Who’s Building the Teams That Build the Future? Recruiting Roundup - October 27
Thirty-five AI infrastructure companies are hiring fast right now. Here is who they are, sorted by the part of the machine each one builds.
Every chatbot answer, every AI-written email, every generated image has to run somewhere. Behind the friendly text box sits a long chain of physical things: chips, cables, power, buildings, servers, and the software that keeps them all busy. For most of the past few years, people talked about that chain the way they talk about plumbing. Rarely, and usually only when something broke.
That has changed. The companies building this infrastructure are now some of the fastest-growing employers in tech. If you want to understand where the jobs are going, you have to look underneath the apps.
A good place to start is a list. Each week, Grace Gong publishes On My Radar, a roundup of companies she is watching closely. This week's edition named 35 AI infrastructure companies that are actively hiring, ranked by how much their headcount grew over the past 90 days. She sourced it with help from Harmonic, the startup database. Full credit to both.
Grace ranked the list by growth. Below, we've regrouped the companies by what they actually make, from the silicon at the bottom to the software at the top. The number after each name is its spot on her original ranking, so you can still see who is growing fastest.
The Chipmakers
For a decade, one company's graphics chips did most of the heavy lifting in AI. These eleven startups are betting that the next decade belongs to chips designed for one job: running AI models after they've been trained. The industry calls this inference.
Training happens once. Inference happens every time someone asks a question, which is why so many founders want to make it cheaper.
This is the largest group on the list, and the most international. Seoul, London, Eindhoven, and Shanghai sit alongside Silicon Valley. If you're a chip designer, verification engineer, or compiler specialist, this is your section.
- SambaNova: AI chips and its own cloud to run them. The fastest grower on the whole list.
- Positron AI: inference hardware, built in Reno rather than the Bay Area.
- Fractile: inference processors out of London.
- Etched: custom chips built for one kind of AI model, sold as whole clusters.
- EUCLYD: efficient silicon for foundation models, from Eindhoven, the Dutch chip town.
- FuriosaAI: AI inference chips from Seoul.
- MatX: chips built specifically for large language models.
- 沐曦集成电路有限公司 (MetaX): GPU chips and compute from Shanghai.
- Tensordyne: efficient inference systems.
- d-Matrix: compute for running generative AI.
- HyperAccel: generative AI inference chips, also from Seoul.
The Plumbers
A fast chip is useless if data can't reach it quickly. As AI systems link thousands of chips together, the bottleneck has moved to the spaces in between: the wires, the connections, and the handoffs. These four companies work on that problem, and several use light instead of electricity to move data. This is niche, deep-tech hiring. Photonics engineers, packaging specialists, and systems architects will find the most opportunities here.
- TYLsemi: chiplet silicon for AI, meaning small chip pieces that snap together into bigger ones. Second-fastest grower overall.
- Volantis Semiconductor: photonic interconnect, linking AI chips with light.
- Mixx Technologies Inc: silicon photonics for AI.
- MangoBoost: acceleration that helps whole AI systems run more efficiently.
The Landlords
Someone has to own the buildings full of chips and rent them out. The industry calls these companies "neoclouds," but it may be simpler to think of them as landlords. They buy expensive hardware, find power and space, and lease computing time to AI teams that don't want to build their own infrastructure.
This is the biggest hiring category on the list, and the most varied. Data center operators, site and energy specialists, network engineers, and salespeople are all in demand. Some, like Firmus, compete on energy efficiency. Others, like TensorWave, compete by using a different chip supplier than everyone else.
- Dapple: dedicated AI infrastructure for large companies.
- Highreso Co., Ltd.: GPU cloud from Tokyo.
- Nscale: full-stack AI cloud, based in Norway.
- Volta: vertically integrated AI infrastructure, meaning it handles every layer itself.
- Firmus: energy-efficient AI infrastructure.
- TensorWave: an AI cloud built on AMD chips.
- Hydra Host: bare-metal GPU servers, rented with nothing pre-installed.
- GMI Cloud: AI cloud for production workloads.
- QumulusAI: a distributed AI cloud.
- Nava: an AI-native cloud platform from Bangalore.
- NexGen Cloud: full-stack AI cloud out of London.
- Lambda: GPU cloud and "AI factories." One of the older names here, and still growing.
The Brokers
Where there are landlords, there are agents. These three companies don't necessarily own the most hardware. Their skill is finding computing power and getting it to the teams that need it, at the right size and price.
It's a matchmaking business, and in a market where chips are scarce, a good matchmaker is valuable.
These are good fits for people with operations, partnerships, or marketplace backgrounds, alongside engineers.
- Boost Run: AI compute provisioning, from Northbrook, Illinois.
- Parasail: compute sourced for inference.
- Andromeda: compute for AI teams.
The Toolmakers
At the top of the stack is software. A rack of GPUs is a bit like a commercial kitchen: impressive, expensive, and useless without someone running the line.
These five companies make the tools that let developers use AI hardware without becoming hardware experts. They also help companies keep their expensive chips busy instead of sitting idle.
This is the most familiar territory for software engineers, developer relations folks, and product managers. If you've written code but never touched a server rack, start here.
- Modal: serverless inference and training. You write the code; it handles the machines.
- Runpod: a cloud built for AI developers.
- hosted·ai: GPU orchestration, deciding which jobs run where.
- Lightning AI: an AI development platform, from the team behind the PyTorch Lightning open-source library.
- Rafay: software for managing GPU platforms.
What the List Tells Us
Read the list by growth and it looks like a scramble. Read it by layer and a pattern appears.
Two of the three fastest-growing companies, SambaNova and TYLsemi, work at the very bottom of the stack, in silicon. The money and the hiring are flowing toward some of the hardest problems in AI: making it cheaper to run, one chip and one connection at a time.
The other lesson is that "AI jobs" no longer means only researchers. These companies need people who understand data centers, supply chains, energy, operations, and customers just as much as they need model builders.
If one of these companies catches your eye, click through to its LinkedIn page for open roles.