Feature
It’s hard to live in current times without knowing who and NVIDIA, AMD and Intel are, and to have at least a vague idea of what they do and why it matters to the development of artificial intelligence. AI is a hungry creature. It wants speed, power and faster connections. To feed this need, a host of startups have sprung up to build better connections to the offerings built by the top AI chip companies, to find better ways to store data, faster ways to move it and clever methods of eliminating the bottlenecks that slow the tech's thought processes.
In the world of nascent backbone architecture, it is a wild west out there. It is like the early days of the internet, and innovation is rampant. It’s hard to keep up, but here are the top upstart AI chip companies — and the technologies they are inventing — that should be on your radar.
1. Cerebras Systems
Cerebras Systems is a start-up based in Sunnyvale, California that builds what it calls the world’s fastest AI accelerator, a small form-factor, 15-rack system designed to speed large language models. Customers around the world use these systems to power AI for enterprises, military intelligence, drug discovery, energy exploration and scientific and language research.
The company’s investors are a virtual who’s who of AI and tech development, with a roster including Sam Altman, Andy Bechtolsheim, Alpha Wave Ventures and Falcon Edge Capital. The company likens current days in AI to those of the early internet, where people were getting online via dialup and, as in those days, processing speed is the thing that will take AI to the next level.
2. Etched
The brainchild of Harvard drop-out Gavin Uberti, Etched claims that its chip — named Sohu — will outperform NVIDIA’s and massively cut inference costs by embedding deep learning architecture directly onto the chip.
The company has recently partnered with Decart to create Oasis, a playable AI model that generates open world games. The company claims that “by burning the transformer architecture into our chips, we can run AI models on an order of magnitude faster and cheaper than GPUs.” The company has raised $120 million in funding and is recruiting talent.
Related Article: 10 Top AI Chip Companies
3. Astera Labs
Astera Labs plans to deliver purpose-built connectivity for AI and cloud infrastructures. The company’s Scorpio X-Series Fabric Switches deliver connectivity between the various parts of AI servers. The company’s switches, controllers and software-defined architecture speed and control the way that a server "thinks," allowing the GPUs, CPUs and other components to work together in a unified way by managing the way they access, pool and store memories.
The company claims that its solution is reducing or eliminating the bottlenecks that impede data delivery and memory connectivity and that leave the high-powered chips in AI servers operating at about 50% of capacity. The company has raised significant funds.
4. Lightmatter
Lightmatter is a world leader in photonic supercomputing, high-performance computing that uses light instead of electrons to process information. Its Passage platform allows incredible bandwidth density and enables AI systems to rapidly scale up. Passage is the world's first 3D-stacked silicone photonics engine and can connect thousands — or even millions — of processors at the speed of light, eliminating bottlenecks and improving scalability and efficiency for advanced AI and high-performance computing.
The company recently announced a partnership with Amkor Technology, Inc to use Passage to speed AI processing.
5. Axelera
Axelera is, according to the company, revolutionizing the field of AI by building a hardware and software platform to accelerate computer vision on edge devices.
The company plans to offer a range of accelerator cards built on its Metis AIPU chip that developers and integrators can use for hardware acceleration in just about any data-heavy industry. Many of these accelerators are currently available for preorder. The company is backed by investors such as the European Innovation Council Fund, Innovation Industries, Invest-NL and more. The company demonstrated its Metis product line at CES 2025 and was named a CES Honoree for its Metis Compute Board.
6. AyarLabs
AyarLabs taps optical I/O technology — a process that transmits and processes data using light — to enhance link density between AI nodes and thereby reduce communication bottlenecks. In short, it's building a connector for AI nodes that moves data with light. Eliminating these bottlenecks lets the nodes connect at scale and perform as if they are one giant GPU. This leads to much greater bandwidth, lower latency and more power efficiency.
The company’s investors and partners include NVIDIA, AMD, Intel, Hewlett Packard Enterprise and Lockheed Martin.
7. EdgeQ
Santa Clara, California-based EdgeQ is disrupting the mobile chip space with its Base Station on Chip. This tiny acceleration card for AI can be used as an element to create all sorts of network technologies, especially mobile ones.
The company envisions an open paradigm for 5G, allowing customers to deploy 5G through software. The company recently partnered with wired and wireless connection company Actiontec and announced the commercial availability of a new wireless small cell.
8. Eliyan
Eliyan’s NuLink technology is designed to eliminate the “memory wall,” the gap between the speed of the processor and the speed of its memory. These walls happen when the processor is faster than the memory. NuLink does this by allowing for super large SiPs on Standard Packaging. Eliminating the mismatch between processor and memory helps AI perform as much as ten times faster. This technology will speed AI as well as automotive, gaming, consumer electronics and server applications.
The company has raised significant funding and is led by Ramin Farjadrad, who holds many patents in communications and networking.
9. Pliops
Behind all this AI is a lot of data. Where do you put all that data? Pliops is a startup based in Santa Clara, California, that aims to tackle this problem and build data storage systems that are up to the heavy demands AI is creating.
Led by Ido Bukspan, previously senior VP of the Chip Design Group at NVIDIA, and with a roster of tech investors and partners that include AMD, Intel, Lenovo, Dell, Oracle and NVIDIA, the company is considered by many experts to be a startup worth watching. The company’s Pliops XDP is, according to the company, the new benchmark for application acceleration. It radically improves how data is processed and SSD storage is managed.
Related Article: What NVIDIA's No. 1 Market Cap Tells the AI Industry
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
10. SiMa.ai
SiMa is determined to make machine learning effortless for the technology industry and help scale it at the embedded edge. To do this, the company aims to address the problems of scale, performance and energy efficiency with a software-first solution that makes machine learning easy to use.
The system-on-a-chip (SoC) solution addresses computer vision problems and improves performance. And it has a push-button software user experience that does not require manual interventions.
