Revival research

    Dead startup ideas worth reviving in 2026

    24 companies that burned $4B+ between them — and the reason each one would be a different business if it launched today. Drawn from our database of 1,091 documented failures.

    Last reviewed Next review August 22, 2027
    TL;DR • dead startup ideas worth reviving • as of August 2026

    A dead startup idea is worth reviving only when the thing that killed it has changed — not when the market simply got bigger. Across 1,091 documented shutdowns, 24 qualify in 2026: mostly companies that died paying humans for work models now do (Olive AI, Munchery), or that had to build infrastructure you can now rent by the hour (Sprig, Clinkle, Better Place). The revival almost never repeats the original product — it attacks the same demand from a narrower, cheaper wedge.

    The test we applied

    Every candidate had to pass one question: has the recorded cause of death actually changed? Companies that died of fraud, governance collapse or founder misconduct were excluded — nothing about 2026 fixes those. What remains falls into six drivers.

    AI capability

    The hard part was a human-labour cost or an accuracy ceiling that models now clear.

    Cost collapse

    The input that made the original unit economics impossible got dramatically cheaper.

    Infrastructure

    The rails the company had to build itself now exist as commodity services.

    Behaviour shift

    Users were asked to change a habit they were not ready for. They are now.

    Regulation

    A legal barrier that killed the business has been removed, narrowed, or inverted.

    Incumbent retreat

    The giant that crushed the original has abandoned, degraded, or over-priced the category.

    The 24 ideas

    Tutorspree · died 2013

    A marketplace matching students with vetted local tutors.

    $1.8M burned
    Behaviour shift

    Why it died

    Acquisition never worked — the company depended on a single channel — and matched pairs left the platform to transact directly.

    What changed by 2026

    Tutoring went remote and asynchronous, which removes the geographic constraint that made Tutorspree's liquidity so hard, and payment-plus-scheduling lock-in is now a normal expectation rather than a friction.

    The revival wedge

    Serve the tutors, not the students: booking, payment, progress reporting and parent updates for people who already have a client list. Disintermediation stops being a threat when the tutor is the customer.

    Still dangerous: Education marketplaces have brutal seasonality and price-sensitive buyers.
    Capital to first revenue: Under $10k
    Revival confidence 4/5
    Full Tutorspree post-mortem Validate this wedge

    Doppler Labs · died 2017

    Programmable earbuds that let you filter, tune and augment real-world sound in real time.

    $51M burned
    Incumbent retreat

    Why it died

    Ahead of the market and out-distributed by AirPods. Doppler had to design custom silicon, its own DSP stack and a hardware supply chain to ship a single product.

    What changed by 2026

    Hearing augmentation moved from a moonshot to a regulated consumer category: OTC hearing aids are legal in the US, and mainstream earbuds now ship live-listen and conversation-boost modes — proving the demand Doppler could not prove.

    The revival wedge

    Do not build the hardware. Build the tuning software layer for a specific hostile environment — restaurants, factory floors, live music — on top of earbuds people already own.

    Still dangerous: Platform dependency. If Apple or Samsung ship your feature natively, your product becomes a setting in someone else's OS.
    Capital to first revenue: $10k–50k
    Revival confidence 4/5
    Full Doppler Labs post-mortem Validate this wedge

    Olive AI · died 2023

    An "AI employee" automating hospital back-office work — eligibility, claims, prior authorisation.

    $902M burned
    AI capability

    Why it died

    Overpromised RPA as AI. The bots were brittle scripts that broke on every payer-portal change, and hospitals could not measure the ROI.

    What changed by 2026

    The gap between what Olive sold and what the technology could do is precisely the gap that language models closed: unstructured documents, payer letters and clinical notes are now machine-readable.

    The revival wedge

    Pick one revenue-cycle task with a measurable dollar outcome — denial appeals, for instance — and price on recovered revenue rather than seats, so the ROI question answers itself.

