

That created a large backlog of potential customers before DX-M1 mass production even started.

The supporting ecosystem has grown alongside it. At COMPUTEX this year, DEEPX products appeared across more than 30 partner booths, including Advantech, ASRock, MSI, AAEON, QNAP, Lanner, IEI and DFI. Global distributors included Avnet and WPG. DEEPX has separately talked about 27 semiconductor distribution partners and more than 50 hardware and module partners.

AAEON provides a good example of how that changes the commercial opportunity. The companies signed a three-year cooperation covering industrial computers, single-board computers and edge gateways. Once an NPU becomes available inside hardware that industrial customers already know how to buy, every sale requires less bespoke engineering from the chip startup itself.

So the acceleration has a fairly concrete explanation. Several years of evaluations built the funnel, mass production opened the gate, and a rapidly expanding hardware channel is making conversion easier.

Is SiMa.ai growing fast enough to challenge Axelera AI and DEEPX?

SiMa.ai is clearly accelerating its commercial push, but today we still lack the hard customer and order numbers needed to put it ahead of Axelera AI or DEEPX.

The company has already crossed an important technical threshold. Its second-generation Modalix MLSoC is in production, and SiMa.ai sells a system-on-module designed as an alternative to commonly used GPU modules. Modalix supports workloads ranging from conventional computer vision to transformers, LLMs and generative AI while operating under tight edge power limits.

What has changed more recently is the volume of deployment partnerships around that hardware.

SiMa.ai launched Palette Neat in June to reduce the work needed to build physical-AI applications. It then announced an integrated drone solution with AVerMedia that is available for customer evaluation with production quantities supported. A collaboration with Mistral Solutions targets autonomous drones, while its newest ARK Electronics partnership packages Modalix into production-ready drone and robotics electronics for commercial, industrial, agricultural and defense applications.

The pattern is clear enough: SiMa.ai is trying to shorten the gap between somebody liking its chip and somebody shipping a machine containing it.

We still have a measurement problem. SiMa.ai has not published anything comparable with Axelera's 500-plus customer count or DEEPX's 77 commercial purchase orders. Its statements about demand, production and partnerships tell us that the business is moving forward; they do not tell us how quickly customers are buying.

For now, SiMa.ai belongs among the strongest challengers rather than at the top of the ranking.

Which edge AI startup has the strongest sales and distribution machine?

Axelera AI currently has the strongest overall route to market, while DEEPX is building its own channel unusually quickly.

For edge AI chips, distribution is easy to underestimate. Most industrial companies do not want to buy a bare accelerator and build the entire computing stack themselves. They buy industrial PCs, modules, servers, cameras, gateways and robotics hardware from vendors they already use.

Axelera has pushed hard into that layer. Its current ecosystem includes companies such as Advantech, Aetina, ASRock Industrial, Axiomtek, Dell Technologies, Lenovo, Macnica, Prodrive Technologies, SECO and Ultralytics alongside distributors and systems integrators. Recent collaborations keep extending the network rather than simply repeating the same early partners.

DEEPX is following a similar path from a smaller base. Its products were shown through more than 30 hardware and integration partners at COMPUTEX, and the company has built relationships with Avnet, WPG, AAEON, Advantech, ASRock, MSI, Lanner and several other embedded-computing vendors. The three-year AAEON cooperation is particularly useful because it aims at broad product integration rather than a single reference design.

SiMa.ai is doing more of this lately as well. ARK Electronics, AVerMedia and Mistral Solutions give it ready-made routes into drones and other physical-AI systems.

The gap between these companies therefore comes partly down to maturity. Axelera already has the widest visible network and the largest disclosed customer pool. DEEPX's channel is younger, but its recent order acceleration suggests that the network is beginning to convert.

If you want more recent data on this point, please see our latest edge AI market report.

Has Hailo already moved beyond the edge AI startup race?

Hailo has effectively reached a different stage of the edge AI race because Microchip has signed a definitive agreement to acquire the company.

Before that deal, Hailo had one of the strongest claims to commercial leadership. Its processors had spread into cameras, industrial systems, Raspberry Pi products and other edge devices. Hailo also expanded from traditional vision inference into local generative AI with the Hailo-10 family.

The acquisition announcement gives us a rare outside description of its commercial scale. Microchip says Hailo brings more than 100 current customers and a developer community exceeding 10,000 users, alongside multiple accelerator and vision-processor product families.

Earlier Hailo material had referred to deployment across more than 300 customers. We cannot calculate a meaningful growth rate by comparing those two figures because the definitions clearly differ. "Deployed by" can include organizations that would not necessarily qualify as current commercial customers under an acquisition disclosure.

Microchip expects the transaction to close subject to customary conditions and regulatory approvals. That makes Hailo highly relevant as a benchmark for what a successful edge-AI startup can become, but awkward as the answer to which independent startup is growing fastest today.

The acquisition also tells us something about the whole category. A large embedded-semiconductor supplier sees enough value in edge AI accelerators, vision processors, software and customers to buy one of the sector's established specialists outright. That is a stronger commercialization marker for the market than another venture round.

Would Blaize be the fastest-growing edge AI company if we included public companies?

Blaize would have looked like the obvious winner based on its 2025 revenue growth, but its latest financial results show exactly why one explosive year can mislead us.

Blaize reported $38.6 million in revenue for 2025, up from just $1.6 million in 2024. That is roughly 24 times as much revenue in a single year. Its fourth quarter alone produced $23.8 million.

The picture in 2026 has been far rougher.

First-quarter revenue dropped to $2.7 million. The second quarter recovered to $12 million, six times the $2 million generated a year earlier, and first-half revenue reached $14.7 million versus $3 million in the comparable period.

