If you were building a consumer hardware startup in 2026, how would you do it? - Can someone understand it in five seconds? - Will they still use it in six months? - What happens when a cheaper copy shows up? @ritwikpavan says answer all three before you commit capital.

If you’re at #PDX2026 this week, come visit us at Booth 3! Tomorrow: Ryan Stevenson is up at 11 AM with a session on the AI-augmented engineer, and @ngpadgett closes at 2 PM on launching a product into the wild. #PDX2026

Tag yourselves

Last week we hosted the SF Hardware Meetup at Flux HQ for #SFTechWeek. This is everything that happened 🎥 Thanks to @michaelraspuzzi for organizing, @ngpadgett and to @mantle3d, @saba_khalilnaji and @fjeg1987 for speaking.

Waterloo interns are human LLMs, it’s amazing

Last night we hosted the 138th SF Hardware Meetup at Flux HQ for #SFTechWeek Thank you @michaelraspuzzi for organizing, @ngpadgett for being emcee and @mantle3d, @saba_khalilnaji, @fjeg1987 and everyone who pitched at the open mic.

THE CONSUMER HARDWARE PLAYBOOK. If I were building a consumer hardware startup in 2026 this is how I’d do it. Start with three questions: Can someone understand it in five seconds? Why would they still use it in six months? What happens when a cheaper copy shows up? Getting attention only answers the first. You need to have a convincing answer to the other two before committing capital. Pick the category carefully. Earbuds, phones, and robot vacuums already have strong competitors. When you ask someone to try an unfamiliar brand, the improvement needs to matter enough for them to take that chance. Slightly better specs are a difficult place to start. Here’s my playbook. 1. Build around a routine people already have. You’re giving yourself a harder job when buying the product also means learning an entirely new habit. Start with something people already do and find the part they’d gladly hand over. Matic uses vision to navigate and clean your home. Eight Sleep adjusts temperature around your sleep. Tonal adapts resistance as you train. This is what makes AI interesting in hardware. The device can handle more of the decisions and adjustments someone used to make themselves. But count all the work it creates too. If someone saves ten minutes using it and spends fifteen getting it to cooperate, you still have a product problem. 2. Find out whether strangers will pay before you build. Put together a mockup, a landing page, and the price you actually expect to charge. Then run Meta ads to people who have no connection to you or the company. Be transparent about what’s built, what’s planned, and how refunds work. You can learn a lot without pretending the product is ready to ship. Email signups are a useful start. Refundable reservations tell you more, but you still have to find out how many of those people finish buying. If you spend $20 in ads to get each reservation and only one in four becomes a purchase, you’ve spent $80 in ads per buyer, before other marketing costs. Simultaneously, talk to engineers and manufacturers. A product people want at $199 might be a very different business if you discover it needs to cost $399. Have that conversation while you can still change the design. 3. Win the one liner before focusing on the bill of materials. Try explaining the product to someone, then listen to how they explain it to somebody else. You’ll learn pretty quickly which part they understood and whether they thought it was worth mentioning. If the explanation needs a full read out of the spec sheet, you probably need to work on it. The same message should carry into the demo. Show the robot getting around a messy room or the resistance changing during a workout. Give people a reason to care before asking them to appreciate the engineering. Then test which images and videos actually lead to purchases. An impressive demo is a great starting point, but it still needs to prove itself as an ad. 4. Plan for someone to offer similar hardware for less. Take a capable competitor in China seriously from the beginning. They may have good engineers, good software, and a better cost structure. What would make someone keep choosing you? - Oura builds insights around your health data. - Eight Sleep uses your sleep history and temperature preferences. - Tonal tracks your training progress. - Skylight becomes part of how a family keeps its week organized. These are useful examples because the experience extends beyond the device itself. Still.. “we’ll have more data” needs a better explanation. What gets better because of that data? Does the customer notice? How long would it take someone else to catch up? Build around those answers, along with reliable hardware, useful integrations, and support customers trust. Being first gives you a head start, so use the time to build advantages that last. 5. Raise for the cash cycle. Super important. Most hardware companies fail because they didn't manage their cash properly. Map out when money leaves the business and when you expect it to come back. Factory deposits, R&D, freight, refunds, and the next production order can overlap in ways that make a healthy sales number misleading. You can sell out and still struggle to pay for the next batch. Raise enough to reach a meaningful next milestone, including testing, production, support, and some room for things to go wrong. Debt is a whole separate task. Inventory financing and venture debt can depend on different sources of repayment. Understand exactly what pays back the loan and what happens if that money arrives late. Before signing, run the numbers with slower sales, a delayed shipment, and a batch that needs replacing. Always negotiate net terms with your suppliers. 6. Keep using Meta to learn what makes people buy. Once the product exists, test real demos, customer experiences, offers, and landing pages. Keep looking for the explanation that helps the right customer understand why it’s worth the money. When an early campaign works, check what happens as you spend more and reach beyond the most enthusiastic buyers. Your CAC target has to fit what’s left after manufacturing, shipping, fees, returns, and support, with money left for the team and future R&D. If subscriptions make the numbers work, be honest about retention and what it costs to serve customers. Revenue expected over several years won’t pay a factory invoice due next month. As spending grows, use controlled tests where practical to check how many additional sales the ads actually create. 7. Work with creators whose audience has a reason to care. For training hardware, look for a coach who can explain where it belongs in a real program. For a family calendar, find someone who can show what happened after their household used it for a few weeks. Give them enough time to form an opinion. They may find a use case, objection, or frustration you never thought to include in the brief. Build longer relationships with the creators who bring customers who keep the product, and get permission to use effective demos in paid ads. Count the whole cost, including fees, commissions, free units, and content production. Judge the partnership by the customers it brings in and what you learn from them. 8. Make the first production run earn the next one. Before placing a much larger order, put a small batch from the factory in customers’ hands. There’s a lot to learn between a prototype working on your desk and someone using the product at home without your help. Watch setup problems, failures, returns, continued use, and the time support spends on each customer. Feed those lessons into the next order. Think about repairs while the design is still flexible. Saving a few dollars on a part can get expensive when replacing it means shipping the entire device across the country. By the time you order the next batch, you should have a better understanding of who buys, why they keep using it, what breaks, and what each sale leaves in the bank. Hardware is not for the weak.