    Still dangerous: Healthcare sales cycles and the credibility damage Olive did to the category. Buyers now demand proof before pilots.
    Capital to first revenue: $10k–50k
    Revival confidence 4/5
    Full Olive AI post-mortem Validate this wedge

    Jibo · died 2019

    A social home robot with personality that families would talk to and live with.

    $73M burned
    AI capability

    Why it died

    A three-year shipping delay, an $899 price point, and a conversational engine that could not hold a thread — undercut by $50 smart speakers.

    What changed by 2026

    The conversational layer Jibo tried to hand-author is now a commodity API. The failure was never the concept of a companion device; it was that the dialogue was scripted and the hardware carried the whole cost.

    The revival wedge

    Target a population where companionship has a payer, not a hobbyist: elder-care check-ins with medication prompts and a family dashboard, sold to home-care agencies rather than to consumers at $899.

    Still dangerous: Hardware margins and support costs. If your revival needs a bespoke device to work, you have rebuilt Jibo.
    Capital to first revenue: $50k–250k
    Revival confidence 4/5
    Full Jibo post-mortem Validate this wedge

    Secret · died 2015

    Anonymous posting inside your real social graph.

    $35M burned
    AI capability

    Why it died

    Anonymity turned toxic fast. Without identity, moderation could not keep up and the network poisoned itself within months.

    What changed by 2026

    Real-time moderation of unstructured text is now automatable at a cost per message that a small team can afford — the exact capability whose absence killed Secret and Yik Yak.

    The revival wedge

    Anonymity with a verified boundary: pseudonymous inside a closed, verified group — a company, a hospital, a school district — where the identity check happens at the door and speech is free inside it.

    Still dangerous: Toxicity is a product design problem, not only a filtering problem. Moderation cost was the symptom; the incentive to be cruel was the disease.
    Capital to first revenue: Under $10k
    Revival confidence 3/5
    Full Secret post-mortem Validate this wedge

    Yik Yak · died 2017

    A hyperlocal anonymous feed of everyone within a few hundred metres.

    $73.5M burned
    AI capability

    Why it died

    Toxic content drove users away, and the panic pivot to real identity destroyed the only reason to use it.

    What changed by 2026

    Same shift as Secret: moderation is now tractable. And hyperlocal demand did not disappear — it migrated to group chats with no discovery layer.

    The revival wedge

    Hyperlocal with a purpose beyond gossip: neighbourhood logistics, campus marketplaces, event-scoped feeds that expire — a reason to post that is not social status.

    Still dangerous: Cold-start density. A local network with 40 users is a dead app, and you need density in every location separately.
    Capital to first revenue: Under $10k
    Revival confidence 3/5
    Full Yik Yak post-mortem Validate this wedge

    Rdio · died 2015

    A streaming music service built around social discovery and taste, not playlists alone.

    $125M burned
    Incumbent retreat

    Why it died

    Outmarketed by Spotify. Rdio was widely considered the better product and lost on distribution and licensing leverage.

    What changed by 2026

    The discovery layer has separated from the catalogue: streaming APIs let you build taste and curation experiences without licensing a single track.

    The revival wedge

    Build the taste layer on top of the incumbents' catalogues rather than competing with them — human and model curation as a product, playback delegated.

    Still dangerous: You live at the mercy of platform API terms, which have been narrowed repeatedly.
    Capital to first revenue: Under $10k
    Revival confidence 3/5
    Full Rdio post-mortem Validate this wedge

    Homejoy · died 2015

    On-demand home cleaning booked in two taps, with the platform handling supply.

    $40M burned
    Regulation

    Why it died

    Bad unit economics plus worker-misclassification lawsuits. Discounted acquisition brought customers who never rebooked at full price.

    What changed by 2026

    The classification question that killed Homejoy has been litigated and legislated into a set of known answers across most markets. It is now a compliance cost you can model up front, rather than an unbounded legal risk.

    The revival wedge

    Sell to the cleaning companies, not around them: scheduling, routing, payments and compliance software for existing crews. You capture the category without owning the labour risk.

    Still dangerous: Retention. Homejoy's deeper problem was that home cleaning is a relationship, and once matched, the customer and cleaner have every reason to leave the platform.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Homejoy post-mortem Validate this wedge

    Shyp · died 2018

    Photograph anything, and a courier collects, packs and ships it for a flat fee.

