The story of AI hardware over the past two years hasn't been pretty — Humane sold to HP and shut down, Rabbit laid off half its staff, Friend was heavily criticized after shipping. One common problem: hardware companies don't know how to build software.
Era, which just raised $11 million, wants to be that layer.
First, the funding
- Total: $11M
- Series A seed: $9M, led by Abstract Ventures and BoxGroup, with participation from Collaborative Fund and Mozilla Ventures
- Earlier pre-seed: $2M, from Topology Ventures and Betaworks
The angel lineup is even more notable: Flickr co-founder Caterina Fake, iPhone keyboard creator Ken Kocienda, OAS founder Tony Wang, former Rabbit CPO ShaoBo Z, and Poetry Camera's Kelin Zhang — a group that essentially represents the "creator class" of consumer electronics and mobile software over the past 20 years. The fact that a former Rabbit CPO is also betting on a company building a software layer for AI hardware is itself a telling signal.
The three founders' resumes say more about the positioning than the company itself
- Liz Dorman (CEO): Previously worked on AI orchestration at Humane, then joined HP after Humane was acquired
- Alex Ollman (CTO): Came from HP, where he worked on enterprise-grade agentic frameworks
- Megan Gole (CPO): Previously at Sutter Hill Ventures, involved in the io project with Jony Ive and Sam Altman
The common thread among the three: they all directly worked on, or closely observed, how AI hardware fails. Humane and io were two of the most closely watched attempts in this wave of AI hardware — the former proved hardware is a dead end, and the latter, which OpenAI acquired for $6.5 billion, is still cooking.
Dorman's own assessment is blunt:
"I want to regain the right to choose my own devices."
She wants Era's software layer to "replace the app layer" — meaning, in the future, your glasses, ring, or pendant won't run a bunch of apps, but will connect to a unified intelligent layer.
What Era actually sells
Not hardware, but a software platform for hardware makers:
- Cross-model routing: Connects to over 130 LLMs from more than 14 providers
- Multi-modal input processing: Voice, images, and sensor data all supported
- Dynamic routing: Automatically selects models based on task, network, and cost
- Connectivity management: Handles offline or weak network scenarios
- Scalable to millions of devices
The positioning sounds like "iOS for AI hardware," but a more accurate description is "Vercel for AI hardware" — you build the hardware, and Era handles model inference, routing, fallback, and billing.
Target hardware form factors: glasses, rings, pendants, and smart speakers.
Why now
The biggest cognitive shift in this wave of AI hardware is that everyone has finally admitted one thing: hardware itself cannot become the entry point for AI.
The reasons are practical:
- The cost structure is inverted: Humane's pin sold for $699 and still lost money because inference costs kept burning
- No one wants to pay a monthly fee for a single AI gadget: Users already pay for ChatGPT, Claude, and Gemini — they won't pay another subscription for the "AI" in their glasses
- Models iterate twice a year, hardware may not change for three years: Hardware makers simply can't keep up with updating their own model stacks
This is exactly Era's entry point: hardware makers need a software layer that can continuously update models, automatically choose the cheapest routing, and avoid being locked into a single provider — building it themselves would be too costly and offer poor ROI.
By analogy: in the early smartphone era, no one could make every phone maker write their own iOS, so Android became the platform for hardware makers. Era wants to occupy the Android-like niche for AI hardware.
But there are real problems
First: 130 models sounds impressive, but how many can actually run on hardware? On-device inference has completely different requirements for model size, quantization, and latency compared to the cloud — simply connecting APIs isn't enough.
Second: Are hardware makers willing to be locked into a third-party software layer? When Anthropic's Claude Design recently caused Figma's stock to drop 6.8%, and its CPO resigned from Figma's board that same week — hardware makers are very sensitive to having someone else control their software layer.
Third: $11 million won't last long given the burn rate of AI hardware. Era's next round will need commitments from several named customers, or the story won't hold.
In a nutshell
All the detours AI hardware has taken over the past two years prove one thing: this business requires a team that understands both hardware and model orchestration. Era's team, based on their backgrounds, barely has traces of both. But whether the business can succeed depends on whether the first batch of hardware running on Era's platform, a year from now, is actually usable.
Sources: CocoLoop, Era raises $11M to build a software platform for AI gadgets (TechCrunch); Era raises $11M from Abstract Ventures and BoxGroup to bring AI to smart gadgets (TechFundingNews)