F.02 Decommission

Had a great time chatting with some amazing hardware teams at @workatthestudio with @ngpadgett If your working in hardware come by! Places like these remind me what SF is really about

Pittsburgh is a hardware mecca. 300k people, 250+ robotics and AI companies, and a conference floor with not a chatbot in sight. @ngpadgett 's trip report from Robotics & AI Discovery Day, plus why Pittsburgh Hardware Meetup is coming back this year: informal.com/blog/pittsburgh…

Your BOM has a memory problem you haven't priced yet. AI data centers are buying up the world's memory chips and the cheap parts in your product are what gets squeezed. If you need EE / Supply Chain support on tap, that's exactly why we exist. Reach out! tinyurl.com/2fv4yx86

Who's going to @RoboPGH's Robotics & AI Discovery Day in Pittsburgh next week? I bought a bunch of drink tickets to their Happy Hour for the Hardware Meetup community. Come hang! luma.com/nubjcg0i

Founders if you’re having trouble raising switch to making money

I've had the same conversation with 10+ early-stage founders in the last few months. Everyone wants to raise a crazy round because a friend or competitor did, VCs will happily fund it, and they've been conditioned to believe anything less is failure. As an industry, we've lost sight of what hype and capital are for. They're supposed to be a means, not the end. Instead, we’ve built a reward structure that pays for signaling value rather than creating it. My latest on status games, kamikaze rounds, and a colonel who never made general because he chose to do something instead of be somebody — and made a bigger mark for it.

A high-end injection mold can cost $400,000-$500,000, with dozens of cavities, hydraulic slides, lifters, ejectors, valve gates and hot runners packed inside hardened H13 or S136 steel. Some molds carry 128 cavities, producing 128 parts per cycle, while precision machining can hold tolerances around ±0.005 mm. Cooling circuits circulate water through channels inside the steel, with temperature control around 1-2°C, cycle after cycle. These aren't designed for a few thousand parts. High-end production tooling can be engineered for 1-2 million+ shots, operating through repeated injection pressure, heating, cooling and mechanical movement without losing much alignment. The industrial ecosystem required to make them is enormous, CNC machining, EDM, grinding, polishing, heat treatment, metrology, mold flow simulation, thermal management and hot-runner engineering all have to work together. And China has become the centre of gravity for this industry. China represented nearly 17.8% of the global injection molding market, while some industry estimate puts Chinese manufacturers at roughly 68% of global plastic injection mold supply. These are the machines that make the machines. For any country trying to build a serious manufacturing base, the ability to produce tooling this precise and complex is fundamental. These molds sit at the core of thousands of factories, determining how cheaply, quickly and reliably everything from cars to medical devices can be manufactured at scale.