    $62M burned
    Behaviour shift

    Why it died

    The flat fee never covered the pickup. Every order consumed a human courier and a packing station; volume made the losses larger, not smaller.

    What changed by 2026

    Consumer resale volume exploded, and drop-off networks now sit inside pharmacies, lockers and grocery stores — the pickup leg that Shyp had to pay for has been absorbed by infrastructure that already exists.

    The revival wedge

    Attack the packing and label decision, not the transport: an AI-priced "photograph it, we quote and label it" layer for resale sellers, using existing drop-off networks for the physical leg.

    Still dangerous: Margin per parcel is thin and carriers control it. You are a feature on someone else's rate card unless you own the seller relationship.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Shyp post-mortem Validate this wedge

    Munchery · died 2019

    Chef-prepared meals cooked centrally and delivered to homes on a subscription.

    $125M burned
    AI capability

    Why it died

    Poor unit economics and massive food waste — production was committed before demand was known.

    What changed by 2026

    Demand forecasting on perishable inventory went from a project to an off-the-shelf capability. Waste, the line item that killed Munchery, is now the most tractable part of the model.

    The revival wedge

    Sell the forecasting and dynamic-markdown layer to existing meal producers and grocers first. Prove the waste reduction on someone else's inventory before you buy a kitchen.

    Still dangerous: Consumer meal subscriptions still churn hard after month three, whatever the food costs.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Munchery post-mortem Validate this wedge

    Clinkle · died 2017

    Phone-to-phone payments without cards or terminals.

    $30M burned
    Infrastructure

    Why it died

    The product never meaningfully shipped. The company raised on a demo and spent years building rails that did not exist.

    What changed by 2026

    Every rail Clinkle tried to build — tokenisation, tap-to-phone acceptance, instant account-to-account transfer — is now an API or a native OS capability.

    The revival wedge

    Payments as a feature inside a vertical workflow, not a wallet. Nobody downloads a wallet; they use the app that already books the job.

    Still dangerous: Consumer payment habits are the stickiest in software. Competing on the wallet itself is still unwinnable.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Clinkle post-mortem Validate this wedge

    Primary Data · died 2018

    A metadata layer that moves data automatically to the right storage tier across vendors.

    $100M burned
    Infrastructure

    Why it died

    Technical overreach and market timing — enterprises had not yet spread data across enough places for the problem to hurt.

    What changed by 2026

    Enterprise data is now genuinely scattered across clouds, regions and vendors, and egress costs and AI training pipelines made placement a line item that finance teams can see.

    The revival wedge

    Start as a read-only cost and placement analyser — show the bill before you ask to move a single byte.

    Still dangerous: Storage vendors bundle this, and enterprise buyers do not let a startup move production data.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Primary Data post-mortem Validate this wedge

    Washio · died 2016

    On-demand laundry pickup and delivery within 24 hours.

    $17M burned
    Infrastructure

    Why it died

    The on-demand model never covered the cost of two human trips per order at consumer price points.

    What changed by 2026

    Route density is now buyable: aggregating laundry onto existing multi-stop delivery routes removes the dedicated-courier cost that defined Washio's loss per order.

    The revival wedge

    Sell into buildings, not to individuals — a residential or hotel contract gives you a fixed pickup point and predictable volume instead of scattered one-off trips.

    Still dangerous: This is still a logistics business with a laundry margin. Density is the only thing that makes it work.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Washio post-mortem Validate this wedge

    Segway · died 2020

    Personal electric transport that would reshape how people move through cities.

    $100M+ burned
    Regulation

    Why it died

    A hype-versus-reality gap: a $5,000 price, no regulatory home on sidewalk or road, and a form factor people found socially awkward.

    What changed by 2026

    Micromobility got legal categories, bike lanes and mass adoption. Segway was right about the demand and wrong about the price, the shape and the decade.