Then came the important part. Blaize cut its full-year revenue outlook to $40 million-$43 million because some commercial engagements did not convert into orders and higher memory prices created additional pressure. The company had previously been pointing investors toward roughly $130 million.

That is a dramatic reset.

Blaize does have fresh business. Its latest results mention a first European order involving several thousand branded units and a binding 2,000-server agreement worth roughly $70 million at current memory prices, with about $20 million expected to be recognized during the second half and most of the remainder in 2027.

Yet the guidance cut shows why we have been strict about separating pipelines, engagements and partnerships from actual orders. A young edge-AI company's apparent growth can change enormously depending on when a handful of large programs convert.

Blaize is also Nasdaq-listed now, so we exclude it from the startup winner. Its public numbers are still extremely useful because they give us a reality check that private-company announcements rarely provide.

If you want more recent data on this point, please see our latest edge AI market report.

Are edge AI startups moving beyond computer vision now?

Yes. Axelera AI, DEEPX, SiMa.ai and Hailo are all broadening beyond the original edge-AI market of running vision models on cameras.

Computer vision remains a big part of the business. Manufacturing inspection, surveillance, worker safety, retail analytics and smart-city cameras are already deployable and relatively easy to understand.

The newer products increasingly target robots, drones and systems that need transformers, language models or multimodal models locally.

Hailo moved into local generative AI with Hailo-10. SiMa.ai designed Modalix for CNNs, transformers, LLMs and generative workloads and is now pushing it heavily into autonomous drones. Axelera's product roadmap has expanded beyond its original Metis platform toward more capable architectures designed for larger inference workloads. DEEPX is developing DX-M2 on Samsung's 2-nanometer process, with the company targeting much more demanding generative-AI workloads at very low power.

Recent partnerships match the product roadmaps. SiMa.ai's activity has tilted heavily toward drones and physical AI. Axelera is working across robotics, defense and industrial systems. DEEPX says its first-year orders span robotics, smart factories, mobility, aerospace, healthcare, security and edge computing.

We are therefore seeing a bigger addressable market emerge around the same chips. The question for these startups is increasingly whether they can become the low-power computing layer inside autonomous physical systems rather than simply accelerate YOLO models inside cameras.

That expansion gives the current leaders more room to sustain growth, although it also puts them closer to Nvidia, Qualcomm, NXP and other semiconductor companies with much larger distribution networks.

Could DEEPX overtake Axelera AI soon?

Yes. DEEPX is the company most likely to knock Axelera AI off the top spot because its recent purchase-order curve is steeper than Axelera's publicly disclosed commercial metrics.

DEEPX went from 27 commercial orders after roughly seven months of mass production to 77 by the end of its first year. That is about 185% growth in the number of disclosed orders over only a few months. More than 60% of the first year's orders arrived during the final four months.

Its disclosed order value has accelerated too. Management discussed roughly $6.6 million of purchase orders in April. The latest figure exceeds $13 million. We are looking at close to a doubling over a short period.

As seen above, the company also has hundreds of earlier evaluations feeding into more than 30 visible hardware partners and a much wider module and distribution network. That gives DEEPX several more opportunities for those early orders to turn into repeat production volumes.

The remaining gap is scale. Axelera says it already works with more than 500 customers, has a broader established channel and has just added more than $250 million of financing. DEEPX has 77 commercial orders worth $13 million. Those numbers describe companies at different points on the curve.

What would change our answer is straightforward. If DEEPX's next set of disclosures shows another large jump in both order value and repeat production orders, we would have much less reason to favor Axelera's broader customer count.

Right now, DEEPX is closing the gap faster than any other company we found.

Which edge AI startup is growing the fastest today?

Axelera AI is the fastest-growing edge AI startup overall today, while DEEPX is growing faster on the hardest commercial metric we can actually observe: purchase orders.

Axelera has the broader case. Its disclosed customer base jumped from more than 150 to more than 500, its hardware is commercially shipping, major embedded-computing companies are integrating the technology, its partner network keeps expanding, and more than $250 million of fresh financing gives it plenty of capacity to support a larger production business.

The latest activity has not broken that pattern. Axelera has continued adding routes into industrial systems and regional enterprise deployments after announcing the 500-customer milestone. The company still describes itself as working with more than 500 customers today, so we have no fresh evidence that the commercial ramp has stalled.

DEEPX makes the decision much closer. Seventy-seven purchase orders worth more than $13 million in the first year of mass production would already be encouraging. The concentration of more than 60% of those orders in the final four months is what makes DEEPX especially interesting. Demand appears to be accelerating as evaluation programs convert and its hardware network expands.

SiMa.ai is advancing quickly around drones and physical AI, but the company still gives us too little volume data to declare it the fastest-growing player. Hailo has reached the acquisition stage with Microchip. Blaize's public financials show both how quickly an edge-AI business can scale and how quickly early growth can become lumpy.

So we would put Axelera first today and DEEPX second, with a much smaller gap between them than the headline customer counts suggest. Axelera has the strongest combination of scale, customer expansion, shipping products and distribution. DEEPX has the sharper acceleration curve.

If the next meaningful DEEPX update shows that the recent order surge has continued, this ranking could flip.

If you want more recent data on this point, please see our latest edge AI market report.

Who is the author of this content?

NEW MARKET PITCH TEAM

We track new markets so founders and investors can move faster

We build living "market pitch" documents for emerging markets: AI, synthetic biology, new proteins, and more. Instead of outdated PDFs or hallucinated LLM answers, our clients get a clean, visual, always-updated view of what's really happening: key players, deals, regulations, and signals that matter. Learn more about us.