    The revival wedge

    The unresolved part is not the vehicle, it is what happens to it: fleet maintenance, battery health and end-of-life economics — the exact line item that killed Bird.

    Still dangerous: Hardware micromobility remains a capital-destroying category. Sell tooling to fleets rather than owning vehicles.
    Capital to first revenue: $10k–50k
    Revival confidence 3/5
    Full Segway post-mortem Validate this wedge

    Sprig · died 2017

    Vertically integrated meal delivery: own the kitchen, own the menu, deliver in 15 minutes.

    $57M burned
    Infrastructure

    Why it died

    Owning kitchens, chefs and couriers at a $12 price point meant the company paid for the whole stack on every order.

    What changed by 2026

    Every layer Sprig had to own — kitchen space, delivery fleet, demand — is now rentable by the hour through commissary kitchens and third-party logistics.

    The revival wedge

    Run a single-menu, single-daypart brand out of rented commissary capacity for a captive audience with predictable demand — hospitals, campuses, large offices — where you can forecast covers instead of guessing.

    Still dangerous: Food waste and the 30% marketplace take rate. Sprig's arithmetic only works if you own demand directly.
    Capital to first revenue: $50k–250k
    Revival confidence 3/5
    Full Sprig post-mortem Validate this wedge

    Napster (1999) · died 2002

    Peer-to-peer distribution that made a global catalogue instantly available.

    $130M burned
    Regulation

    Why it died

    Legal injunctions. The distribution was brilliant and the rights position was indefensible.

    What changed by 2026

    The licensing framework Napster refused to build is now standard, and the same "everything, instantly, for a flat fee" promise is the dominant model in music. The idea won; the company did not.

    The revival wedge

    Apply the pattern to a catalogue that is still fragmented and un-licensed at scale — training data, academic material, archival footage — and build the rights layer first, not last.

    Still dangerous: Rights-holder litigation remains the fastest way to zero. Licence before you launch.
    Capital to first revenue: $50k–250k
    Revival confidence 3/5
    Full Napster (1999) post-mortem Validate this wedge

    Quirky · died 2015

    Crowd-sourced invention: the community submits products, votes, and shares the royalties.

    $185.3M burned
    Cost collapse

    Why it died

    Burned cash and did not sell enough products. Quirky committed to tooling, manufacturing and retail distribution for far too many SKUs at once.

    What changed by 2026

    The expensive half — industrial design, CAD, renders, packaging, listing copy — is now generatable, and small-batch manufacturing plus marketplace distribution removed the retail commitment Quirky took on.

    The revival wedge

    Community invention with no inventory: validate demand with pre-orders and only tool the one SKU that clears a threshold.

    Still dangerous: Physical products have long cash-conversion cycles. One tooling mistake still eats a year of runway.
    Capital to first revenue: $50k–250k
    Revival confidence 3/5
    Full Quirky post-mortem Validate this wedge

    Iron Ox · died 2023

    Robotic indoor farms growing produce closer to the consumer at grocery prices.

    $98M burned
    Cost collapse

    Why it died

    Robotic greenhouse capex could not beat commodity produce prices — the robots were amortised against lettuce.

    What changed by 2026

    Robot arms, vision stacks and controlled-environment components have commoditised, and the crop mix has moved on: the economics work for high-value plants, not staples.

    The revival wedge

    Same technology, different crop. Pharmaceutical botanicals, seedlings and propagation, or research-grade plant material — where a kilo is worth hundreds, not a few dollars.

    Still dangerous: Energy cost per kilo. If your revival still competes with an outdoor field and free sunlight, the arithmetic has not changed.
    Capital to first revenue: $250k–1M
    Revival confidence 3/5
    Full Iron Ox post-mortem Validate this wedge

    Zume · died 2023

    Robots cooking pizza in the delivery truck so it finishes baking as it arrives.

    $500M+ burned
    AI capability

    Why it died

    The robots never worked reliably at speed, and the company pivoted through several unrelated businesses while burning half a billion dollars.

    What changed by 2026

    Food robotics narrowed its ambition and got good: single-task fryer, wok and assembly units now run commercially in kitchens. The failure was the truck, not automation.

    The revival wedge

    One station, one motion, sold as equipment leasing to existing restaurant operators with a labour shortage — not a new consumer brand.

    Still dangerous: Restaurant operators buy on payback period. If you cannot show it under 18 months, no kitchen installs it.
    Capital to first revenue: $250k–1M
    Revival confidence 3/5
    Full Zume post-mortem Validate this wedge

    Better Place · died 2013

    Battery-swap stations so EV drivers exchange a depleted pack in minutes instead of charging.

    $850M burned
    Cost collapse

    Why it died

    Wrong timing and over-ambition: it required a national station network and carmaker-standardised packs before there were enough EVs to use either.

    What changed by 2026

    EV penetration is no longer the bottleneck, battery costs have fallen by roughly an order of magnitude since 2013, and swap networks operate commercially in China — the model has been demonstrated, just not in the West.

    The revival wedge

    Forget consumer cars. Swap works where the fleet is standardised and downtime is expensive: delivery vans, two-wheelers, port and warehouse equipment on fixed routes.

    Still dangerous: Capex and standardisation. Better Place died of both, and fast charging keeps improving underneath you.
    Capital to first revenue: $1M+
    Revival confidence 3/5
    Full Better Place post-mortem Validate this wedge

    Path · died 2018

    A deliberately small social network capped at your 50 closest people.

    $74M burned
    Behaviour shift

    Why it died

    Could not grow against Facebook and Instagram, and a contact-upload scandal cost it trust at the worst moment.

    What changed by 2026

    Public posting collapsed. Sharing moved into group chats and close-friends lists — users adopted Path's thesis, just inside other people's products.

    The revival wedge

    Small-group sharing built around an artefact rather than a feed: a family archive, a trip, a child's first years — something with a reason to return that is not vanity metrics.

    Still dangerous: Social products need network effects, and messaging apps already hold the small group. You need a job the group chat does badly.
    Capital to first revenue: Under $10k
    Revival confidence 2/5
    Full Path post-mortem Validate this wedge

    Vine · died 2017

    Six-second looping video as a native creative format.

    ~$30M (acquisition) burned
    Behaviour shift

    Why it died

    Twitter neglect and no creator monetisation. Vine's biggest stars left for platforms that paid them.

    What changed by 2026

    Short-form looping video became the dominant consumption format on the internet. Vine was right about the format and wrong about paying the people who made it.

    The revival wedge

    Do not rebuild a general short-video app. Build the payout and rights infrastructure for creators inside a vertical the big platforms monetise badly — education, trades, clinical training.

    Still dangerous: You are competing for attention against recommendation engines with a decade of head start.
    Capital to first revenue: $10k–50k
    Revival confidence 2/5
    Full Vine post-mortem Validate this wedge

    Juicero · died 2017

    A connected appliance plus a subscription of pre-portioned, traceable fresh produce packs.

    $118.5M burned
    Behaviour shift

    Why it died

    A solution in search of a problem: a $400 press whose packs could be squeezed by hand. The appliance added cost, not capability.

    What changed by 2026

    The subscription half of the model was never the failure — pre-portioned fresh-food subscriptions are now a normal grocery behaviour. The hardware was.

    The revival wedge

    Ship the consumable subscription with zero hardware, in a category where portioning genuinely is the pain: clinical or post-operative nutrition where dosage accuracy is the product.

    Still dangerous: Cold-chain logistics and churn. Fresh subscriptions die on delivery cost, exactly as the food-delivery cohort below did.
    Capital to first revenue: $50k–250k
    Revival confidence 2/5
    Full Juicero post-mortem Validate this wedge

    Aereo · died 2014

    Rent a tiny antenna in a datacentre and stream your local broadcast channels to any device.

    $97M burned
    Regulation

    Why it died

    A Supreme Court ruling killed it overnight — the court held the service was a public performance requiring retransmission licences.

    What changed by 2026

    The legal barrier remains for Aereo's exact architecture, but the market moved around it: broadcasters now license direct-to-consumer streaming, and free ad-supported channels are a mainstream distribution tier.

    The revival wedge

    Do the licensed version of the same job — aggregation and discovery across free, ad-supported and broadcast feeds — rather than the unlicensed workaround.

    Still dangerous: Content licensing is the whole business, and rights-holders set the terms. This is the one entry where the original cause of death is only partly gone.
    Capital to first revenue: $50k–250k
    Revival confidence 2/5
    Full Aereo post-mortem Validate this wedge

    Every idea at a glance

    Company Died Burned Why it's back Capital needed Confidence
    Tutorspree 2013 $1.8M Behaviour shift Under $10k 4/5
    Doppler Labs 2017 $51M Incumbent retreat $10k–50k 4/5
    Olive AI 2023 $902M AI capability $10k–50k 4/5
    Jibo 2019 $73M AI capability $50k–250k 4/5
    Secret 2015 $35M AI capability Under $10k 3/5
    Yik Yak 2017 $73.5M AI capability Under $10k 3/5
    Rdio 2015 $125M Incumbent retreat Under $10k 3/5
    Homejoy 2015 $40M Regulation $10k–50k 3/5
    Shyp 2018 $62M Behaviour shift $10k–50k 3/5
    Munchery 2019 $125M AI capability $10k–50k 3/5
    Clinkle 2017 $30M Infrastructure $10k–50k 3/5
    Primary Data 2018 $100M Infrastructure $10k–50k 3/5
    Washio 2016 $17M Infrastructure $10k–50k 3/5
    Segway 2020 $100M+ Regulation $10k–50k 3/5
    Sprig 2017 $57M Infrastructure $50k–250k 3/5
    Napster (1999) 2002 $130M Regulation $50k–250k 3/5
    Quirky 2015 $185.3M Cost collapse $50k–250k 3/5
    Iron Ox 2023 $98M Cost collapse $250k–1M 3/5
    Zume 2023 $500M+ AI capability $250k–1M 3/5
    Better Place 2013 $850M Cost collapse $1M+ 3/5
    Path 2018 $74M Behaviour shift Under $10k 2/5
    Vine 2017 ~$30M (acquisition) Behaviour shift $10k–50k 2/5
    Juicero 2017 $118.5M Behaviour shift $50k–250k 2/5
    Aereo 2014 $97M Regulation $50k–250k 2/5

    Reviving one of these? Check it against the graveyard first

    Our validator scores your version of the idea against 1,091 documented failures, the open-source projects that may already have built it, and verified pricing from comparable products — so you find out whether the cause of death really changed before you commit a year to it.

    Frequently asked questions

    What makes a dead startup idea worth reviving?

    One test only: the specific cause of death must have changed. A bigger market is not a revival thesis — if the company died of delivery costs and delivery still costs the same, the idea is not back. Every case on this page names the original cause, the concrete shift since the shutdown, and the part of the risk that is still unresolved.

    Why do failed startup ideas succeed the second time?

    Usually because someone else paid for the infrastructure. Sprig had to build kitchens, fleets and demand; today all three are rentable. Olive AI hand-scripted document handling that language models now do natively. The second entrant inherits the rails the first one had to fund.

    Is it safer to revive a failed idea than to start a new one?

    It is better informed, not safer. You get a documented post-mortem, a known buyer, and evidence that demand existed. You also inherit any structural problem that was never about timing — thin margins, disintermediation, or platform dependency — which is why each entry here lists what is still dangerous.

    How were these 24 ideas selected?

    From the IdeaProof failure database of 1,700+ documented shutdowns. We kept only companies whose recorded cause of death maps to a driver that has measurably changed since — cost collapse, AI capability, behaviour shift, regulation, infrastructure, or incumbent retreat — and discarded cases that failed for reasons still fully in force, such as fraud or governance collapse.

    How much capital do you need to revive one of these?

    It depends entirely on the wedge. The software re-entries on this list — curation layers, vertical workflow tools, forecasting for existing operators — can reach first revenue under $50k. Hardware and fleet revivals such as battery swap or robotic growing still need seven figures before the first customer.

    Keep researching